SanDisk Aktie

Good afternoon, and welcome to SanDisk's First Quarter Fiscal Year 2027 Earnings Conference Call.

[Operator Instructions]

Please also note, today's event is being recorded. At this time, I'd like to turn the floor over to Ivan Donaldson, Vice President of Investor Relations. Please go ahead.

Before we begin, please note that today's discussion will contain forward-looking statements based on management's current assumptions and expectations, which are subject to various risks and uncertainties. These forward-looking statements include expectations for our technology and product portfolio, our business plans and performance, our capital allocation priorities, market trends and opportunities and our financial results. We assume no obligation to update these statements. Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations.

We will also make references to non-GAAP financial measures today. Reconciliations between the non-GAAP and comparable GAAP financial measures are included in the written materials posted in the Investor Relations section of our website.

With that, I'll turn the call over to David.

Thanks, Ivan. Good afternoon, and thank you for joining SanDisk's Fiscal First Quarter 2027 Earnings Call. We began fiscal year 2027 with another strong quarter, creating value for customers and shareholders and delivering results above guidance.

Our strategy is working and our execution is delivering strong results. Revenue for the quarter was $11 billion. Non-GAAP gross margin was 85.1% and non-GAAP earnings per share was $49.51. We generated $5.7 billion of adjusted free cash flow and returned $5 billion to shareholders through share repurchases. Our confidence in the long-term trajectory of the business is grounded in 3 enduring strengths: technology leadership, deep customer partnerships and disciplined financial management.

These strengths are reinforced by a structural increase in NAND demand driven particularly by AI and by customer signals now extending beyond fiscal year 2030 through our new business model, our NBM partnerships. Together, they position us to invest with confidence in our technology road map, strengthen and expand our customer relationships, and deliver growing and durable free cash flow while returning capital to shareholders. It starts with our technology leadership. BiCS 8 is the backbone of our current production portfolio in the industry benchmark for performance and density across TLC and QLC. BiCS 8 introduced our CMOS directly bonded to Array, our CBA architecture, enabling the performance required by AI data center customers in establishing the framework for future generations. BiCS 9 pairs the proven BiCS 8 cell array with next-generation CMOS significantly increasing interface performance while retaining a mature memory architecture. BiCS 10 10 advances the full node architecture across both TLC and QLC with a 1 terabit TLC die designed for performance-intensive workloads in a 2-terabit QLC die delivering industry-leading density in twice the read and right bandwidth of BiCS 8. Together, these technologies will extend the BiCS 8 foundation to meet the evolving performance and capacity requirements of next-generation infrastructure. Beyond the core road map, high-bandwidth flash, or HBF, reached an important milestone in the fiscal first quarter with the tape-out of our first HBF memory die with fabrication now underway.

As we outlined at Investor Day, HBF is a new memory architecture for inference that targets HBM class read bandwidth with substantially greater capacity to support larger models, longer context lengths and growing KV caches. It is designed to complement HBM by bringing NAND class capacity closer to the processor, and we remain on track to deliver samples of our first HBF-based products next calendar year. HBF remains a future innovation and is not included in the long-term financial model we introduced at Investor Day. As customer adoption, product requirements and manufacturing readiness become more clearly defined, we expect to provide further perspective on the market opportunity. Customers recognize the strength of our innovation and increasingly view SanDisk as an important partner in achieving their growth objectives.

Through our NBMs, we continue to evolve customer relationships into long-term partnerships grounded in innovation, collaboration and shared commitment. Our most recent customers' engagements with the world's leading AI companies indicate further upside in demand for our technology. Our market-leading technology product portfolio and customer partnerships are the backbone of our strong financial position, which allows us to reinvest to maintain our technology leadership and expand supply through R&D-focused CapEx. Each of these strengths is reinforced by tailwinds from AI, the need for scalable memory and storage is expanding as inference workloads grow, making NAND increasingly integral to AI infrastructure. During inference, each generated token must reference the context that precedes it, creating a growing KV cache that serves as the model's working memory. Longer prompts, more complex reasoning, multimodal inputs and persistent user context materially increase these storage requirements. HBM and DRAM support the highest performance tier, while flash provides the capacity required to retain broader context efficiently and avoid costly recomputation. We believe this is a secular dynamic that is expanding NAND's role across the AI stack. The memory hierarchy defines the role of each NAND architecture.

TLC provides the lower latency, higher rate performance and greater endurance required for active data sets, check-pointing and latency-sensitive KV cache workloads. In our AI lab comparison, we presented at Investor Day, systems utilizing SSD delivered approximately 3x the token throughput while using approximately 75% less energy than comparable systems relying on volatile memory alone.

QLC further increases storage density for capacity-intensive applications, such as data lakes, vector databases that power AI search and retrieval repositories. Between these layers, a growing persistent context tiers, establishes data to be retained and recalled rather than regenerated. As inference costs declined, we expect the adoption and usage to expand, supporting our view that KV cache will represent approximately 1/3 of the AI data center flash market by the end of the decade and that the AI data center opportunity will approach the size of the NAND industry today.

Turning to our end markets. In data center, our portfolio addresses the full range of AI infrastructure requirements from performance-intensive enterprise SSD to high-capacity platforms. Demand remains strong and continues to grow, and we estimate that data center has expanded from approximately 20% of industry bids early in a decade to roughly half of the industry bits today. This structural shift in customer mix is reshaping industry economics and is central to the transformation underway at Sandisk. This quarter, data center again delivered strong growth and continued to outpace the broader market as we allocated more supply to address the growing demand from this end market. Edge is a large and strategically important market for Sandisk, spanning smartphones, PCs, automotive tablets and an expanding range of physical AI applications.

Automotive is becoming an increasingly important growth vector in our Edge business and a meaningful long-term growth opportunity. We have deepening engagements with leading customers reinforcing our confidence that automotive can become one of the largest markets within edge over time. Our consumer business remains a differentiated strategic asset, combining direct brand engagement with hundreds of millions of end users and broad shelf presence across the world's leading channel partners. The consumer end market provides a durable foundation for our business. while its direct connection to the market informs product development and provides early insight into shifts in end user demand.

Taken together, these results reinforce the durability of the model we outlined at our Investor Day. We have established a strong level of profitability and cash generation, and our focus from here is consistent execution across the portfolio.

With that, I'll turn the call over to Luis for a detailed update on our business and outlook.

Thank you, David. We're executing these strategies and plans that we outlined at the Investor Day, and we're delivering attractive financial results. In the first quarter, we closed 2 NBMs because. Both deals closed before our fourth quarter earnings call, we announced these partnerships on that call. We're in active conversations with existing [indiscernible] customers that are evaluating extending existing agreements and with new customers that are evaluating new NBMs. These long-term partnerships provide greater visibility into demand and financial performance, supporting a more durable financial model.

We will remain patient and disciplined in selecting the right long-term partners as we define the optimal percentage of our bids that we allocate to NBMs, which is slightly higher than our current levels. Revenue for the first quarter was $11 billion, up 23% sequentially and 378% year-over-year. This compares favorably with our guidance range of $10.3 billion to $10.8 billion. Both volume growth and higher pricing contributed to revenue growth, with approximately 60% of the increase attributable to volume growth and 40% to higher pricing. Data center revenue was $4.8 billion, up 61% sequentially, outpacing estimated market growth for the quarter. Our leading products and strong customer partnerships including 3 U.S. hyperscalers underpin our continued expansion in this fast-growing end market. Edge revenue was $5.8 billion, up 7% sequentially, reflecting resilience in premium and mainstream markets, partially offset by contraction at the lower end. Within Edge, we saw strong growth with automotive and IoT customers.

Consumer revenue was $397 million, down 29% sequentially as the market adjusts to higher pricing. We remain committed to serving all 3 end markets with our diversified portfolio that optimizes our performance over time. Non-GAAP gross margin for the first quarter was 85.1%, up from 84.6% in the prior quarter, driven by higher volumes and pricing. This compares favorably with our guidance range of 83% to 85%. Non-GAAP operating expenses were $513 million, representing 4.7% of revenue.

Non-GAAP operating margin was 80.4% compared to 79.2% in the prior quarter. Non-GAAP EPS was $49.51, up from $39.25 in the prior quarter and $1.22 in the same quarter of the prior fiscal year. This compares favorably with our guidance range of $44 to $46. We closed the quarter with 154 million diluted shares outstanding.

Key GAAP to non-GAAP reconciliation includes $91 million in stock-based compensation, representing 0.8% of revenue and $513 million gain on equity securities primarily attributable to our investment in Nanya. During the quarter, we repurchased approximately 3.2 million shares of our common stock for $5 billion. Over the past 2 quarters, we have repurchased 6 million shares, representing approximately 4% of our total shares outstanding. Our repurchases reflect our conviction in the long-term value of the business and our disciplined approach to capital allocation. We continue to invest appropriately in the business while returning 100% of excess cash to shareholders.

Moving on to free cash flow. Cash flow from operations was $6.4 billion. Gross capital expenditures totaled $378 million, representing 3.4% of revenue. We expect gross capital spending to increase during the balance of the year and represent mid-single-digit percentage of revenue for fiscal year 2027. During the quarter, we generated $5.7 billion in adjusted free cash flow, representing approximately 52% margin. We closed the quarter with $5.1 billion in cash and cash equivalents and no debt.

Moving on to guidance. Our confidence in the long-term health of the market is reinforced by the structural growth of our -- of the AI infrastructure and the expanding role of NAND across inference workloads. Before turning to our second quarter outlook, I will announce that the Sandisk Board of Directors has approved an [ 8-for-1 ] stock split of the company's common stock. The additional shares are expected to be distributed on November 19 and trading on a split-adjusted basis is expected to begin on November 20.

For the second quarter of fiscal year 2027, we expect revenue between $12.2 billion and $12.6 billion. We expect non-GAAP gross margin between 84.5% and 85.5%, which includes modest like-for-like price increases. On a pre-split basis, we expect non-GAAP EPS between $55 and $57, assuming 151 million fully diluted shares.

On an 8-for-1 split-adjusted basis, this equates to non-GAAP EPS between $6.88 and $7.13 and assuming approximately 1,208 million fully diluted shares. For context, we expect a strong market in calendar year 2027. Data center is expected to be, again, the engine of growth and represent over 55% of the market, with bit demand growing about 40% year-on-year enabled by higher AI-driven capital spending. This is after more than doubling in 2026. Our NBM pipeline exceeds our supply throughout the end of the decade. With that, let me turn the call back to David.

Thank you, Luis. We continue to execute on the strategy we outlined at Investor Day, reinforcing our confidence in Sandisk's long-term trajectory. Our technology leadership, deep customer partnership and disciplined financial management, supported by a structural increase in NAND demand driven by AI, position us to serve our customers, deliver growing and durable free cash flow and return capital to shareholders. Over the past year, we have transformed the earnings and cash generation power of the business and established a strong foundation for durable performance. We remain focused on disciplined execution, sustained value creation and delivering against our long-term financial objectives.

With that, Ivan, let's open it up for questions.

[Operator Instructions]

Our first question today comes from Ben Reitzes from Melius.

I appreciate it, Dave. I wanted to ask two, if I could. One on the NBMs. You signed 2 before the last call. So I'm just wondering if we're going to have a lot more of them because it seems like you're already capacity constrained to agree to them. And then do you mind reminding us what the bits under contract will be in the next few years, if that's changed at all? And then I have a quick follow-up.

Yes, I'll make a few -- thanks, Ben. We appreciate it. Always good to hear from you. So I'll make a few comments on NBMs and then Luis will fill in. Yes, we've got a very strong pipeline of NBMs and you kind of hit on all the issues there. As the portfolio gets more subscribed and more committed and we're at half the bits for this fiscal year and 2/3s for next year. It's a fitting exercise to get the next tranche of them in there, but the pipeline is very strong. I think we said last quarter -- I know we said last quarter that we're going to be patient and consider the next set of customers. And that's exactly what we're doing.

And we expect more of them over the next several months, several quarters as we continue to allocate the rest of the portfolio. So the pipeline is very strong and growing, quite frankly, and it's become the predominant way we deal with our customers. And I expect that the percent of the portfolio committed to continue to go up over the next couple of quarters. So Luis?

Yes. I mean just a few things I think David covered most of it. We feel very good about where we are with NBMs. As you know, we've been very selective, very strategic in who we chose to sign these deals with. We're looking at the companies we have and today, 95% of our NBMs -- sorry, with companies that are investment grade, and they generated about $130 billion in operating cash flow. So we feel that we've made the right choices. And as David just mentioned, we'll continue to be patient.

We'll continue to be selective and the pipeline is terrific. So you should expect us to make -- continue to make progress. I think in my prepared remarks, I said that we expect the bids to continue to grow we had shared with you that about 50% of our bids are under NBM for '27 and about 2/3s for '28, and we expect that number to continue to go a little bit higher from the 2/3s.

Ben, I'll just make one more comment. You got to ask two questions. I'll make 2 comments. We're also balancing customers that are coming back for more like that is a real thing. And in general, we're hearing from our customers that they have underestimated the amount of NAND they need to run their infrastructure. And so we have multiple customers coming back for a second round, and then balancing that against bringing new customers into the portfolio. So there's a lot to sort out, and these are big complicated agreements. But we feel very, very good about where the pipeline is, where the portfolio is at, the products are resonating. And we'll continue to update you as we build out the portfolio of NBMs over the next couple of quarters.

Okay. Great. And just a quick follow-up. You may have seen that Intel guided down for the next quarter due to PCs and it really the impact of the memory is more than anybody thought, at least with their guidance, I don't know about anybody else. Just wondering if you are seeing NBM interest from those in the PC market more and how you -- if there's -- if that market, with regard to some of the device makers, is something that will become an NBM or is actually now getting more attractive because they're willing to pay more. How should we think about that as it looks like they're starved for product and missing numbers in that segment?

Yes. I mean, Ben, we have -- we cannot talk about specific customers. What we've said in the past is we do have NBMs in the data center business. It's data center-centric. But we also have customers on the edge. And we've offered this construct to any customer that wants it and start a process where it makes sense. So we're not trying to restrict it from any customer. We're open for business.

So Ben, the other thing I'll say to this. I mean, we are looking at this in our own reviews. We know that units were down this year. They're going to be slightly down next year, low single digits. I mean these 2 calendar quarters right now are really the trough of the year-over-year compares where they're the most negative. And so this is where we're feeling the biggest amount of change in that market on a unit basis on a year-over-year basis. All that said, we expect the PC exabytes to grow next year. And in fact, we think our internal models are every segment in the NAND market, PCs, smartphones, data center, will all show positive growth next year because of content per device continues to go up as part of that is mixing to premium. So there's no doubt there's been a market adjustment.

In many ways, we're in -- as I said, from a unit basis, we're in the teeth of that on a year-over-year basis, but we see improvement from here.

Our next question comes from Joe Moore from Morgan Stanley.

Great. It seems like the underlying story here in addition to just how good the market is, is your presence in data center, you're up almost 20x year-on-year, up 61% sequentially. Can you talk about -- obviously, the market's good but it seems like your participation is also a lot better. Is that because of BiCS 8? Is that just because the market is different? Just what is it that's allowing you to get to those levels of penetration. And if data centers half of bits at some point, would you get to that in your own mix as well?

Yes, Joe. So thanks for the comment. We feel very good about where we're at in Data Center. I mean anything where you see the kind of numbers we posted year-over-year is a combination of things. I mean it starts with the fundamental NAND node. BiCS 8 is a very good node. CBA gives you very clean CMOs, which is very good interface speeds, all those kinds of things. We built the right products. We've had a TLC product in the market for several years now. It was just a couple of quarters ago that we started selling our QLC product for revenue and that we expected a faster ramp and that's happening. So we got the right product at the right time based on the right technology.

And then over the last quarter or so, there's been a lot more refinement of the AI inference architecture is what I would say. When I look at the market and we kind of decompose it internally, customers are just getting more refined about how they're using NAND, where they're using TLC, where they're using QLC all different phases from KV cash to staging to fast data lakes and all the sub parts of those markets, inference and agent orchestration, model pretraining model post training, there's just a general. We're seeing NAND be applied in a more, I would say, nuanced way.

In the past, QLC is being used in the right places and you're even seeing QLC sub customers even move into like KV cache where it can be supported. On the other side of the spectrum, we're seeing SLC being pulled into the architecture. It's not really in sales yet. It's more in architectural conversations. And it's not a performance type of conversation. It's more endurance think like during agenetic workloads, the amount of endurance required. And in that place, it's kind of where can we substitute NAND for DRAM. And we think those kind of architectures will be very sticky over time.

Once you make that transition, you're going to stick with it given the price differential. So we see all of these things working together and I think these NBM set us up for these very deep and strategic level conversations with our customers about how they're using our product deep discussions with our engineers about how they could use our product to solve different use cases and all of that is leading to a tremendous pull on our portfolio. And like I said, our biggest customers, we've already signed one second tranche of an NBM and we're in discussions in others. So the whole setup is very strong, and we feel good about it, and we appreciate your recognition of the performance on a year-over-year basis. It really is quite outstanding.

That is surprising. And where do you think the mix goes long term in terms of good [ overhead ]?

So yes, we're working hard to catch up to have our portfolio mix equal what the market bits are. And we're going to continue to make progress throughout calendar year '27. We'll see if we actually catch up by the end of the quarter, but I think we'll get to a substantially higher number on mix than we are if we look back at calendar year -- our fiscal year '26. We'll continue to make progress here.

Our next question comes from Mark Newman from Bernstein.

On the -- if I could just first talk about price and volume in the guidance. You talked about in the fiscal Q1 that you just reported, I think you said 60% of the revenue growth from volume and 40% from pricing. Baked into your guide, I mean, how do you expect that mix to progress? Any comments on pricing? I appreciate that you've got a very high portion of volumes on these NBMs but still wanted to get some clarity on pricing, I think there's a comment that modest improvements in pricing in Q2, any further clarity on that? And then I have a follow-on question on HBF and some of the new NAND technologies as well.

Yes. So let me take the first one. Yes, you're right. We reported -- about 60% of the revenue growth came from volume and 40% from pricing in Q1 roughly and we do expect both to contribute in the next quarter. We normally don't break this mark into its pieces. But we do expect some modest pricing as we put in our prepared remarks. So we do expect both to contribute during the quarter, and we expect pricing to continue to increase modestly throughout the year.

That's helpful. Fair enough. And follow-up question is on HBS and other new NAND technologies in the data center. So there was a comment about some progress on HBF, which was great to hear. But I wondered if there's any more comments on progress and discussions with partners and customers, anything there. And the other thing, Dave, you mentioned about some of the other new NAND technologies in the data center storage next and then there's also the CMX context memory. Just what's your view on all these types of new NAND tiers in the data center, do you think they all win? Just any kind of comments on the interplay of all these and the roles each and how Sandisk plays in and the timing would be really appreciated.

Sure. So Mark, no the conversations with potential customers on HBF are robust and continue to be -- we'll continue to have those conversations as we go into putting product samples in their hands that they can start to use and understand how it's going to fit in their system. So we'll continue to update on that, but it's kind of an always on conversation.

Your second question. I think this is like very, very healthy for where AI inference is, which is, as it scales out, it needs to be -- the architecture is just being relentlessly refined for economics and that's a very good thing because it's going to make -- it's going to scale, the more economic it is, the better -- and that's pulling NAND into more parts of the architecture because NAND is the most scalable semiconductor technology, right? So we're seeing -- like I said, we're seeing -- if you would go back a couple of years ago, it was all about going for more density again, going from TLC to QLC, we're still doing that, right?

That's still a big part of the strategy. But now there's some of it even going in the other direction. How do I use -- how do I get more endurance in specific parts of the architecture is the endurance in the performance I get out of NAND good enough to solve the use case at a part of the architecture, where I can use it over competing technologies that may be significantly more expensive and difficult to acquire.

So I would say this is a very, very dynamic discussion. There's a lot of ideas on the table. I think there should be a lot of ideas on the table. I think they'll all be pursued. And I think that the most important thing is to stay close to the people that are spending the money to scale the infrastructure. Because in the end, they'll make the decisions of what's right for them based on the use cases they're trying to solve at scale and where they're putting all this money behind it. And we're deeply -- we're deeply engaged in those conversations, and I expect them to continue to be very robust to have a lot of different ideas about how we could structure our technology in different directions to fit more and more use cases and building a product, it's always the same thing.

I want the right technology at the right time and the right intensity at the right cost and really not anything less or anything more. And that's what building a product is all about because you can deliver it at the right price and then you can get the right return on it. So that process is full on in AI architecture working with our customers.

Our next question comes from David Gibson from MST Financial.

Just wondering -- thanks very much for your time today. Just a quick one. Just wondering whether you're looking to accelerate the BiCS transition from 8 to 9, 10 and so forth as a way to accelerate the bit growth in the short term and then satisfy some of that very strong customer demand.

No. I mean we're committed to the bit growth we talked about, mid to high teens. Those nodal transitions require a lot of planning and tool purchases and where lines are set up. So we're still -- we have that fab plan set up for the kind of bit growth we've been talking about for a long time. So no plans to change that at this point.

Okay. And just 1 follow-up. You talked about earlier about allocating more supply to the data center in Q1. Are there any limits to increasingly allocating your bits to data center?

Well, I mean it's -- that's very -- in some ways, yes. I mean out of the fab, you come out with a certain mix of TLC, QLC, certain dynamics, all these kinds of issues. So it's a constant fitting exercise of getting the right product to the right customer and there's a lot of magic that goes on in the organization and Luis' organization is a big part of that working with our sales teams of what is the available inventory, what's coming out of the fab, what's the mix? And where is it all going to go? Do you want to say a little bit more about that.

Yes, I think the only thing I would add, obviously, changes depending on the time frame we're looking at over the short term, it's almost impossible to do anything. It takes 120 days or so to produce a wafer. But obviously, as you look further out, there are more degrees of freedom.

Our next question comes from C.J. Muse from Cantor Fitzgerald.

A question on the comment around your NBM pipeline exceeding supply through the end of the decade in the slide deck. I guess can you elaborate on that? And how is that changing kind of your thought process around what bit supply will look like for the industry over the next handful of years? And how are you adjusting, if at all, your CapEx plans. I know you talked about mid-single digits, which implies roughly $2.5 billion to $3 billion gross CapEx, fairly low kind of number. So I would love to hear kind of your thoughts on how the visibility is impacting your thoughts around bit growth and supply.

Well, I mean, I think my first thing I say about, C.J., is the process is working. I mean it was only 2 or 3 quarters ago, we talked about even instituting this kind of business model, and we had some customers that were like-minded, and we made some very good progress. And here we are a couple of quarters later, and it's -- we talked about this after our Investor Day, we had customers coming to us offering NBMs. And that now becomes -- it's not the only way we do business.

There's lots of customers with lots of different volume levels and different products, and we want to meet our customers where they are. There's no doubt about that. But it's very good to see that we have a robust pipeline now through the end -- through 2030, and we're even getting beyond that, as we talked about of customers that are really willing to give us their demand signals and financial commitments. And to the question earlier, those kinds of things will allow us -- if we choose to eventually start changing supply and demand and all those kinds of things, but we're just not there yet, right?

We got to go through negotiations on all of those.

We've got to get commitments. We've got to get everything all aligned. But I would say the process is working well. We're very happy with where it's at, and it gives us a lot more visibility into how we plan. But the process is still underway, right? It's still underway. And we'll -- as I said, we'll continue to update you over the next couple of quarters as we make progress.

Just a quick follow-up for Luis. On these contracts and considering some have deposits, are you able to use those deposits for buybacks? Or is that cash restricted?

Yes. In theory, we could, C.J. But as we explained at Analyst Day, a vast majority are financial guarantees, which are held by third-party financial institutions. So they are not in my balance sheet. The only thing that's in the balance sheet is prepayments and we -- that's by far the smallest amount. But there is no legal restriction on that.

Our next question comes from Karl Ackerman from BNP Paribas.

It's encouraging to see increasing NAND content per XPU and AI servers to support at cash demand, which I think compares to reductions in memory content and some server SKUs. But Dave, could you discuss how you see enterprise SSD content per server progressing in 2027 in the context of this agentic AI proliferation because it's certainly becoming a larger portion of your portfolio?

Yes. That's a big question, Karl. I mean, look, what we're doing is there's no doubt in inference agent orchestration that our product is tied to kind of CPU demand in -- for those type of applications. I mean we know at a pretty granular level the amount of demand that we expect per CPU, something 60 to 90 gigabytes per CPU is what we see in those kind of applications. I haven't run the math against what the server TAM is and kind of what the elasticity is between those two, but we can follow up on that. But as I said earlier, we continue to see more NAND being pulled into the inference architecture in all parts of the architecture and in very different ways, right? TLC, QLC and as I said earlier, now, people looking for high endurance solutions that will fit -- work -- fit parts of the workload where they don't need to use DRAM and they can use that type of NAND. So it's all very encouraging, and we'll continue to stay close. But we'll do some work on the 2 TAM analysis and what the trade-offs are between them.

Sure. If I may, a quick follow-up. Do you have an updated view or progress on your goal for enterprise SSD share? I mean can you achieve perhaps a low to mid-teens share in 2027 because you're seeing quite success right now?

Yes. We're probably not -- I mean, in the past, many years ago, we put some share goals out there, and I've tried very hard not to put share goals out there because sometimes it leads to kind of perverse choices in the portfolio. We want to maximize the value of the portfolio across all time horizons, near term, midterm, long term. So we want to keep some balance into it. Clearly, we're going to mix higher into data center. I mean you see that. We talked about that earlier. I think 17x growth year-over-year. You don't see a 17x very often in a business at this scale. And as I said, I expect that to go higher as we go through '27. In the balance point, in some ways, will be end will depend quarter-by-quarter or what we see and then what NBMs we signed, right? Those will give us a big -- a lot of visibility over the next several years where we see that mix going. So it's going higher, but I'm not going to put a specific target out there.

Our next question comes from Vijay Rakesh from Mizuho.

Just a -- great quarter and guide. Just a quick question. As you look at data center, [ 2026 ] looks like you mentioned demand bit growth, which is massive. As you look at this BiCS 8, BiCS 9, BiCS 10 transition, does that help you drive some bits to meet that demand? How is the supply on that side? I guess and I have a quick follow-up.

Look, I mean, BiCS 8, BiCS 9. BiCS 9 is a big part of a data center strategy of being able to get faster interface fees with the same cell and kind of what CBA gives us that flexibility. We talked about that at Investor Day. So in that sense, yes, those are big parts of the strategy. There's no doubt about it. So as I said earlier, it's about getting the right product to the right customer and the TLC product has been performing very, very well for quite some time now, and now we're seeing the ramping of QLC and we'll see some ramp of BiCS 9, but BiCS 10 is still a little ways out, right? I mean we announced BiCS 10, we're still ramping BiCS 8. So we've got a ways to go before we get out to some of those future BiCS nodes.

Got it. And then on the endurance side, as you mentioned, some of your peers talking about SLC NAND out there. Any thoughts how that plays out next year? Obviously, that could tighten up the NAND market as well in addition to what you're seeing in just overall demand growth, I guess?

Yes, I think it's a little early to have a very strong view on that, right? We are talking to some customers that are giving us some signals of demand for SLC. I mean, clearly, you start making SLC and you're going to take away. There's a clear trade ratio of 3:1 for 1, whether it's TLC or QLC. So it will impact the overall supply of the market depending on how much SLC is used and how much really gets deployed. But it's a little early to put a specific forecast on it right now.

What I would say, it's -- it's an architectural discussion and a request from customers of better understanding the technology and how it could fit part of the use case as they build out their inference architecture. So we'll update you as we go over the next couple of quarters, and we'll see how this develops. But definitely something to keep an eye on.

Our next question comes from Jim Schneider from Goldman Sachs.

I was wondering, maybe on a more tactical basis, your consumer business obviously has seasonality in Q1 historically. And obviously, the PC business does too, but data centers is in a, sort of, secular growth mode. So Luis, based on your visibility at this point, I was just wondering if you could maybe give us any kind of color on whether you'd expect Q1 to be sequentially up, down or sideways roughly?

So fiscal Q3, you mean?

Yes, the March quarter.

Yes. So if you look at historically, right, our Q3 is normally our weakest sequential quarter. So last quarter, for example, our bits were down 19% sequentially. So If I -- we're not guiding Q3, but for modeling purposes, I would be conservative in your sequential growth from Q2 to Q3.

And then just as a matter of principle, on the buybacks, it's good to see the continued execution there. I'm just sort of curious as you sort of start to run down the authorization. Would you expect the Board to re-up their authorization on a quarterly basis? Or maybe you could provide us any more color on how you're thinking about executing on the actual buybacks given a much lower stock price today?

Yes. So we're -- we have a very clear strategy, 100% of excess cash is going to go back to shareholders. We're going to execute that aggressively or appropriately, and we know how to get a hold of the Board. We have meetings every quarter. And so I don't expect any impedance on that side of the -- from us about how we execute. I expect us to continue to be -- I don't expect us to be, we will be continue to be faithful to our strategy, which is we're going to return 100% of excess cash to shareholders. We're off to a good start, I think, over the last couple of quarters totally. I think we bought back 4% of the market capital of the shares outstanding of where we started, and we're going to continue to execute that program appropriately.

And our next question comes from Krish Sankar from TD Cowen.

This is Steven calling on behalf of Krish. The question I have for Dave. I guess looking at your CSP customer base, I kind of wanted to get some color on what the prevailing demand currently is between TLC versus QLC bits? And kind of going forward with the NBMs in place, does that TLC versus QLC mix shift meaningfully later this decade?

We're still like at that 70-30 number on TLC to QLC, TLC has been a big driver of the portfolio. We are hearing more from our customers about a reacceleration of QLC in different use cases. And so we'll see how that plays out, as you said, through the end of the decade. But yes, that's where it is. I mean that 70-30 number is a pretty good number. And of course, there's some fluctuation quarter-to-quarter a little bit up and down here and there, but that's where it's been, and that's where we expect it to be at least for the next -- the rest of this fiscal year.

Let me put that time horizon and then we'll see what customers are thinking about QLC as we continue to have those discussions. And I will say the ramp -- the Stargate product has been -- which is our QLC product, I don't use too many internal names here, but it's been very well received. It was not that many quarters ago that we were talking about going through qualifications, that's all 100% behind us. Customers are ramping. It's been a very, very well-received product in a very key part of the portfolio.

Got you. And as a quick follow-up, a question on the mobile edge market. Just given the double-digit declines in that market this year as well, but albeit concentrated in the kind of entry and mid-tier of the market. I'm kind of curious like from the demand signals that some of your customers in that work that are giving you, can your mobile bits still grow next year overall?

Yes. I think let's talk about the market first. I mean you got it. I mean, the low end under pressure, but that leads to a unit decline but a content per unit increasing. So when you add it all up, we expect growth of exabytes in that market next year on a slightly declining unit count. As far as our portfolio, I mean, we don't -- it's kind of like a data center. I don't want to make projections on individual markets.

We will manage the portfolio with agility and flexibility. We've been very clear that we want balance in the portfolio. But we're going to optimize the financial outcome over the short, mid and long term kind of integrate those to get the best outcome, and we'll update you on what the mix of mobile is as we go forward.

Our next question comes from Wamsi Mohan from Bank of America.

Yes. In your slide deck you say leading AI companies could indicate further demand upside I was wondering if these are AI labs or Neo clouds? And are any of the credit concerns around the new clouds filtering back to you in any way that you can see at the moment? And I have a follow-up.

Yes. We haven't seen any of that. To the contrary, we're seeing very strong demand from our customers. And as I said, if I look at our largest customers, they're basically all investment-grade companies. right, generating very strong cash. So we're not seeing any impacts from any credit issues whatsoever if they are out in the market and demand from our customers is, if anything, strengthening versus where it was a quarter ago.

So to clarify there, Luis, when you say leading AI companies that you're offering to AI labs over there?

We sell to everybody in our business. We sell to hyperscalers, we sell to model makers, we sell to everybody.

Okay. And then just on linearity in the quarter, it seems like receivables picked up a little bit. Is that something that we should expect because of payment terms as more bits go under NBM? Or was there something else in the quarter?

No. We basically saw a little bit of a change in customer mix. And that has a little bit of an impact on our receivables a little bit on linearity, but it's mostly customer mix.

Our next question comes from Amit Daryanani from Evercore.

I guess, David, just to start with on the capital allocation side, as you transition towards becoming a more predictable NBM driven business model, does that change how you think about your capital structure? And historically I think memories perceived to be more cyclical so balance sheet flexibility has been important. But as you become structurally, I think, less cyclical, would that support a willingness to take on leverage and potentially accelerate the capital returns or is it comfort still with the $5 billion a quarter buyback which is fairly impressive. I'm just wondering, at what point do you say we can take on more leverage and take advantage of the fact that we're less cyclical?

Yes. I mean we put a financial model out. I mean we just talked -- we just spent 3.5 hours talking about this at our Analyst Day, and we committed to a financial model through the end of the -- through 2030, which I think was pretty impressive. And we plan to execute that, and we're very comfortable with the strategy we have of staying debt-free, very strong balance sheet and then returning -- investing in our business. We have a growing business and lots of opportunities for internal innovation, which we're going to continue to invest in like HBF and 3D Matrix memory in addition to the BiCS road map and all the great products we build.

And after we do that, we're going to return 100% of the cash flow to shareholders, and that's the plan, and we're going to go execute that for the foreseeable future. And so we get up at another point and tell you we're going to change it. But I don't anticipate that -- I don't anticipate that happening in any kind of near-term time frame. Fair enough.

And then maybe just on HBF and rather trying to ask you to frame how big the TAM could be. Could you just talk about what the scaling variable really is every year as inference models and context windows expand, should we think HBF capacity to grow in line to GPU compute or potentially faster and I think you said you will start initial sampling in '27. What's -- what are the key milestones from there to becoming a meaningful revenue ramp as well?

Well, I mean, so it's a little early for your first question. But on the second question, we get to die back. We have the base die, which we'll tape out, which is the controller, we'll put them together, put it in customers' hands. And then they start to see how the product is going to work in their infrastructure. We've obviously done a lot of modeling on that, but actually having the actual product and then doing the integration work. That will then lead -- that kind of feedback from the customer will then lead to be able to start to answer some of the questions that people have. Like what's the market opportunity how fast is it going to happen? What's the TAM? Where exactly is the architecture going to fit. We're going to follow our customers on that. But the big step is to actually put a product in their hands.

And then that leads to a whole another set of conversations with them we actually have the device that we're -- instead of running simulations, we're seeing exactly how it performs in their infrastructure. And does it deliver the results of what they're expecting. And we'll also -- that also helps you continue to expand the set of use cases, right? At this point in development process, you want to be a little bit careful you don't engage too much because you're overwhelming your engineering team with 1,000 ideas of how the product can be used.

So as we get -- as we get actual device that we can put in people's hands, we can open the aperture on those kind of conversations, which will, I think will lead to more use cases for the product. But that's that's still TBD. So I think a lot of things start to happen when you get to that next milestone of actually having a real product. You can -- people can start to -- they can start to use and understand how it impacts what they're trying to do, which is deliver a world-class inference architecture at the best cost point possible. And we think this is a fantastic technology for doing that. But we'll see when we get there and we produce it.

Our next question comes from Mehdi Hosseini from SIG.

David, I want to go back to -- sorry, for the audio. David, I want to go back to what you said earlier, you said that the AI inferencing is being refined, having an impact and more SLC needed. And my question to you is, is this having an adverse impact on adoption of TLC or the growth of TLC or SLC is additive to the same demand by TLC and QLC?

No, I don't think it's a cannibalization of other NAND products. I think it's an expansion of the TAM into areas that you need different attributes of the technology. You need different endurance, you need different performance. And we can take our product and we can run anything out of the fab in an SLC mode. So there's a lot of flexibility in the product. And I think our people are discovering that and starting to understand how can I apply this in ways where I get I get the attributes I want for the problem I'm trying to solve.

So it's just a more refined view of the architecture, but I don't see it as cannibalizing other parts of the NAND stack. If anything, I see it is -- and I want to be careful on this because it's still early, but is it good enough to replace places where I'm using DRAM, right? Can I get enough endurance? Can I get enough performance if I run NAND in an SLC mode because the cost differential is like huge, 20 to 40x. So I take it as a very healthy sign in many ways, Mehdi. One is like very, very real about how you scale an architecture. When you scale an architecture, you got to do it in the most cost-effective way possible.

And again, NAND is the most scalable semiconductor technology. And so it doesn't surprise me that people are finding more ways to use it. In many ways, I think scalable technologies kind of attracts TAM, and I think that, that's what's happening here as these architectures are getting more detailed and more refined.

Okay. Let me come from a different angle. And if there is this demand for more enduring SLC and as customers wait for commercialization of HCF. What is the JV's flexibility in adding some SoC capacity to meet that short-term demand.

No, we don't -- I mean, we don't -- there is no -- I mean I wouldn't think of it as SLC capacity. You can take any -- ourselves are flexible enough that we can run them in whatever mode we want, SLC, MLC, TLC. So we can supply the market with the wafers we have, right? That's how we think about it. And I understand the question of wanting to go to the supply question right away, but it's just like super early to that. These are conversations at this point and customers exploring how they can use technology starting to get some idea of what their demand would be if you use it this way. It's something to keep your eye on. It will impact the market about that because SLC is a little bit of SLC, it is a lot of QLC, but more -- I guess more to come, Mehdi.

And at this point, ladies and gentlemen, we'll be ending today's question-and-answer session. I'd like to turn the floor back over to management for any closing remarks.

All right, everyone. Thanks for joining us on the call. We appreciate it. Great conversation, and we look forward to seeing you throughout the quarter. Thank you.

Thank you.

And with that, we'll conclude today's conference call and presentation. We do thank you for joining. You may now disconnect your lines.

Good morning, everybody. Welcome to the Goldman Sachs Communacopia Technology Conference. My name is Jim Schneider. I'm the semiconductor analyst here at Goldman Sachs. It's my pleasure to welcome Sandisk today. We're very happy to have CEO, David Goeckeler; and CFO, Luis Visoso. Welcome, guys. Thank you for being here.

Thank you for having us. We're happy to be here.

Excellent. I think you want to start with the safe harbor, please.

Yes. Thank you, Jim. We will be making forward-looking statements in today's discussion based on management's current assumptions and expectations, including with respect to our technology, our product portfolio, our business plans and performance, market trends and opportunities and our future financial results. These forward-looking statements are subject to risks and uncertainties. We assume no obligation to update this statement. Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations. We will also be making reference to non-GAAP financials and reconciliation of GAAP to non-GAAP results can be found on our website. Thank you, Jim.

Excellent. Let's get started. So maybe, David, let's start by helping us understand at a high level, how you believe the role of NAND flash memory is kind of changing in the build-out of AI infrastructure, both in terms of what NAND enables as we shift to inference and how large the data center portion of your NAND market is going to get?

Yes. So I think as -- I mean, we've been getting this question for a long time about how does NAND play in AI, right? That was the question in the training phase of AI, and it's less obvious there. But certainly, as we move into inference, we're seeing that the scalability of NAND, the idea that models are getting bigger, [ context ] lengths are getting longer. You just get -- from a solution perspective, you're going to be driven back to more scalable technology, right? And so NAND has a huge role to play. And as inference continues to expand, and it's fantastic. It seems like every week that goes by, there's a more capable model and more interesting things that we can do with the technology. And inference is just going to continue to expand and be proliferated around the world.

And NAND is a big part of that, right, where KV cache becomes a big part of that. RAG is a big part of that. So our technology becomes front and center about how you -- not front and center, but a big part of how you scale this technology. There's a lot of technologies involved in this, obviously. And that's really changing the dynamics of our market, right? So it's great how inference is driving that. Data center has always been a big part of the NAND market. It's always been a big component and a growing component. But now it's really growing much stronger. And we're kind of at a watershed moment where NAND becomes -- our data center becomes more than half of the NAND market.

So when you see something become more than half of the market, it tends to change the way the market works. And the use case for how the customer monetizes our product, how it's used, the way it's consumed, you're starting to see that ripple through not only the demand, demand is great, but it's allowing us to get at some of these business model issues of the way the market has always worked and kind of shift that in a way, which I think is the benefit to the supply side and the demand side of the equation, and we can talk more about that.

Yes. I don't want to get to that. But I did want to ask you relative to data center specifically, what are the biggest opportunities for the company over the next 12 to 18 months tied to AI? How is your market position differentiated relative to your peers?

So it's a big opportunity for us. I mean, we've traditionally been a business. We've kind of had this evolution of the business if you look over a long arc of time, like decades. Big consumer business, great brand. We still have a global brand. We invest in the global brand. I hope you're all consumers of our product as in your everyday life. Made the transition into client, very large client business. In many ways, you can look at that as the way -- at least the way I look at it is like the Western Digital era was a client, Tricia client business that moved into NAND. And Sandisk had a world-class portfolio there and continues to have a world-class portfolio there.

In seminal innovations like the DRAM-less, client SSD, all these kinds of things, our opportunity was to move into data center. And over the last 3, 4 years, we've been building out that portfolio. And that portfolio has now shown up over the last couple of years exactly at the right time as the data center now becomes more than half of the market, all the things we've talked about, inference is really scaling very aggressively. So that puts us in a very good position because we have a great portfolio.

First of all, we have fundamentally really good NAND technology. We can talk about that, BiCS8, BiCS10. We've got the compute and storage enterprise SSDs kind of architectures that have been built in the last couple of years. So very, very strong architectures on the controller side. And that allows us now to tap into this side of the market that's more than half of it, half of the TAM. So it creates a tremendous opportunity for us to grow the company and to change our mix. And then we've got some other technologies like HBF that are on the horizon that may be a couple more years down the road that are also big opportunities for us.

We'll get there too.

We'll get there -- there is a lot of stuff to get through.

But you recently made the argument you just referenced it before, the industry is moving away from a near-term spot market governed by short-term supply-demand balance to one that's more characterized by long-term supply agreements, more stable pricing. Help us understand what that means to the company and why you think that shift is happening right now? Is it just the demand shift that you mentioned before? Or is there something else at play?

So it's always been a market that -- I came into this market about 6.5, 7 years ago, and it's always been a market that surprises me a little bit because there's this huge time horizon between the supply side and demand side. Like the supply side is making decisions for like 10 years. Over the next 10 years, you got to build a fab. Like that's not -- it's a very large undertaking. And once you build a fab and you turn it on, you got a lot of fixed cost, you tend to want to run it all the time. And so you really need to make sure there's consistent demand for that product.

And then on the demand side -- so the supply side has this time horizon where you're thinking in decade. And on the supply (sic) [ demand ] side, it's like we'll kind of say what our demand is for the year, but then we'll negotiate price every quarter. And if we agree to price every quarter, then we'll agree to transact. So you kind of have this huge gap between the planning horizon of being a little bit dramatic here, but like a decade on one side and 3 months on the other. So when you put those 2 things together, it's hard to get them to match, right?

Somebody is always like not in a great position, either we've got too much and pricing is very low or we don't have enough and I can't get it and all this kind of stuff. And it's like, why are we doing this to ourselves? And that's -- when the market is driven by consumer-type products, that's not that surprising. -- the customers in -- if you're providing consumer-like products, you tend to have more ability to shape your demand. Maybe your business model has a component of it of selling more of our product to sell more of our product. When I buy a device, I can choose how much of our product we want. So the very dynamics of the way the whole market worked led to this kind of business model that we kind of rationalize the price every quarter.

Okay. So now we're transitioning to data center, all the reasons we talked about earlier, very different. The business model of the data center operator is not quite so transactional, right? It's like I'm building a data center, I'm monetizing it in a different way, and I need your product to build my data center. I'm not really -- I can't really shape my demand on a quarter-by-quarter basis. It's just kind of growing all the time. And so you tend to have customers that are more interested in consistency of supply as opposed to what is the price every quarter. And so -- and as that market became more than half of the TAM, that became like the major way now you're going to start to do business. And so we recognized this a couple of years ago as this was happening.

And this year, especially as we see the growth, we went into these conversations and we said, let's find some willing partners that want to change the way we do business. We want more visibility and more predictability. They want more visibility and more predictability. And they were sending us demand signals for years in advance where I would look at those demand signals and say, it's unlikely you're going to be able to get everything you want because your numbers are big if we're going to transact every quarter. So we found willing partners that can say, "Hey, you're going to get what you need, which is certainty of supply. We're going to get what we need, which is visibility, more predictability of demand, then we can decide on economics where that works for both of us and then we can kind of put that in place."

And that's what we've been doing. And I think that not every customer wants to do business that way, which is fine. There's a range of ways we can do business. But what we're seeing and what we've talked about the last couple of quarters is just literally in the last, I would say, 8, 9 months, Luis, that we've been talking about these, we've been able to shift now the predominance of our supply over the next couple of years to that more predictable model, which is where we want to get to. It helps us with that investment decision.

Yes. Now I want to ask you specifically about supply for a minute. You said you -- very clearly, you want to maintain mid-teens growth in supply over the next year at least. I think a lot of investors worry about what happens if your competitors get more aggressive with their additions and what happens if that number goes to like 30% for them. So how would you frame the risk of faster supply growth over the next 2 to 3 years? And how would you and Kioxia, your JV partner respond to faster supply growth by them?

Yes. It's hard to get into these kind of what-if games like what if somebody else does something uneconomic, should you do something uneconomic first? I mean that's not the way I think about it. The way I think about it is we're trying to get -- we're trying to get our incentives aligned with our customers. And those incentives are changing, right? They want more predictable supply, we want more predictable demand. And we're going down that path to elongating this view of supply and demand, but just how we rationalize those 2 things and give us more predictability.

I mean, I suppose that could lead us to a place where we have enough visibility that we had 10 years of visibility and that could influence the way we supply the market, but we're not near that yet. And I don't think it's really worth -- I don't really think about how we're going to respond to something that could happen like that because I don't think that's what's going to happen. I think the market is changing in a way where it's beneficial for everyone, the supply side and the demand side. And we're going to continue to refine that business model.

And that's a great business. I think the business model that we put out at our Investor Day, it's been a month ago now. I am glad. I think that's a great business model, right? And I think if we can walk into that business model and we have and we can get the business practices in place, that's where we want to land. That's where we're going to land.

Yes. Now another related worry that I hear from a lot of investors is China specifically. You've had CXMT been more [indiscernible] YMTC as well. To the extent they're better capitalized, how would you frame the risks of that competitive threat, both technically and from just a pure supply perspective coming from China?

So let's start with the technical side of it. I mean, look, if you're in the technology business, you have to have great technology. So -- and you have to assume your peers have great technology as well. So you don't want to discount anybody in that respect. But we're going to make sure we always have fantastic technology, and that's where it starts. And with us, between us and our JV partner, we produce roughly 1/3 of the world's NAND. That gives us great economics of how we invest in R&D. It means we can invest at that level.

And that -- if we do our job correctly, and I think we will and we have a track record of doing that, we will always have really, really good technology. I think other peers in the market, there's other very capable people in the market that will invest as well. But if we continue to do that and especially if we stay very capital efficient, we'll be in great shape. Now particularly about China, I mean, China is an interesting market. It's kind of a China for China story at this point, and I expect that to continue. And -- but that's a very big and dynamic market. There's plenty of room for them to grow into that market before it becomes anything beyond that.

Yes. Okay. I think we've also talked about technologies. We've often heard about the importance of HBM memory in the AI server architecture and design. I think it's fair to say that there's been this high-bandwidth memory that's really important. You talked about last year, this notion of high-bandwidth flash. You talked at your Investor Day a little bit more about it. So help us understand broadly what you think high-bandwidth flash could mean for the industry? Is this technology complementary to HBM in the DRAM space or potentially cannibalistic to some extent?

So I would say, first of all, I mean, one of the things -- a lot has happened to us in the last 18 or 19 months since we launched the company, right? It's been a great ride. And one of the things that's most satisfying is in February of '25, we got up and we have been working on this technology for quite some time already, and we were trying to decide should we talk about it. And we eventually decided, hey, we're asking people to invest in the company. We should talk about all the interesting things we're working on. And we kind of put this idea out there of high-bandwidth flash. And I think everybody looked at us and said, like what are you guys talking about, right? It's like, are you talking about HBM because, I guess, maybe it had HB in the beginning and people thought it was a substitute or something like that.

And that's not what we were talking about. But what we were saying is we are kind of shooting for this day of inference. -- where we could bring this very scalable NAND technology to play in AI and specifically in inference. And back in those days, it wasn't that long ago where the predominant conversation around AI was all around training. And HBM is spectacular technology, right? It's incredible technology. And our goal is not to -- we're not recreating HBM. That has its place in the world, very, very important. But as we go to scale inference, we thought, hey, these models are going to get very large. I think the team did a good job of anticipating some of these problems. We're going to need more density. What does NAND bring?

NAND is the most scalable semiconductor technology. It brings a lot of density. And I think some of the insights were if you're a NAND designer, you've been kind of told your whole life, give me more density, right, give me more density. I think in the R&D side of our business, many years ago, people said, well, what if we -- we're anticipating this world of inference where bandwidth is going to be more important and endurance is going to be more important. So can we take our fundamental NAND design and twist it in a way and make some innovation where we can increase the bandwidth, we can increase the durability, and we can really start to play in this world where inference comes along because we know we can deliver the density, right?

That's one thing HBM has -- DRAM has a certain density level. It has unbelievable characteristics because there's only so much density you can get. NAND, we get a lot of density, and we have a road map for more density in the future. We talked about it at our Analyst Day, like we can produce the bits. It was about getting the other parts right where it played a role in the AI architecture. And so that's the journey we're on with HBF. It's not necessarily a substitute. It's just different. And it gives the providers that are scaling inference a different way to think about, hey, I could get a whole bunch more density, and I could get this bandwidth.

And if I can just get the system to work right to internalize that, I can really scale this inference and change the economics in a fundamental way. And that's what it is. So it's not a replacement for anything else. It just provides a different level of innovation to solve this kind of memory wall problem that now is talked about an enormous amount. Again, 18 months ago, we didn't talk about it that much, but we've always kind of believed that AI is essentially a memory-bound problem and how could -- we believe we're very big innovators in memory. How could we apply all of that innovation capability that we have as a company into this inference question, and that's what we're doing with HBF.

It's very exciting. We still got some time to go, right? We're just taping out the die, and we want to get in a position where we can put samples in customers' hands next year, and then they'll figure out how to integrate it into their system, and then we'll see what the real potential is. But we think it's very, very exciting technology. And again, over the last 18, 19 months of all the things that have happened to us, I think this is one of the most exciting for me because it's like pure innovation. You introduce something new that people hadn't thought about and we're moving the whole market in a way where now people are talking about a lot and they're seeing the potential of it, and that's exciting.

Fair enough. Now we kind of already covered the catalyst for what's changed the market structure and the need for these longer-term agreements at least for some customers. I want to talk some more specifics about that and maybe give Luis a chance to speak up a little bit. But also, help us understand, you've talked about the structure of these contracts, multiyear in nature, price ceilings, price floors. How do you kind of conceptualize this? Was it done in conjunction with your customers at a broad level? How should we be thinking about sort of like going forward over the next several years? What are the risks to what you're leaving on the table if pricing continues to go up? And then what protection do you have if pricing goes down in the broader market?

Yes. So as the market was changing, as David explained, our customers were coming to us with very large demand of NAND. And we just didn't think there was a way to transact every quarter. David mentioned that as well. So we're trying to see how do we move from a quarterly negotiation of [ bizarre ] into a long-term situation where we have predictable demand. And it happens that they were looking for predictable supply because they're making huge commitments, they are buying land, they're making energy contracts and everything. They want to make sure that NAND is arriving at exactly the right time.

So we started small and we started playing with the idea and over time, we -- over really a small period of time because it's only 9 months since we signed the first deal, we have 8 very strong partnerships. And I think that number 8 is very important because we've been very selective. We don't want to have a ton of agreements. We want to go and be very strategic for a few of those customers. And they have requirements, which are very meaningful and going all the way to 5 years. And we -- that was the first negotiation.

How do we get to an agreement on supply-demand certainty and for how many years and how many exabytes? We would spend, I would say, 2/3 of our time agreeing to that foundation because a 5-year agreement is actually an agreement with details by year, by quarter, by month, think about TLC, QLC, high cap, low cap. So it's very deep in terms of the agreement. So we know exactly what they will need and where we need to produce. And then once we have that foundation, we talked about pricing. And we wanted to make sure that pricing was attractive for us and attractive for our customers, and we came up with this construct. There will be a time -- and this is generic because there are -- each agreement is different. But think about a time in which prices are fixed.

And then there is an ability for us to capture upside if prices go up and an ability for our customers to capture some downside. But we wanted to be careful on the downside to make sure that we were protected. So we set that price in a way that was still highly attractive for us.

And the final point of the negotiation is, well, how do we make sure that you're going to be there until the end because I may have to burn some bridges with other customers, and I want to make sure that you're actually going to fulfill your purchase obligations. So we came up with these financial guarantees with third-party financial institutions, which basically will flow those funds to us if they don't execute their part of the agreement.

So we feel very good about the agreements. And as I've mentioned before, just as we signed -- we were about to sign one of them, one meaningful one and the customer came back and say, "Hey, my engineering team is actually changing the signal, they need a lot more." And we're like, okay, well, let's get this done. Let's give your team a little bit more time and not unexpectedly, that they came up with more requirements. And so we continue to have these conversations, which I think is very important because we are establishing a relationship, very strategic partnership that we expect will last for many, many years to come.

Yes. And then you've talked about sort of 2/3 of your fiscal '28 BiCS supply being booked under these NBMs and '29 levels at that rough level or maybe moving higher than that. In a downside scenario, you've talked about locking in that 80% gross margin level for pricing on average. Is it correct to think very simply about 2/3 of your business in a downside scenario being 80% gross margin and the remaining 1/3 at prevailing prices?

Yes, that's an easy way to think about it. I totally agree.

Okay. Great. Maybe want to -- well, I do want to ask another couple of technology questions, but maybe just kind of like wanted to talk about your targets because that kind of goes hand-in-hand with what we just talked about here. For fiscal '28 to '30 back at your Investor Day, you talked about revenue growth in the mid- to high teens, gross margins of 80%, operating margins of 70%. Given the amount of business covered by your HBMs -- NBMs during that period, what are some of the overarching assumptions you've made with respect to market supply and floor pricing to sort of get to those targets?

Yes. We didn't start with the market. We started with our relationship with our customers, right? And we have a good idea of how to model our NBM business because we have, again, close relationships. We have expectations. And we model a range of scenarios where prices would be higher, prices would be lower, how much we renew, how much we expand, how many NBMs do we have. And then we did the same for the non-NBM business under what reality will those -- that part of the business will be operating. And we blended the 2. And as I said, we looked at many, many potential outcomes.

Then David and I sat down and we looked at what do we think is the most likely scenarios, right? And we had several potential outcomes, and we came to the conclusion that this is a reasonable expectation, and that's what we shared with you in our Analyst Day. We think it's a great model, right, 80% gross margin all the way down to 50% free cash flow margin on an ongoing basis. That is a model we like, and we wanted to share that with you.

Let me just make one comment on this. And I think Luis and I we didn't sit down once, we sit down like 5 times a day and talk about this, but we always knew we had an incredible franchise, right? Whether you look at it from a technology point of view or from a business point of view. From a technology point of view, our technology is used in like every interesting device that's built in the world, right? Everything, like smartphones, laptops, drones, cars, data centers. It just goes on and on and on. And in many ways, it's kind of an evergreen market. Like there's always something new people are thinking of robotics. It kind of -- so it's always been just fantastic technology.

We've been able to produce it at scale. It's highly scalable technology. We talked about at our Investor Day, just through the application of what I'll call intellectual capital, new nodal transitions, we can grow at a compounded 27% growth rate if we look at a 10-year period, that's incredible, like that's not new CapEx. So then you start to transition into the business model. We always knew that the dynamics of this business were incredible for free cash flow generation. If you can get the economics right then you can get the cyclicality to settle down. And so when we went into these NBM discussions with our customers, we had a willing partner where it's like they're going to get a great deal out of this, which is they get predictable supply of incredible technology and they have unbelievable businesses, very enviable businesses.

And we're getting out of this predictability of demand that we're feeding into something we know is an incredible economic model. We can scale this technology very efficiently, and that allows us to build a really, really attractive financial model for all of us that invest in the company can drive an enormous amount of free cash flow margin. And that's what we were shooting for, and we've been able to kind of land that to the level where we could get up a month ago and say, "Hey, we have enough confidence in the contracts we've signed and the way we've navigated this business that we can commit to this financial model." And we think that's a great place to be.

Excellent. I want to touch on capital for a second because I think you put up an interesting chart at your Investor Day, basically showing like per petabyte. The industry spent about 2.7x more than you over the recent cycle, which kind of speaks to the level of cost reduction you're able to drive. So how are you thinking about how that plays into sort of your mid-single-digit capital intensity target? And then on the flip side of that, as we think about free cash flow, you've talked about 100% of free cash flow return to shareholders. Talk a little bit about like why buybacks make sense now and whether dividends might ever make sense in the future.

Yes. So the first one is kind of this thing I was talking about. We knew we had an unbelievable franchise. Like we knew we had incredible technology that we could produce very efficiently. And like the relationship we have with Kioxia is a very, very strong relationship. It gives us scale where scale is important. In our business, scale is important in R&D. I think most people think about scale in production. That's important, too, but R&D is really important because if you have the highest share, that just means you can invest more engineers. Like roughly in technology, big global technology franchises, I'm not talking about the start-up world. I'm talking about big mature markets, you can roughly afford to invest R&D commensurate with your market share, right?

And so if you're the biggest player, you should be able to invest more money, which means you should have the best technology. And those -- that chart you just showed -- you just talked about, we can produce BiCS more capital efficient than anybody else is the output of 25 years of doing that together of like this consistently applying the most R&D, very effective with a goal of how do I -- not just how do I produce the most BiCS or how do I get the most density? How do I produce the incremental terabyte or incremental petabyte of supply with the least amount of capital. That's kind of the goal for the R&D team.

And they've been doing that for years and years and years, and they've been doing that in intensity that's higher than anybody else because of our market share. And when you do that, you end up with charts like you just described. That gives us like this -- what I was talking about earlier, we always knew we had this incredible foundation. If we just get the economics correct around it, that's a big part of it, right? And we got that right, and that's a big advantage for us, and it's going to continue to be a big advantage. And I think one of the things you should take away from our Investor Day, we've got line of sight on that road map for BiCS technology for years to come. There's people working way down the horizon on that, which is exactly what you would expect us to do.

Now I think your second question was around how are we going to return capital to shareholders.

Yes. David mentioned we meet 5 [ times ] a day. This is one of the other topics we talk about, which is what is the best way to return cash to our shareholders. At the end of the day, your company and we're here to create value for you. We believe that the best way to do it is through a share buyback program. We said we were going to execute that, and we generated $5 billion in cash in Q4, and we bought $4.5 billion of that back. So we're basically executing exactly as we told you we would do.

And we continue to do that. We believe -- we continue to believe that's the best way to do it going forward. But we're not -- we're open-minded. If this ever changes and there is a reason why we should pay a dividend, we'll consider that. We don't think that's the right solution at this point in time, but we're open-minded.

Yes. And we just got a minute or 2 left, but maybe the way I'd like to close is sort of ask, you've met with a lot of investors since your Investor Day, including at this conference and other venues. Maybe talk about like something you think is still a misperception among investors and how you expect that to change over time?

Yes. I don't know if things are a misperception or not. It's just a question of -- some of this is just a question of time horizon, what do you believe and everybody is afraid to say it's different this time or if you say it's different this time, that's like a code word that you don't know what you're talking about or something like that. But we believe very strongly in the model we put together. Like we believed very strongly back in February of '25. Like I mean, we bet our whole careers on this company. We thought this was a tremendously underappreciated asset.

Like I said earlier, we looked through all of the noise around the way the business had worked in the past, and we said tremendous technology, unbelievable foundation, own the whole stack, right? No margin leakage and technology -- I mean, like we have the most spectacular customers in the world. It's like the who's who of everybody in technology uses our product, which is an incredible place to be, it was about just getting -- making a few changes in the way we did business, this huge gap I talked about earlier between we're negotiating pricing quarterly, and we're making investment decisions. We have to live with it for 10 years. We got to close that gap.

And then on top of all that, we had this transition of going to the data center. And how do we use that transition to make that business model change and really expose the value of this franchise. And I think we're like 2 or 3 quarters into doing that. And there's a belief like, oh, this is so good, it can't be true. It can't last. Well, why can't it last, right? It's just -- this is what we're creating. And I think what we hear from people is just really trying to understand. They help me really understand these contracts. How does this really work? Why do you have this belief? Because we've seen this so many different times, and it never seemed to work in the past and help me develop conviction that it's going to work this time.

And all I can say, I don't know maybe Luis could say it, too, we have a lot of conviction. We had a lot of conviction back in February of '25, and it's worked out okay. We still have a lot of conviction. In fact, I said when I opened, I think, our Investor Day, I finally feel like I got to the point after 6.5 years of managing this franchise where the real value creation is going to happen because we've gotten to that point where we're returning cash to shareholders. And we have a model where we can turn revenue into cash, and that cash is going back to the shareholders, and we have a lot of conviction in the durability of that. And it's just -- everybody is making their own decision of what the durability of that is.

Very good. Well, I think we're out of time, but really appreciate being here with us. David, Luis, thanks for being here.

Thanks Jim. I appreciate it. Thanks for your time.

[Tech Difficulty] Tech Conference. Welcome, everyone. Asiya Merchant. I'm here. I lead Citi's tech hardware and supply chain on the research side. Really excited to have Luis Visoso here, CFO of SanDisk. I'm sure this is one of the more exciting sessions here at Citi's Tech Conference. So I'm going to start off here with a few questions. Any safe harbor or something?

Yes. Really quickly.

All right.

We'll be making forward-looking statements in today's discussion based on management's current assumptions and expectations, including with respect to our technology and product portfolio, our business plans and performance, market trends and opportunities and our future financial results. These forward-looking statements are subject to risks and uncertainties. We assume no obligations to update these statements. Please refer to our annual report on Form 10-K and other filings with the SEC for more information on the risks and uncertainties that could cause actual results to materially differ from expectations. We will also be making reference to non-GAAP financials and a reconciliation of GAAP to non-GAAP financials can be found on our website.

Great. Thank you. All right. I'm going to kick it off here with a few questions. I'm going to leave some opportunity for investors to ask a few. Please do raise your hand so we can bring the mic to you. All right. Thank you again for being here.

Thank you. It's great to be here.

Of course, you guys -- I'm going to have to credit David and yourself, Luis here for these NBMs that've taken investors' interest by quite unexpectedly. I know you guys have been talking about it, but when actually people did see it in your results, and it's been 2 quarters, plus you had an Investor Day as well, where you talked about it. So I'm going to start it off with that because that seems to be on top of everybody's mind.

Just to recap, you expect NBMs from a bit perspective to be 50% of your bits this year, and you're talking about 2/3 into fiscal '28. Just where do we end up with the steady state? Are you thinking about this being like a 70%, 80% through your -- I think at the Investor Day, you talked about it through fiscal '30, sort of your long-term target overall when you lead that out. But in terms of NBM bits, like what do you think could be the steady state going on after fiscal '28?

Yes. I love the question and just what we're trying to do here. And you really take a step back because when we started, we were transacting every quarter, right? And it was tough because you started the quarter and you did not know what our customers wanted. You got an idea of what they wanted, but not necessarily agreement on price. And it was very tricky to manage and to guide a forecast. I remember one of our first Board member meetings, one of our Board members asked, what's your backlog? I'm like, what do you mean backlog? That doesn't exist in this industry. So it was not kind of profitability was really challenging. So starting this journey, on building the new business models, I think, was the right thing to do.

And we started talking about that maybe 3 quarters ago, and it was an idea and we were working on it. We had several customers engaged. And we signed our first NBM, I think, was in January. We were taking small steps and really warming up and learning on how to do this. And it was a shorter one and it was with a strategic customer, obviously. But we then fast forward to where we are today. And I think the relationship with these customers has dramatically changed. I think as you and I have talked, we've been very selective. We didn't want 20 or whatever. We wanted to focus on those few customers that we thought are going to win in the market that are very, very strategic for us. And those customers that are going to continue to take more bits as we go along the journey.

So I don't know what the magic number is of where we want to be, right? We keep on learning. We have these conversations with customers and they tell us their needs. We work on financials. Again, we're working on an attractive business that's durable. That is the single most important thing to do, and we do both. Doing one or the other doesn't make sense. We want an attractive business that's durable. And therefore, as we negotiate with customers, we are constantly redefining where do we want that 50 2/3 to evolve into the future.

And then just within that framework right now, you have 50%, you're the 2/3 fiscal '27, fiscal '28. What about the one that's left over? Should investors think about that as any kind of upside above sort of whatever the pricing is within these NBMs? Is that sort of upside relative and it could be a function of the spot price or the contract -- broader contract price that's going on the market?

Yes. I think there are some customers which some of them just are never going to embrace the new business model, and we will sell bits that we have for them in the market at market prices. There is also a small percentage of our business that goes to very strategic customers that are just relatively small. And they just -- the new business model may never be a good solution for them. So we think there will be a proportion of our business that will stay in the non-NBM business, and that will be based on spot prices.

Okay. And then just a little bit on margins. So obviously, investors are captivated by your current margins. And you've said that even at these floor levels, which are within these NBMs, there's a floor and a ceiling. But even at floor levels, you're hovering around 80% gross margins, which is very, very attractive. When you talk about economic sharing mechanism at these, like just help investors understand what happens on the upside, right, because there's a ceiling, but then there is what happens on the downside. And then how do we think about the transactional part of that, which are non-NBMs?

Yes. So when we started to work with our customers and started to lengthen the time of these new business models, we thought, well, a fixed price is going to be tricky, because one of us is not going to be happy. If prices go up, we wouldn't be happy because we wouldn't capture it. If prices go lower, our customers would be uncompetitive versus what others would be paying in their markets. So we said, well, there would be a mechanism, this floor and ceiling around that fixed pricing. So think about it kind of parallel to the fixed pricing and kind of some upside protection and some downside protection. And the good thing is, as you've mentioned, we feel good about margins even at the low case scenario. The non-NBM business will continue to price at spot prices. Well, whatever we think -- the value we offer to our customers and what we can negotiate with them -- and even if you are an NBM customer and you want upside, obviously, we're going to see what is the right price for that business.

Okay. Which reminds -- brings us back to the fact that you've had a couple of customers come back to you after those initial NBMs. I think you've talked about that in your last call. So when we think about these customers who are coming back, are these more hyperscalers? Are these non-hyperscalers but still in the data centers? Are they edge customers? And then how should we think about these NBMs when you renegotiate them? Are they just incrementally higher price now? Is it the entire agreement that got redone? Is it part of that agreement that was just the upside? Just maybe help investors understand.

Yes. So there are 2 contracts that we changed, but there are many NBM customers that have come back to us. And some of them have just bought business outside of the NBM, which is great. Now we're strengthening this relationship, which moves us from being a tactical vendor to a very strategic partner. And I think that has many implications, including if they ever had to make choices, we are closer to them, and that will give us a preferential treatment. We expect that will give us a different treatment.

But to your question, yes, 2 customers have come back. One of them extended the duration and the other one just added a ton of volume to the NBM that we had initially signed. This was a very interesting contract because they all take time. Somebody was asking me, why do they take so much time? Well, because they are pretty big, right? It takes longer to buy a house than to buy something small, I guess. But these are big commitments and customers take them very seriously. And even when we were signing the first version, they started to talk about upside and they said, well, engineering is coming up with these numbers based on these models, inference usage models and the numbers are just too big, and we need more time to get them. Why don't we sign version 1?

And then we'll see what happens over time. And not surprisingly, they came back and said, "Hey, we actually do need more. We are ready to commit, and we changed that." So the contract is not super long, but these are robust contracts with a lot of different clauses that protect around supply and demand certainty, right? So there is a few things. The only things that changed in the new contract were quantities, and we did adjust price to the prices that we thought were attractive for them and attractive for us.

Okay. All right. Just while we're on the topic of data centers, and I know there's a lot of questions on data centers. But remind us, 38%, I think, exiting fiscal '26, which was your last quarter. What -- how do you think about the business overall? Do you think 40%, 50% is kind of where you think sort of end state where data centers and obviously, you have your retail business and then you have your edge business?

Yes. So we started late with data center, right? We -- when the financials were a little tougher, we had to make choices and we prioritize edge versus data center and just didn't have the right products. And all of that changed when we had BiCS8, which is an amazing technology. And then we brought these products into market with we call Carrera for our compute product and Stargate for QLC, right, so for storage. So now we have the right products, which are amazing in terms of performance, density and energy consumption. So we're very happy. We reported $3 billion in revenue from these businesses in the last quarter.

But as you said, we're only 38% of our business going to data center and the market is very soon going to be 50%. We're still -- in my view, we're still a little bit underrepresented in data center. We expect this will continue to evolve. But we believe in a portfolio. We believe in having presence across markets. We think that's the right thing to maximize value over term -- over time. And just like we're committed to our consumer business. It's -- well, some people talk about consumer at edge. We talk about consumer really what you and I would buy in an Amazon.com or Best Buy or those type of stores. And we believe in that portfolio, it's very important.

Yes. All right. If I can, just before I switch off from the -- when you're talking about QLC technology, you talk about data lakes, you talked about at your Analyst Day, you had some very interesting projections as well on how big this data center could become in terms of exabyte consumption for the overall industry. You do have -- and you're also talking about a cost structure that you're -- with your BiCS8, BiCS10, BiCS11. You do have Chinese competition, YMTC. They're very, very aggressive. They have been aggressive for the last few years. Are they also gaining a lot of hyperscaler attention? Like how do you think about that competition from hyperscalers, which is where a lot of the growth that you've talked about for the industry?

Yes. So I -- it's difficult for me to talk about their business because I don't know -- I mean, I know as much as you can read externally, there is some information through the IPO process, which seems to indicate to me or at least my understanding is they are more of a component selling a lot of components. So I don't know how much they are developing their eSSDs. What I can tell you is our relationship with hyperscalers is very strong and robust and growing. So we feel about our competitive position in the market despite of [indiscernible].

Yes. And just talk a little bit -- I know you guys don't like to talk about cost per se. But just as you think about your migration, BiCS8, BiCS10, BiCS11, how are you thinking about your moats? Cost obviously should be attractive for your hyperscaler customers. But broadly speaking, how are you thinking about your moats as you're transitioning on that technology road map from current BiCS8, BiCS10, BiCS9. And I know you talked about BiCS9 at the Investor Day.

Yes. So we don't think it's a good strategy to talk about our cost downs, particularly our negotiating prices with customers. Having said that, we do see some benefit in cost as we continue to innovate in our portfolio. As we drive new nodes, we continue to improve yield. And our single most important advantage is how much CapEx is required to spend to generate new nodes. Alper has this -- isn't it amazing when your R&D leader talk about CapEx efficiency, right? That's what you want. So he has this beautiful chart that proves that over time, we've been consistently spending less CapEx per additional bit, which makes us very competitive because then that flows through obviously lower depreciation, lower unit cost. So we think that, that's a competitive advantage, which goes back to our ability to continue to drive scaling.

So we're very focused on driving cost as a competitive advantage. We think it's very important. You could argue a little bit less critical right now because margins are better, but we are very focused on driving cost down. There are -- within our cost, as you track our unit cost over time, there will be mix impacts, right? As you move more of your business to data center, those products are more costly per unit. So the gross margins are still very interesting and attractive, but the unit cost could be higher because there is more components, more DRAM, more components in general. So there will be a mix element of that and DRAM and just as a few other components have increasing costs. So the -- what we call non-memory cost keeps on going up.

Great. And then you also have your CMOS, which is your bonding technology, direct bonding technology. Just help us understand, when you think about your mid- to high-teens bit growth kind of how you're thinking about overall. How much of that is a function of just the mix towards more of the bonded arrays? How much of that is sort of just wafer migration towards these higher nodes, stacking, et cetera?

Yes. So I would say it's 100% driven by node transition. So if you look at our history, we've been producing fewer wafers over the years from where we were all the way to a point where we started under utilizing the fabs, whatever, 18 months ago, maybe a little bit less. And then we've been ramping back up. But going forward, we don't expect that we'll need any wafer additions to supply the market. So all of this is driven by more efficient, more productive wafers where you can fit in more bits.

There was this picture we showed at Analyst Day where you saw one on top of the other, and you can clearly visible see it's just more efficient from a space point of view, right? So just every wafer can produce a lot more bits. So that's our single driver. It's not -- your bonding is very interesting because now you can be more productive in your R&D because now you can innovate on the array or on the CMOS. You can do one or both. So it just gives you a lot more flexibility back to your BiCS9 comment. But that's not how we drive productivity on the wafer. It's the array itself.

Okay. And then when you talk about BiCS10, I know you talked a lot about what was interesting from that with it was bandwidth, it was power efficiency. Like any one particular thing that hyperscalers are more focused on as you -- is it bandwidth? Is it power efficiency? Maybe if it's density, it's TCO? Like where do you think your technology? Is it all of the above where it's like...

I think it's 3 of them really. But if I had to say one that's increasing in importance is energy efficiency, right? And we know the reasons for that. But they are -- it's difficult to see them sacrificing one over the other. They want it to be an end. But if there is one increasing, it's that one.

Okay. And if I can just maybe switch a little bit towards the non-hyperscale or the non-data center customers, I think I should say. If you think about the edge, that's still a very important part of your business. You've talked about it, smartphones, PCs. Just as you think about maybe some return to growth here in calendar '27 in the back half of [indiscernible] from a unit perspective. And when you think also on top of that content growth, we obviously have iPhones launching here, the new iPhones, foldables, et cetera. How are you thinking about that end market? And what are these customers telling you about from a -- whether it's a return to growth on units or content overall for the market?

Yes. So first, I would say their financial results seems to be good, right, or very good, if I will say. Some of them reported last week and some have been reporting over the last several weeks. And I -- it seems like their revenue, their margins, their free cash flow seems to be pretty robust. So I think that the edge customers are adjusting to the fact that memory was subsidized and now it's no longer being subsidized. I think what's happening is that the low-end models, right, they are just not viable anymore, and they are really shifting towards the higher performing -- medium to higher performing models, which makes more sense.

So if your business model was built around very cheap NAND to subsidize your business model, that just doesn't make any sense. I don't think you want us to subsidize those business models. So we're not -- and so what gives us confidence that this will grow in the future. So again, once you eliminate the lower-end models, we believe that next year, we should expect to see medium -- low single-digit unit growth and maybe a little bit more on content growth, just particularly focused on the models that will remain, which are the medium and the higher end.

Okay. And do you see, I don't know, AI at the edge, maybe driving a little bit of that content growth? Is it too early? Or how are you thinking about that -- or your customer conversations at least that they're driven -- content growth is driven by AI on the edge there?

Yes, I wouldn't say we're seeing too much of that. AI could be a driver of replacement of all PCs, for example, but not a major factor. So far, I would see it more on the data center.

Okay. And then the mid- to high teens kind of bit growth that you guys have talked about, like when you think about the industry, and I know you have a market intelligence team that kind of runs through all these numbers. Just if the demand is stronger than expected, we obviously heard from Neoclouds, we have our own AI model here at Citi. We've upped numbers there more recently. If demand is stronger and all your competitors alongside SanDisk are kind of looking at this, I think investors always worrisome about overbuilding, over capacity, what if they're building for 35% bit growth or 25% bit growth versus mid- to high teens bit growth. What gives you the confidence that we're not going to repeat what probably we've seen in past cycles?

Yes. So I don't know what others will do, right? We track them as much as we can. There is a lot of external press related to that. What I can speak to is what we're doing. And we've been consistent over the last 24 months. We believe that the market will sustainably grow mid- to high teens, and that's what we're executing to. I think you can easily make the argument that you should invest for more. And then a minute later, you would find out that that's probably not a good thing to do. So we think that, that's sustainable over the long term. As you know, there is very little or nothing we can do to change supply over the short term. So really, you're looking at a 10- to 15-year horizon, right? And am I making a bet to increase our CapEx? And I don't think that makes sense for us at this point in time.

Yes. And investors also -- just to your point exactly that about demand, like why couldn't demand be higher than that mid- to high teens? Just I know you guys have been very consistent. I've heard David, even when he was back at Western Dig talk about NAND bit growth kind of in that range. What -- and since then, we've seen AI adoption, right? What gives you confidence on that long-term mid- to high teens bit growth? Why couldn't it be stronger?

It could be stronger. But then you are subsidizing business models that don't make sense, right? So it depends on -- everything is going to be balanced at the end of the day, they cannot be more demand than supply, right, by definition. So you're selling at a price that makes sense for your customers and for us, and we believe that we're achieving that now, which was not the case in the...

A little bit on HBF. You guys talked about it. That's interesting. Obviously, it solves a need or a business case is there for it. You've talked about it. You introduced it a couple of years ago, then at your last analyst event, you talked about more developments in it. Just remind investors again, where are we? When should we start to see this as really in your model, in the business model itself?

Yes. I love the quote from Alper. He's in love with HBF [indiscernible] thing. So we're very interested in the technology, right? So the first thing we needed to do was build a consortium. So when we announced that we're working on HBF, we saw a few companies raise their hand, particularly SK Hynix called us and they said they were interested in partnering with us in developing the standards. So we're not developing products together. We're just developing standards. And SK Hynix obviously knows a few things about HBM being one of the largest producers of HBM.

So we welcome that and particularly given that it requires some changes for our customers. So having 2 strong players partnered together made a ton of sense. And then since then, Google and Meta joined the consortium, right? I mean -- so you have 2 manufacturers and you have 2 potential or 2 users of this technology that could be very interesting. So I think that was one of the key steps. The next step is to get products in our customers' hands. So what we will do in 2027. And then we'll see from there, right? We'll see the reactions. We're encouraged by the conversations and the relationships we are having with our customers, but we'll see how things evolve once they test the product and get more excited.

And then just, again, for those who may not have be fully aware of what HBF does, like do you see that on these workloads like more sort of a DRAM replacement or HBM and transitioning more of the workloads to HBF, like in these customer conversations that have evolved over the last few, I don't know, weeks, months?

Yes. The idea is to get much closer to the CPU or GPU to be able to manage all the inference memory that's needed so that models are useless if they need to compute everything over and over again. And NAND is a perfect way to solve the problem and HBF has the ability to get much closer to the CPU, GPU, as I just said, and serve that purpose. So that is the intent. So you saw a problem without any bandwidth trade-offs and you get a lot more density into [indiscernible].

Okay. And then KV caching, again, I think SanDisk put out some projections out there in terms of the demand for KV caching at your Investor Day, I think it was roughly 35% of the workload by 2030. So -- but it's -- I think you also talked about it just being very sensitive to a whole bunch of factors, whether it's token size, how many users, the parameters that are being used. Just what are puts and takes to kind of that output that you put out that it could just be 35%? I mean, it could be much greater, could be lower? Like how are you thinking about overall NAND demand? And where you talked about data centers could consume as much NAND as you produce today?

Yes. I mean the model is very sensitive. And every time we've revised it recently, we've gone up. But I think to your point, the 2 variables that are most sensitive is number of concruent users, right? How much is AI being used at the same time? And second, for how long and how much data do you need to save for your context in your KV cache. And those are the 2 most critical variables, and they seem to be going up, right? So that's very encouraging as more AI is used, customer service, software development, whatever, right, all these areas, the more context you have, the better job AI does.

Okay. I'm going to pause here. Now it's a full house. All right. [indiscernible] question. Can we please bring the mic?

I understand there are a lot of demand for the LTA with you guys. But I'm just wondering, so far, the NBMs are mostly with the cloud customers. But are you seeing any demand from edge customers are asking for the LTAs? Because given that we're hearing from Asia that Apples are looking for 3-year to 5-year LTA with NAND makers even without a price cap?

Yes. We're open to new business models with any customer, right? And we've had conversations with anybody that wants to engage with it. The only constraint is it has to be within the framework that we talked about. There has to be a time commitment. There has to be a volume commitment. There has to be price and there has to be a financial guarantee. That is very important. Otherwise, we're back to the old model of negotiating pricing every quarter. And frankly, it's -- we have no interest of doing that if we have the alternative, which is to do a new business model with those elements. We do -- we said at Analyst Day that we do have customers, at least one customer in the edge, right? We didn't say who it was and -- but we're very excited about that.

Right. And you did also say hyperscalers, right?

U.S. hyperscalers.

Okay. All right. Well, let's talk about the fact that you have a whole bunch of investments as well. I mean I think you've talked about Kioxia just had some announcement at [indiscernible] had an announcement about how much to invest. Can you just remind investors again strengthening your positioning here in AI and the fact that you are under-indexed to the data center market, you're trying to at least be on par with the industry. What -- how many -- what should investors think about the pace of investments and the ones that you recently announced, do you think that would help you to get to sort of that 50% mark for data center?

Yes. So we're investing capacity towards the mid- to high teens growth that we've been talking about. The dollar amount will increase, right? And why? So if you go back to where we started the BiCS8 transition, we come from a world where we were underutilizing our fabs. So by definition, you have tools available. So you -- when you transition, you first do the most efficient transition, right? So it was less dollars needed. So the dollars are increasing a little bit, but still within the percentages that we talked about, percentage of revenue, and it's still for the same percentage growth on NAND year-over-year. So we haven't changed that. We just did the most efficient ones first and then you transition your model going forward. And that should enable our ability to continue to drive our data center...

Okay. And then when you think about HBF and some of these newer technologies that maybe could drive substantial growth, when you think about any incremental capital investments because it seems like HBF wasn't really included in that long-term target model that you shared. And I think investors were looking back saying, well, what if it does take off, are you then going to be investing some more? Or is that still within that, I think, CapEx as a percentage of revenue that you outlined at the Investor Day?

Yes. So we covered the OpEx for HBF. We don't have the revenue, the gross margin, the gross profit and the CapEx. So we need to figure that out. And I think we need just a little bit more time to understand the customers' acceptance, their needs, how fast they're willing to go. So we'll figure that out, and we'll be very clear about that with you. The good thing is the foundation is BiCS technology. This is BiCS8. It's not that we need to invent a new node. We have the node. It can't be produced in existing tools. So it's not something completely out of what we're already doing because it's BiCS8.

Okay. And then even on the OpEx side, Luis, I mean, you've come down, I think you said like, what, 5% or so, if I'm not mistaken, for OpEx as a percentage of revenues. Like how comfortable are you with that? Why couldn't it be higher if some of these technologies like HBF? I understand it's kind of BiCS8, but still there could be a lot more development work that needs to happen.

Yes. So we're funding our R&D to the point where we believe we're creating value. So we're fully funding our BiCS program, right? We jointly agreed to the program with Kioxia, and we fund that to the level we think it makes sense. We also fund [indiscernible], right, who takes the NAND and develops products for eSSDs and for client and for all of that. So that's a totally different organization, which we also fund. And then we're funding 2 big bets. One is HBF and the other one is this new memory technology that briefly -- was briefly discussed by Alper in the Analyst Day. So we're investing in our core, 2 eggs, Alper and [indiscernible], NAND and products. We're investing in to kind of extend the core into the future, particularly focused on AI with HBF.

So we think we're properly funding the business. We also made an investment to extend the JV. We feel good about that. We made an investment in Nanya to get the DRAM. We invested $1 billion, something that's worth probably close to $2 billion right now. right? So it continued to go up after the quarter closed. So we're happy with those investments that we're making. But I think if there was an attractive investment, we'll definitely consider that. I don't want you to think we'll be constrained by a number in the model, right? We believe that's the right model, but we will continue to invest in the business to maximize value creation.

And then one of the things that I heard at the Memory Summit, there are some alternatives to talking about KV caching, there are some alternatives that are out there. I mean I think one is the camp that you're in. I think Kioxia has some alternative. Just there is so much change that is happening. Like when you think about your own investments, how do you make sure that you're investing at the right pace or the right technology just because there's so much change that's happening on the architecture itself.

Yes. So our teams are well connected externally. They are looking at all these trends. And we think we're balanced right now in terms of where we want to be, but we'll keep on assessing. And if there is ever a need to make a change, we'll share that with you. But we think we're properly funding the business.

Okay. And then coming to capital deployment. You had a big announcement at the earnings as well. You're talking about essentially generating a lot of free cash flow here as you did in your last quarter, I guess, just returning it all pretty much to shareholders the way I see it. Just -- I think there was a couple of questions on the Investor Day where people were just trying to understand when you mean excess cash, is that really just free cash flow and how to think about it? And the pace of CapEx -- the pace of capital return, sorry, I mean you had a big buyback in the last quarter. Should one expect at similar levels as we are going through fiscal '27?

Yes. So we generated $5 billion last quarter. We bought $4.5 billion. So that's pretty much all excess cash. Maybe we can't get the forecast perfectly, but we think we like how we're executing. I think one of the things you should take is whatever we tell you is what we intend to do, and we're executing that, and we'll do that. The Board continue to authorize more spending, and we'll be updating you every quarter as we do that. We did talk to many investors to try to understand what was the best way they thought. We should return cash to shareholders. And we think that at least at the current point in time, we believe that share buyback is the best way to do it, the most tax efficient and it makes sense.

We may consider different things down the line. I'm not saying we're going to make a change or anything, but we're always open to continue to adjust our strategies going forward. But yes, we will continue to buy our shares back our excess cash. And I think that's an important message that we should have.

Yes. Anything on M&A, like that's something that you think might be needed in the portfolio? I'm not talking about gigantic M&A, but even small tuck-ins here and there, technology purchases that you think would be attractive.

Yes. We like the portfolio, but we'll always keep an eye on and be open-minded about opportunities.

Any other questions in the audience? Okay. Maybe I can ask a little bit about the market sizing itself. I think your market team talked about -- I think you guys talked about $500 billion as a market in calendar '27. I think people were immediately doing the math. So what is SanDisk's share of that. But just talk to us about how you kind of got to that $500 billion. And then on top of that, I think at least in your target model, you talked about mid- to high teens revenue growth through the long-term model, which was in line with the long -- mid- to high teens bit growth, which assumes like a pricing environment that is very, very favorable beyond fiscal '27 or beyond calendar '27 as well. So just help us understand what's underpinning that market size. And as you think about fiscal '28 to '30, a very favorable pricing dynamic that's kind of baked into your long-term target model?

Yes. I mean it's very interesting because this industry has been around $50 billion for a long period of time. And we talk about cycles, but those cycles were small variations within that $50 billion. And then all of a sudden, AI and these data center players come in and the market grows from that $50-plus billion to $200 billion, $300 billion, and we continue to see a path to continue to grow. So AI is driving that. AI is -- data center is by far the single most important driver. They continue to invest CapEx. Every time we hear an update on CapEx, those numbers continue to go up.

And they seem to be generating significant revenue growth out of that, right? Their revenue growth keeps on accelerating. So that's the biggest driver, right? More volume and prices obviously have changed, and that's enabling that growth. But we're assuming there, obviously, data center continues to be the bulk of the growth. And then Edge after this reduction in units then goes back, as we were just saying, to grow on the low single digits in units and content.

Okay. And any -- like are you tracking any milestones in inference adoption or anything you're thinking about that?

We track a lot of them, yes. We are -- we want to understand the market as much as we can. And frankly, the more data we get is from this relationship with our customers, right? And the more engagement we can have and how exactly are you using our products and what are you seeing? So that helps us a lot.

Okay. Last 5 seconds, Luis, like what do you think investors are still missing about the same?

Yes. I think the key question is sustainability, right? And it's difficult to prove a negative in the future won't happen. So I think our focus is to make sure we execute, right, and build that credibility so you guys can change your view. But that's what we're trying to do. And I think the long-term models, the investments in our supply chain, hopefully, those give you some confidence, but that's what we're trying to do.

Great. Thank you, everyone. Thank you, Luis.

Thank you. Appreciate it.

Thank you very much for joining us today. Before we begin, please note that today's presentation will contain forward-looking statements based on management's current assumptions and expectations, which are subject to various risks and uncertainties. These forward-looking statements include expectations regarding our technology and product roadmaps, our new business models and multi-year customer partnerships, our business plans and performance, market trends and opportunities, and our future financial results.

Please refer to our most recent annual report on Form 10-K, our quarterly reports on Form 10-Q, and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations. We will also make references to non-GAAP financial measures today. Reconciliations between these non-GAAP measures and the most directly comparable GAAP measures are included in the earnings releases for the relevant periods and in the appendix to the presentation materials, which are posted in the Investor Relations section of our website. Please welcome Vice President, Investor Relations at SanDisk, Ivan Donaldson.

Thank you very much. I just want to say thank you to everyone for being here today. I've been in this industry for 22 years, and the journey with SanDisk has just been astounding. So it's been an amazing ride. We have an amazing management team and Board of Directors, amazing employees across the globe, and we're just really excited to be here today. I'm going to talk a little bit about the agenda just really quick. So we'll go over the, obviously, company overview, strategic vision and future for the company.

Then we'll go through the technology roadmap or really the way I think of it as our innovation engine in the company, which is just astounding. We'll also take a step and look at the industry transformation, what's happened, how we got here essentially, which is pretty phenomenal. And then we'll go into a deep dive of the AI infrastructure, essentially why? Why is this happening? Why are we seeing such a step change in demand and where we see that going forward? What are some of the key variables and dynamics for that?

And then followed by the financial model from Luis and really trying to deliver that we have a tremendous opportunity to drive shareholder value well into the future, and we hope that's going to be the takeaway from today. And then at the end, Alper will come back up and talk about sort of future innovation roadmap and where we see that going to conclude, again, where we see some future technologies and emerging memory opportunities, followed by Q&A. We tried to take into account a lot of your questions over the last year of what you guys are most interested in. So hopefully, we'll get to all of those. So again, I appreciate everyone being here. Thanks for your support.

[Presentation]

Please welcome Chairman and Chief Executive Officer at SanDisk, David Goeckeler.

All right. Welcome. It's great to be back here in the room where we launched this company 18 months ago. A lot has changed in that time. And today, we're going to talk about the company going forward. I can tell you, I was -- as I was talking to some of you as we were preparing or just this morning and gathering, we are talking through a lot of stories of stuff that has happened over the 6.5 years, whether it's roadshows or conversations we had about this franchise. I joined Western Digital, of course, in March of 2020, actually the same week that COVID started.

That's probably people know about that a lot more than they know about me starting at Western Digital. But we started on this conversation, I think, that's been going on for almost 6.5 years now about where we were going to take this franchise. I thought from the beginning, this was just an unbelievable franchise that we could really unlock the value, but it took some time. It was going to take -- they're big markets. It was going to take a lot of moves about how we got there. And I can tell you, as I stand here today, I feel like I've finally gotten to the starting line of where the real value creation is going to happen.

Now that may seem like a pretty big statement based on what's happened since the launch. But I think when you walk away from here today, you'll see kind of the conviction we have in what is really the earnings power of this franchise and why going forward, we finally got things structured in a way we can really start to reveal that on an ongoing basis. All right. I'm going to set some context here on kind of how I think about the business, kind of the big picture of how all the different things we're thinking about, how we integrate it. This is a business where you can't just think about one thing. There's like 2, 3, 4 variables that are in motion at all times, and it's about getting them balanced and how are we going to make changes not only to the technology, we're always changing the technology.

Like, that's something that goes on and on all the time. We're world-class at that. And we're going to hear from the people today that are really driving that. And that's an incredible story all by itself. But then also how are we thinking about the business model around that? How are we structuring the business? How are we changing the relationships with our customers? How are we allocating capital in the business, all these kinds of issues to create just on an ongoing basis, relentlessly create intrinsic value in the franchise that will be revealed as we continue to execute the business.

All right. So that's really kind of -- there's kind of 3 big categories to that. The first one, look, I was going to -- when I was putting to get this talk together and I was thinking about this, I was going to spend some time going over what we committed to back the last time I was on this exact stage. So what was that? February of 2025. We made a bunch of commitments on what we were going to do with the company. I think it's just fair to say, like, it went pretty well. I think most of the things we said we've delivered on. We talked about we wanted to win in data center. We needed to establish data center as a major growth pillar of the business.

That had been a long, I think, an issue with the company for quite some time. I think we're getting there. I think we just delivered significant outsized growth in the last fiscal year. And I think you're going to see that continue as we go through FY '27. We talked about we're going to focus on the consumer business. And hopefully, you stopped by outside and saw the products. They all look very different. It's the same products, different branding, showing up a different way. This is just an incredible brand that I think is in a bit of an unpolished gem and we're continuing to work on it.

We've gained 2 points of global share in that business. We'll talk a little bit about that business today about why it's so important to the business model as well. We made some statements about -- we made some controversial statements at the time, quite frankly, we got up here and we said, "Hey, pricing is going to inflect positive in the second half of the year." And that became a big talking point. No, it's not; yes, it is; no, it's not. Yes, it is. It turns out when we got to the end of the year, things were going in the right direction. And if anything, we significantly undercalled it.

And so anyway, I think things went well. Those of you that believed in the company back then and invested in us, we really sincerely appreciate that. We take that very, very seriously to be good stewards of your capital. And we worked very hard and you got a good return, and we're very happy about that. But that's the past, right? We're not going to talk about that. That's -- we can't go back to that point in time. That was a special point in time. It's not coming back. But what we can do is talk about going forward and how we're going to create value from here going forward.

And I can tell you, everybody you're going to see up here on stage just has unbelievable conviction that this franchise, we're finally starting to reveal the true earnings power of it. And that earnings power is going to go on for a very, very long time. So we're going to -- we'll talk about that. All right. So let's start with this first. When we say we have this capital allocation strategy, number one, we're going to invest in the business. That's always the most important thing, invest in the business. So what are we talking about?

And I'll preface this by saying there are a lot of things we committed to doing back when we launched the company, I think we were -- like I said, I think we were largely successful in making progress on those. But there's a lot of things over the last 18 months and especially the last 9 months, where we were presented with some opportunities to really change the business, a lot of momentum, and we took advantage of those. And I think we have fundamentally restructured the business. And we kind of put a pin in the map on this day early this year, realizing we could see what was happening in the business.

We could see we are restructuring the business, and we're going to need to stand up and explain it all to you because it is very different to what it was back in February. So anyway, with all that said, let's talk about all the things we did and kind of how we thought about investing in the business as we went. So, the first thing, the most important thing, we're a technology company. Like, if you don't have great technology, you could probably, like, not -- you shouldn't be doing what we do. You always have to have unbelievable technology. So this is always the most important thing we're going to invest in.

And I think we're in the best position we've been in, in a very, very, very long time as you look across the portfolio. Something I've been doing my whole career is investing in technology and thinking about this multi -- what I think of as a multi-horizon innovation investment plan. You can't just think about what's going to happen next or what's going to happen this year or next year. And you can't just think about what's going to happen 10 years from now. You got to think about all of it. And how do we invest across this entire horizon to make sure we have the right technology today, tomorrow, 5 years from now, 10 years from now.

And this is where I think we've -- just when you look at our business, it starts with what Ivan said, kind of the engine of the company is the BiCS roadmap, is the fundamental NAND roadmap. If you don't get that right, it's kind of hard to make up for it at the system level. It's kind of hard to, like, hide that. You have to have really good NAND technology. And we're constantly investing in many, many generations of NAND technology. We don't talk about all of them all the time. We just announced BiCS10, I think, a couple of weeks ago.

By the way, we announced BiCS9 yesterday, and I'm sure everybody is confused. Like, why did you announce BiCS10 like 3 weeks ago and then you announced BiCS9 yesterday because 9 is before 10, and we thought you would have announced that first. So Alper will explain that. I think it's a very important point to understand, like, the technology strategy is changing based on the fundamental BiCS architecture, and he'll explain to you how we now have multiple ways we can move this technology. So we're investing in BiCS9, BiCS10, -- there's actually people working on BiCS13 right now.

So one thing you should have confidence in, and this is with our partner, Kioxia, we have a long roadmap of really strong fundamental NAND technology. One of the big advantages of the JV, the big benefits of the JV is we invest together on R&D. And together, we're 1/3 of the market. So that means we can invest as much or more than anybody else in the market in making sure we have the best technology. So that's going to be there for a very long time. Alper will go through that. We also build the systems capabilities. The BiCS investment gets you through the wafer through the fab.

The wafer comes out of the fab, I have to do something with it. You could basically -- we could just go sell all the wafers, but we actually turn them into systems ourselves. So we have people working on all of the controllers, how to build SSDs, how to build all the different products in the markets we operate in across consumer, across edge and now across data center. So you're going to see Khurram up here today. He's going to be talking about AI in the data center. And it's really his team that builds all this stuff. So just enormous systems expertise across all of these markets.

And I think this is one of the big -- if you look at the foundation of the company as a technology, one of the positions that we now are in and one of the reasons I have so much conviction about the future is we now have optionality across the entire market. We have a very unique consumer franchise. Edge, we've always been very, very strong, PC, smartphone, IoT, all of the things on the edge. And now we're very strong in the data center.

So we've got all this optionality, and we're going to keep all that optionality, right? One of the big things of the strategy of the business is make sure that all of this technology remains very relevant, very on point, to continue to drive this innovation across all these different markets. So that's always going to get invested in. Now we like pick our head up a little bit, look a little further down the field. One of the things we talked about 18 months ago as we announced this product or this strategy around High-Bandwidth Flash. If you would ask me, like, a lot of stuff has happened in the last 18 months. This may be one of the things that is actually I'm the happiest about.

We basically stood up here and said, "We're going to build this thing called High-Bandwidth Flash." And I think the reaction was pretty much, "What are you guys talking about? Like, nobody has any idea what you're talking about." what is High-Bandwidth Flash? And we only knew high-bandwidth memory. But we were kind of targeting this idea that, "Hey, when we get to inference -- like, AI is a massive opportunity." At that point, all the focus was on model training, and appropriately so. But I think we were looking down the field, our team -- this wasn't me, this Alper and his team -- a tremendous amount of insight to see that, hey, at some point, we're going to move to this inference phase.

You're going to have to scale this, and we're going to have to come up with a different memory architecture or storage architecture for inference to really scale. And we have an incredibly important technology that we can bring to the party. And so we announced it, and we said we're going to form an ecosystem, and we are going to start building this product. And Alper will be here later. That's the last talk that we're going to have today. He'll be here later and give you an update where it's at. But I think if anybody was at FMS last week, they probably saw there was a lot of activity around High-Bandwidth Flash. And even last year at FMS, High-Bandwidth Flash was awarded, like, the most innovative technology in the industry.

So that's a horizon. And I know a big question is going to be when is it going to ship and all this kind of stuff. But -- and we'll get to that. We're not going to get to that today. So it's a little bit of a spoiler alert. But we're getting there. I think the ecosystem is being developed. People are coming to the table, a lot of very good discussion. And so we'll -- that is on the horizon. And then we look even further down the field, we had this idea of 3D Matrix Memory, a little longer-term project, continue to make progress. So we're basically -- this is like the first priority, invest across all of these things and make sure they're all healthy and bring them to market to the extent that we're getting the feedback that they resonate with our customers.

All right. The second thing we talked about was get to this idea of a cash positive balance sheet, right? It's not exactly a novel idea, but it's like we need to get the debt out of the company. And I think this is something we're very happy about last -- it happened faster than we probably thought. Market -- we got a receptive market. And we got to this position where we don't have any debt. We have significant cash reserves. And one of the things Luis is going to talk about later today is this franchise, I think one of the things you guys are all seeing, when you start to scale this franchise, it really is good at generating free cash flow. That's a good -- that's good, right?

That what we think our job is to generate free cash flow for all of you. So we're in a position where we can generate a lot of free cash flow. What are we going to do with it? Luis will talk about that when he gets up here. But we feel like we're in a very good spot there. All right. Now there was -- that was kind of like things we need to do every day. But once the business started turning, we started thinking about, okay, what else can we do to invest in this business? That's what we want to do first. First thing we want to invest in the business. So the way I think about this is we constantly go through a process of how do we systematically derisk the business? How do we systematically make investments? We're basically derisking the future.

And we -- there are a couple of things that were a big part of that, right, that we were able to get done. First one was extend the joint venture. It's one of the first things you saw us do, we invested over $1 billion with our partner, Kioxia, which was a recognition of the scale of this joint venture. But we -- one of the first things we spent our money on was making sure we had production of NAND from 2030 to 2034, right? That was very important to us because we have a tremendous amount of conviction in the future of this franchise. And if you want to be in the NAND business, you need to have a NAND fab. That wasn't always the case, by the way. Like, part of the issue with the industry in the past is you could procure NAND very inexpensively because the people that own NAND fab seem to be selling at prices that were not basically the marginal cost.

I think that world has gone, quite frankly. My personal view is not coming back, and it's going to be very difficult to be in the NAND business unless you have access to a NAND fab, which we do. And we have it at scale, and we have all the R&D benefits of that and all the manufacturing benefits of that. The other thing I hear sometimes is, oh, NAND is not that hard, NAND is a commodity, anybody could do it, right? Well, if that's -- if you actually believe that, I would encourage you to go to Yokkaichi and take a look at the fab that's there, right? And if you want to duplicate that yourself, find a lot of money, and I'll see you in about 10 years, right?

It is extraordinarily difficult to be in this business, and this JV is a huge part, a huge strategic asset for us, and we took the opportunity to extend it when we could. Now one other thing we did in this kind of derisking, we knew that, look, we are -- we want to play in the data center market. Why do we want to play in the data center market? Well, it's -- you guys know all that. It's a very attractive market. But also, it was the market that's going to help us change this dynamic with our customers. We want to go from this kind of negotiate price every quarter -- this kind of highly transactional, highly volatile business. And we want to turn this into a business where we kind of dampen that cyclicality.

We have more long-term relationships. We have more long-term visibility into what demand is going to be. And the customers that are most likely to do that are the data center customers, right? And there's a lot of reasons for that. We can go into that later if we want in the Q&A. But we have a willing partner that wants to go down that path with us. So if you're going to be big in the data center business and you're going to grow that business, you need access to DRAM, right? And we're no different than anybody else, right?

The DRAM market is very tight, as they say. And so we needed to make sure if we're going to go to our customers and say, "Hey, we want to strike a 5-year agreement on selling enterprise SSDs, and we're going to put this huge contract together that's worth tens of billions of dollars," we need to make sure we have access to all the pieces to actually fulfill that contract. So this became extraordinarily important to us. It maybe wasn't as clear to all of you at the time that we were putting all these building blocks in place that was leading to this different contractual relationship with our customers, but that's kind of what we were doing. So we took -- we had the opportunity, we took a 4% equity stake in the company. It's worked out okay.

I think we just took -- Luis is smiling down there, our CFO. I think this quarter may be one of the few times where our GAAP earnings are higher than our non-GAAP earnings because we recognized, like, an $800 million gain on that investment. So that's been going well. But we didn't invest in it for the return. It's great that we get that. We invested in it because we need access to the technology. So look, I think the net of all this is we've taken the opportunity over the last 18 months to really make sure the foundation of the business is just incredibly solid. We have the right technology. We have the right roadmap. We have the right innovation. We have the right relationships. We have access to all the things we need for years and years into the future.

And then on top of that, we're going to think about how do we change the business model. How do we get this franchise where it's sustained value creation over the long term. And there's a lot of questions about that. I remember when I took this job, we went through the separation, some of you were actually pulling me aside saying, "Dave, what are you doing, right? Why are you taking this job? Do you realize this industry has never made any money?" And I'm like, I got it. Like, we'll figure it out, right? We're going to get there, right? We can change things and we can get a better outcome. That's one of the things we really believe in -- change things and get a better outcome. So how do we think about that, what we're going to change?

This is kind of how I think about it. The 3 imperatives for sustained value creation, right? So there's not -- I've talked about these before, but I'm going to just go through them a little bit. Number one, we have to increase profitability. I think if you go back to that time of February of 2025 when we launched the company, this was probably the big debate. I actually love all you guys because there's always a debate. No matter when we have a conversation, there's always a debate. And there's always a debate about something. And as soon as you get past that debate, there's a new debate, right? So there's always a debate. But the debate, I think 1.5 years ago was, could you get to profitability where you needed to be.

The debate today is the second issue: how do we reduce cyclicality. Now it's like, oh, okay, Dave, like, we get it. Like, we understand profitability is at a level it's never been before, but it's just a matter of time. Just for a matter of time, faster things go up, the faster they come down, all these kinds of things. So it's like now it's all about cyclicality. And we're going to talk about that, right? I think we're doing some things. Again, over the last 9 months, we've been extremely intentional about the way we run the business and the way we structure our relationship with our customers to try and reduce cyclicality. We're not saying cyclicality is going away, right? The whole world is cyclical. There's a business -- the whole business is cyclical.

What we're trying to do is get this wild cyclicality of this business, just dampen that down, get more sustained value creation. And if you do those 2 things, in every business, you got to grow. You got to get consistent revenue growth, right? And this is hard. In most businesses, this is hard. It's hard to grow. Like, usually, you run out of TAM. And so you have to go acquire and do all kinds of things. And we'll talk about in this business, it's actually quite different. All right. So let's dive into these just real quickly. Oh, one last point, what I said earlier -- you have to do this across all time horizons. It's not about maximizing value for the next 2 weeks, right? What's pricing going to do in the next 2 weeks? That's important. You got to do that.

You also have to do it across the midterm, the long term. So we're constantly thinking how do we balance these things across all time horizons to get to this point of sustainable value creation, right? So it's important. So you got to think in 2 dimensions. And a lot of the questions I get, a lot of the questions we get are -- tend to be about one of these things independent of the other 2 in about a certain time horizon. And just so you know what we're always -- when you ask us questions, what we're always doing is trying to translate your question into, okay, how do I think about that question across all of these variables in all time horizons and give you an answer that makes sense because answering point questions doesn't really help advance your understanding of the whole franchise of what we're trying to do.

All right. Let's just go into profitability. Like I said, honestly, I think profitability was always the one of these 3 that was just like hiding in plain sight. Like, when I was managing the franchise when they were together, and we went to separate the company, it was this issue: will this business really ever create value over the long term? And to me, that seemed like I just didn't have a question about that, right? Because I thought the intrinsic value was always there. The question was, could you get the business model right, the way the business worked, the way you engage with your customers -- like, could you change that to actually let this intrinsic value come out. And it really kind of starts with this kind of observation. It's kind of a very simple observation, but I think some people forget it sometimes. We own the whole stack. We do everything, right?

This is not a fabless semiconductor company -- like, that's impossible. And so we own the NAND IP. So the fundamental IP it takes to build NAND. What I said, one thing you should take away from the fact that I just said we have people working on BiCS13, which is like going to be launched sometime, I don't know, don't come back and ask me 100 questions about I just said when BiCS13 is going to be launched, but you're talking like well beyond 2030. People are working on technology a decade into the future, right? So this is like, just on its face, extraordinarily difficult to do. And there are hundreds of engineers that have dedicated their lives to building some of the most sophisticated semiconductor technology -- and that's just one part of this chain, right?

So we have all of that. And we have it at the largest scale, again, because of our collaboration with our partner, we have it at the largest scale of anybody in the world. And if somebody that's managed technology franchises for decades, very large technology franchises, your market share makes a big difference in how much you can invest in R&D. You can basically invest R&D commensurate with your market share. So what that says is when you get bigger, you should get better if you're doing your job right. And I think you're seeing that show up. You see that show up in our roadmap, and Alper will go through that. Once you have the NAND IP -- we don't call somebody else up to, like, build wafers, right? We have our own fabs with Kioxia. Like I said, they're quite spectacular.

They're some of the largest fabs of the world. The scale we operate at is just incredible. So we do all the front-end manufacturing. Wafers come out of the fab. What I said earlier, we have another engineering team, hundreds of people that are working on all the systems expertise. How do I take this wafer, cut it up into die, put it in an SSD, put it in an enterprise SSD, a client SSD, put it in something that can go into a car, whatever it happens to be, we have all those people, too. We're doing all that work. And as I said, across all markets: consumer, edge, and data center. And then we have the back-end manufacturing as well. We go to Malaysia -- we have a big factory there. We have a factory, again, with a partner in China. We do the back-end manufacturing, too. And then, of course, we do the whole go-to-market piece. So -- there's a lot of debate. I said it. I'm adopting all of your language. There's another debate, right? So there's this thing like, oh my gosh, your margins are so high. Your margins are higher than the fabless guys.

Well, we do a lot more than the fabless people, right? Nothing against them. They're great companies, incredible companies. But when you look at this, there's just minimal profitability leakage across this whole model. So that's why I say this thing has been kind of hiding in plain sight the whole time. The issue was the business practice was wrong and you never could see it. And not only was it there, but we've been doing this for like 25 years. It's like crazy. Like, we've got like 25 years of paying engineers, like hundreds of millions of dollars a year, right? They're not -- these engineers, they're not cheap. They're like expensive.

They're a little bit temperamental sometimes. They can be hard to manage. They're wonderful people. I'm one of them, but it's like not something you just roll out of bed and do. So we've been investing in that for years, decades, decades, decades. And quite frankly, one of the things I see today when people say, "Oh, like, this is just memory, it's a commodity," I'm like, yes, you try it. Like, you try to do this stuff. It's like incredible what people are doing. And so we've been investing in this for like decades. In fact, Alper will be out here later. He leads this team. Most of the people -- a lot of the people in that team have been at SanDisk, like, way before the Western Digital phase, right?

That was just like a little phase of the company. So we've got like a huge amount of expertise there. And then on the manufacturing side, we've been investing billions of dollars. Again, for 25 years, we've just been investing billions and billions of dollars of building out this incredible scale manufacturing capacity. So in many ways, like I said, I think this profitability thing was just hiding in plain sight. This thing is like a coiled spring that's been compressed for like 25 years, and it really hasn't produced and now it's producing. And now it's about how do we sustain that over long periods of time.

And again, I think -- hopefully, one of the things you take away from today is that the way we're engaging with our customers, the way we're -- the level of strategic engagement has fundamentally changed that allows us to completely reveal the profitability of this franchise. Now, a little bit, what did it take to get here? Like, what was the -- how did this get unlocked? Again, if I go back 2 years ago, it was like, oh, Dave, can like -- can NAND ever be 50% gross margin again? Like, yes, of course, it can, right? So I think we've kind of put that to bed. But the issue was we just have to proactively manage this business from the supply side.

I think '23 showed us -- like, the great meltdown of the industry showed us that if you manage this from like, "I have new technology, I should release it," that's going to lead to a bad economic outcome for us. And I'll go through a little bit of that later about why that's the case. So, a little more proactive supply management, and pretty soon things come into balance. And this next statement may be a little bit controversial. I actually think the market is adjusting to these dynamics quite quickly. And I think that's one of the things we're here to talk about, right? The market is changing actually quite quickly in the way we engage with our customers around securing future supply. It's moving rapidly from this quarter-to-quarter to multi-year timeframes.

All right. Let's talk about cyclicality, the current debate, right? So how do we think about cyclicality? So you may be surprised, I said earlier, the first place I think about this is the consumer business. It's one of the reasons why the consumer business is so important, right? Why is the consumer business important? 351,000 points of sale around the globe, right? This global brand equity. You can almost go -- it's fun in this job, you almost go anywhere in the country, anywhere in the world, and, like, people know our brand. Last time I was in China, I -- like, there's always a photograph. I pulled them over and said, "Hey, open your camera." He opens his camera; he's got a SanDisk card in it. It was great, right?

So people know us all over the place -- billions of products sold in the last decade. The dynamic range of this company is incredible. We sell, like, a single product to billions of people, and we sell billions of dollars worth of products to single customers. That's kind of what the state is across the whole thing. But let's look at the financial dynamics of this. We never -- I don't think we've ever showed this chart before. But this is industry gross margin going back to the first quarter of 2017. So there's the cyclicality for you, and you can see the great 2023 washout that kind of impacted how we changed the business in a major way.

But if we plot consumer gross margin on top of this, you see that it generally tracks the business over almost all time periods, but especially when you start to see the down cycles, it insulates you from those, right, because it's just a broad-based market. So it's like this shock absorber on the business that allows us to generate a consistent level of profit on the base of the business over long periods of time. And you can see -- I don't expect '23 to come back, by the way. Like, I know some of you are waiting for that to come back. I don't expect that to come back. But you can even see even in the worst -- the darkest, darkest days maybe in the history of the industry, this business was still producing profit, right?

So a very important business, will always be an important business and it's a great asset to the company. But it's not enough, right? The issue is it's not big enough, not big enough to insulate the whole business. It's great. It gives you some protection. We need to make it bigger. By the way, that's the key of why we're taking a brand-first approach to this business. It's about growing it, right? Financially, it's a great business. And there's actually some little tricks behind the scenes that make it very profitable because we can use more of the wafer in this market than we can in other markets. There's lots of little things like that, that goes on.

But it's about growing this business now. And like I said, since we launched the company, we've gained a couple of points of share globally. It's hard to do. We've rebranded the whole portfolio again. Janet is here -- she leads the business -- stop by outside and see all the great new products. So it's not enough. So this is where we really went into this whole concept of new business models, right? And we've talked a lot about this last couple of quarters. We've tried to be very transparent on what we're doing, right, as transparent as we can be given these are like confidential customer relationships.

I'm not going to go through these in great detail because Luis is going to go through these in great detail. And he's the best person to go through them in great detail because he's actually the guy that's negotiating them. So that should tell you one thing about how important they are. It's like literally the CFO of the company. And now there's a lot of people around -- there's a big team -- but he's the one that's like got his hands on the steering wheel of what all the terms and conditions are and what we'll agree to, and engaging at this most strategic level with our customers.

So I'm not going to go through all this detail. Luis will go through it. The one thing I'll tell you, like I've just heard a lot of stuff about these agreements. Just incredible -- there seems to be people that are very informed about the way these work that have never read the contracts, right? It's incredible. I watch TV and people say, "Oh, there's more holes in these things than you can imagine." Like, really? Like, how would you possibly know that? Like, there's only -- there's like the number of people that actually read these -- you don't really have to take your shoes off to count the number of people that have read these contracts, like it's like just both hands, right?

They're just like incredibly detailed, incredibly strategic, very, very consequential, and they're different than what has been done before. We have a lot of confidence in that. There's -- there may be a lot of reasons you choose to invest in this company or not invest in this company, but don't choose because you believe something about this just based on the history of the way the industry has always worked. Take the time to understand how these things really work. And I think you'll get some of the conviction that we have about why I can make a statement, I feel like I'm finally at the starting line of where the real value creation is going to happen on a sustained basis.

Now, I will say this long-term customer visibility, demand visibility, and pricing stability is replacing this quarter-to-quarter world. It's a crazy industry. Everything is negotiated constantly. I only have to go back like 3, 4 quarters. Everything is negotiated. We come into a quarter. We're still negotiating with our customers on what they're going to take and what the price is going to be. We have like 3 months of visibility. And we're making investments that are 10 or 15 years long, we have 3 months of visibility.

Within 2 quarters, we've gone from 3 months of visibility to over 4 years of visibility. So what I said earlier, I think the market is adapting very quickly to these changes. I think that's 100% true. 2 quarters is very quick to have customers of this scale, and we can say things like 50% to 2/3 of our supply is under agreements, and we have like average visibility of 4 years. That happened in 2 quarters -- kind of amazing.

And I can tell you, even since our earnings call last week, the number of customers that are coming to us, and now they're proposing the agreement, not us. We started with we would go to them and we would propose like, hey, how do we put an agreement together, get more visibility? We got some people that wanted to work with us, and we got it done. Now it's turning -- not everybody, don't get me wrong, not everybody. But now I'm starting to see customers are coming in and proposing, "Hey, how about we do a 3-year deal? How about we do a 5-year deal? Here's what we'll offer on pricing. Here's kind of how we'll do the financial guarantee." So this market is changing very, very rapidly.

All right. The third part of this is there's still high-value businesses where customers are not going to sign multi-year agreements, right? We're going to stay engaged in these businesses. We want to stay engaged in the whole market. I think one of the very brilliant things about the NAND business, it's kind of an evergreen business. There's always something new, right? There's always something new. There's always a new device -- like, this is the magic of innovation.

Somebody is always thinking of something brilliant. We don't know what they are. But when they think of those things, if they have to store data, they're going to use NAND. So we want to have -- make sure we have some of our supply available to make sure we can play in those markets. There's very important customers and very important businesses that maybe don't have the scale. So there'll always be a way to play in this other part of the market. And quite frankly, there's also going to be some customers who just want to do business the way it was done a year ago or 2 years ago, and we'll continue to engage in those markets to the extent we have the supply to make it available.

So we're going to meet our customers on their terms. And we want to stay -- like I said before, we want to stay engaged in the whole market. And one of the things about this business, again, one of the things that encouraged me a lot about this business a couple of years ago when we were talking about coming to work here full time, was we have a spectacular set of customers. I mean, we have like the who's who of technology companies in the world is our customer base. So that's amazing.

We want to stay engaged with all those customers to the extent we can. All right. So that cyclicality, I know it's the current debate. I'm sure we'll continue to debate it, but that's kind of our view of why we do think this market is changing very, very quickly. And this is a very intentional strategy to try and change the relationships and kind of bridge this gap of, like, on the supply side, having 3 months of visibility, right?

But we have to make 10-year investments in fabs, and we end up in this vast middle ground where it seems like nobody is happy, right? Either the supply side is like scrambling because we're not making enough money to invest, or the demand side is upset because they can't get everything they want. It's hard for me to see how that strategy works for anyone. And I think we're now rapidly converging to something that is a little more sane and we can all get enough visibility to make sure everybody can get what they want.

Like, again, what I tell customers, like, you want to buy NAND, you're in luck. We sell NAND. And not only do we sell NAND, we have the whole stack, right? We have all the IP, all the production from beginning to end. So I think we're getting to a very different spot. All right. I'm going to cover one more thing. And I think as an investor, it is important. It's important for you to understand. We've talked about it before and maybe you already understand it, but I just want to bring it out.

We talked about growth. And again, my experience in running a lot of technology franchises, this is kind of the hard one, especially when you get to very good economics because when you get to good economics, you kind of start running out of TAM, right? And so you have to then figure out, how do I expand my TAM? How do I develop new products? How do I move into new markets? How do I go acquire different companies? All that stuff, I've done all that stuff. It's all expensive and it's very difficult. We have kind of the opposite issue.

Like, we said we're going to commit to grow [indiscernible]. We're going to grow supply. We're going to grow volume mid- to high teens. And then we debate, is that too big or too low, right? Oh, it could be higher. Well, it could be lower, right? It's like some -- last year, if we went to February of 2025, people would probably say that's too high, right? Pricing is going to go down.

We got to January, and it's like, oh, that's way too low. We can go higher. But we're committed to this kind of mid- to high teens volume growth. When we look at the whole market, we think this is sustainable over long periods of time. And when we talk about mid- to high teens growth, I think one thing that's important for all of you to understand, it's become clear to me over the last couple of weeks.

When we talk about mid- to high teens volume growth, we talk about that as an input to a process of developing like a whole fab strategy. And we were talking yesterday, and I asked somebody on our team a question, which I'm not going to tell you the answer to. But the question was like, what's going to be our BiCS10 mix at the end of the decade. They pull up a spreadsheet and they tell me, we have a plan already. Like, we have a fab plan like years and years and years in advance. So this investment is the input into that plan. That plan over long periods of time, grows volume mid- to high teens. Now the output on a quarter-to-quarter basis or on a year-to-year basis is going to have some variability in it because things change in the quarter you're in.

If you just pick 2 endpoints and you say, "Oh, you're growing faster, you're growing slower." It depends what endpoints you pick, right? CAGRs are very, very sensitive to endpoints. But again, one thing we're committed to this grow volume mid- to high teens. Some quarters, it's going to be less. Some quarters, it's going to be lower. Some years, it's going to be higher. Some years, it's going to be lower. But over long periods of time, this is what we're going to grow. And that's an amazing place. If you can get the economics right, you get the cyclicality dampened and then you grow, that's an unbelievable franchise, right? And you'll see it in the business model that Luis talks about. But this is important. How do we grow? How do we grow? This is really important for as an investment point of view. How do we grow?

So you go back to the BiCS roadmap, Bit Cost Scalable roadmap. We say BiCS all the time. Sometimes people don't know what it means, right? It's right there in the title: cost scalable, right? How do we build this scalable technology? So I was sitting at my desk a couple of weeks ago, and I just pulled out some material, and I said, let me do some calculations. Let me look at this. And I looked at about a 10-year period, right? About a 10-year period from calendar year '20 to calendar year '30. And I looked at kind of the plan of launching nodes across that 10-year period or -- let's call it 9 years, 5 nodes over 9 years, right? BiCS5, 6, 8, 10, 11. Again, there's a BiCS9 in there. Alper will explain that.

But these are the big major nodes. The average generation-to-generation bit growth per wafer was 54%, right? It's a pretty impressive number. Every time we put a new node in the market or every time we turn the crank on that innovation engine I talked about, we get 54% more output per wafer. It's kind of amazing. Now the issue is we don't do a node every year. So 5 nodes over 9 years, you can do the math on what that is per year. When I was a young executive, they sent me to PR training. And one of the things they told me in PR training is never make the statement, "you can do the math," right? Because nobody does the math or nobody can do the math. So now all of you can do math, I'm pretty sure. But I'm going to do the math for you.

And so when you back that out to a yearly CAGR of productivity growth, it's 27% -- so through the application of innovation, we can grow output per wafer at a rate of 27% a year, which, by the way, tells you, you can't just release nodes whenever they're available. Otherwise, you're going to flood the market with supply and then you're going to have another 2023 situation. But that's in the past. But what this tells you from an investment point of view, what it tells me from an investment point of view is growth is primarily driven in this business by the application of intellectual capital, not financial capital. It's the paying of those engineers and those NAND designers to continue to drive that roadmap forward is what is going to drive the growth.

Now, there is more CapEx. Each node is more steps, more steps is more tools, more tools is slightly more cleanroom space, right? But you will -- you can see from this equation why we're constantly like you have to reduce wafers on an ongoing basis because the technology you have is so productive. So it's like getting the business model around that really, really high-powered engine, getting that right, and it's an incredible business. And what this means is, again, from an investment perspective, the ability to take bits, turn them into revenue, and then revenue to free cash flow is quite high. And I think at scale, you have a franchise that has an extraordinary ability to generate free cash flow margins, right? And you'll see a little bit of that later on. All right. That's where we are. That's the big picture.

I'm going to turn it over now to the people that are really driving all of this fundamental technology and financial greatness. So Alper -- as Ivan said, Alper is going to come up. He's going to talk about the NAND roadmap. As I said, like, you're not going anywhere in this business if you don't have the right roadmap. And he is expert on this. We have a lot of people working on this. He'll tell you about BiCS8, BiCS10, where that's going.

We made this statement on our earnings call last week. Earnings calls are always fascinating events because something always happens you don't expect. I love earnings calls. Everybody -- like, I talk to my peers, and I'm like, "Oh, I've never heard somebody say they love an earnings call." Like, we love earnings calls. We get to talk about our business. But we made this statement that said we think the market is going to be $300 billion this year and $500 billion the year after that, right? And like, oh my gosh, like, we never heard that number before, and they started backing into thinking that was like a revenue forecast for '27. It wasn't, right? It wasn't.

The one thing you need to understand about that number: it includes China. So when you include China, it skews the numbers if you're trying to back into everything else in the world. And that's an exercise left to the reader to figure out what that is. But Eric Cherrstrom -- we have a team on market intelligence. This is why we were able to stand up here last year. And we said with conviction we thought pricing was going to inflect in the second half of the year.

And we -- and that was very -- again, another debate, very debatable. And the reason we had conviction in saying that is because the work Eric does and his team. So we thought we'd give you some visibility into how he thinks about this and kind of how he thinks the big picture of how this market has kind of resettled over the last 2, 3, 4 years. Very important to understand this. Not only is SanDisk changing, the playing field we're on is changing dramatically. And I think when you understand the dynamics of that, it starts to unlock some of this value creation as well. All right. Khurram is going to come up. He's going to try -- he's going to talk about AI inference. There's obviously a massive tailwind to the business right now.

There's a lot of questions, a lot of conversations about KV cache, how is NAND used in inference. He's going to try and demystify all that a little bit, how our products fit into the data center. How do we think KV cache is going to grow in the future. And then Luis is going to come. He's going to wrap it all up into the business model. He's going to go through the NBMs in a little more detail. And just like last time -- we're going to put HBF -- we'll talk a little bit about 3D Matrix Memory, too. We're going to put it at the end. And the reason we put it there: it's not in the model yet, right? When it's in the model, we'll tell you. But until then, what Luis is talking about is the model for all the core business, and you should see these as like future innovations that we'll continue to update you on.

But I think Alper will give you a very good view of all the progress that's happened in the whole world around HBF in the last 1.5 years. It's really been really quite exciting. All right. Thanks for your time. Thanks for being here again, and I'm going to turn it over to Alper to get into the technology. Thank you.

Please welcome Chief Technology Officer at SanDisk, Alper Ilkbahar.

Thank you. Good morning. Great to see all of you. Welcome. Welcome. I'm really excited to be with all of you here and talk about the memory technologies our teams at SanDisk are driving. So let me start off highlighting the main pillars of our technology strategy. Our #1 priority is to keep our exponential scaling engine running. In semiconductors, especially in memories, scalability is the most critical factor we are looking for. And you're going to hear us talk about scalability, the importance and the role it plays in our business over and over today. And we're going to talk about how our 19th generation strong scale engine keeps going and how our roadmap extends well into the next decade and beyond.

Next, we are laser-focused on what our customers are looking for, which is performance, power efficiency, and density. You're going to hear from us how we are leading in every single one of these metrics and how we are delivering the most capital efficiency amongst our peers in the industry to deliver superior financial results. And finally, we strive to innovate to amaze. We are innovators, we are engineers, technologists, and we are really looking for ways of improving all the applications and products every single day, but not only the existing ones, but we're also innovating in creating new applications and new markets every single day.

With that, let me start jumping into the next slide here. This is a slide that I shared with you last year in February here on the stage. And at the time, I shared with you how since 2001, our teams have delivered 17 generations of NAND technology.

So today, only 18 months later, we have added 2 new generations of NAND technology, BiCS9 and BiCS10 to our roster. This data tells me 2 important things. The first one is the pace of our innovation is accelerating to match the demands of the markets. And number two, NAND flash being the most scalable semiconductor technology is actually the only technology that can match the exponential growth of AI. And that's why you're seeing and will continue to see the increased adoption of NAND in AI architectures.

Last year, I also showed you this slide to explain the vectors that are driving and fueling our scaling engine. At the time, I had talked about how we are prioritizing the more technically difficult but significantly more capital-efficient ways of scaling, which are lateral scaling, logical scaling, and architectural scaling. These are our priorities over the easier yet financially more challenging and costly vertical scaling, which is adding layers. Our strategy hasn't changed, and we are really pleased with the results we're getting through this strategy. And it's exemplified in this slide.

I shared similar data with you last year. At the time, the data ran through 2024. This year, we added 2025 as well. And what we're showing here is capital intensity of SanDisk and our JV partner, Kioxia, and compare it against the capital intensity of the industry. And we are defining capital intensity as how much CapEx we have to put there to get an incremental petabyte of bit output. Of course, the higher it is, the worse you're off. So you're trying to minimize your capital intensity.

And the white line here is showing the average of the industry divided by our numbers, so the ratio to that. So now this data, what it's telling you is in every data point, by the way, here is backward looking for 3 years and averaging that. In 2025, the industry spent on average, 2.66x more capital than we did to generate the same output, 2.66x more. So this is the capital efficiency our strategy is delivering.

Looking at the same data through a different lens. Here, I'm showing you the percentage output -- bit output of each of the peers in the industry versus the percentage of the CapEx they spend every year. Actually, we are looking here in a period of '21 to '25. So this is a 5-year period we are looking at. So on the left, you're going to see that between us and our joint venture partners over that 5-year period, we produced 29% of the industry's bits while spending only 13% of the capital, 29% versus 13%.

So when we look at these ratios for each player one by one, I captured that data on the right side. What you're going to see is that our capital efficiency is just about 2x that of our nearest competitor. And this is possible through our technology strategy that I just highlighted through our execution of that incredible technology roadmap, our scale, our operational capabilities and excellence as well as the intense focus we have on tool reuse, all of which make this possible and delivering superior financial results for all of our investors and shareholders.

Now as we are pushing our technology forward and pushing the limits of scaling in every single generation, we actually advanced the technology across 2 dimensions. The first dimension is what I'm capturing in the x-axis here. Every generation gives us more bits. This is happening through pursuing those 4 vectors of scaling. Next -- and you have seen earlier, every generation, we get 50% to 60% more bits, and that's happening through scaling.

The other frontier we pursue is what I'm capturing in the y-axis here, which is performance and power efficiency. And we get those improvements in every generation as well through device and design innovation. Now you have seen this roadmap before. I'm mostly going to talk about BiCS8 and 10 today on that roadmap. But as you heard earlier, our teams are already working on BiCS11 and beyond. And the scaling engine continues running, and we see actually no end to the scaling limits in the foreseeable future. So we're going to run this for a decade and longer, of course.

Now we have talked earlier about our CBA technology, which is our hybrid bonding technology where we're able to combine 2 different wafers to make a single wafer. By utilizing this innovation, we are actually augmenting a derivative roadmap, as I'm showing here. This CBA-enabled roadmap allows us to take an existing technology node and push its performance and power efficiency to the next level by combining the memory array technology of that node with the next-generation CMOS technology and combine the 2 wafers together and get to the next level of performance.

So you may ask, why is that relevant? Why is that important? A great example for where this is needed is actually happening in the data centers right now. The storage interfaces in the data centers, and what I mean by storage interfaces, think about our enterprise SSDs, they run on a standard interface called PCIe. And there's a transition that happens industry-wide from, say, Gen 4 to Gen 5 to Gen 6. These transitions used to happen every 4, 5 years in the past. That time allowed us to essentially move from one technology node to the next one, ramp that technology and maybe ramp the next one as well so that you would have plenty of supply and transition our products gradually into the next generation of this interface.

But with the advent of AI, these transitions started happening significantly faster. And when the demand turns on in the data center with the volumes that we're looking at, you have to enable that transition extremely fast. So you may not even have time to ramp your next-generation technology node to meet that demand. So what do we do with this CBA-enabled derivative roadmap? We can take our existing technology node and very quickly move it to the next performance level and make that transition happen, move our entire portfolio very quickly to what our customers need and customize the silicon very rapidly.

And the beauty of it is that it can be done with very minimal additional capital spending because we're leveraging the existing nodes' memory array technology, which is where most of our capital sits. So this is a super efficient way of moving to the next level of performance. It gives us incredible operational flexibility. It gives us great capital efficiency, and it gives us the ability to meet our customers' requirements very quickly.

So it's an awesome innovation, a technology competitive advantage that we can leverage and create incredible competitive advantage for our business. So I'm going to talk about BiCS9 in a quick bit. But before I get to it, I want to take a quick look at BiCS8 because BiCS8 today is the backbone of our current production. And it is the industry's gold standard. We introduced the CBA technology -- the hybrid bonding technology -- first time with BiCS8, and it has given us tremendous competitive advantage in performance, density, as well as power efficiency. And it turns out that these are exactly what our AI data center customers were looking for. So when we compare BiCS8 against some of our peers' performance and power efficiency numbers, we've seen a tremendous gap where we had a huge advantage.

And here, I'm comparing our BiCS8 against our peers' 2xx-generation memories. Of course, our peers are moving forward. They're announcing their next-generation products, which we call 2yy or 3xx products. And they are improving their performance. But when we look at the power efficiency, we're seeing that their power efficiency is really not moving a lot better. Now, let me put on what's coming, BiCS10, and show you how BiCS10 is going to compare against these. So this is where BiCS10 is coming out to be.

BiCS8 was amazing. BiCS10 is going to be even better. And I think it's going to be the gold standard for the AI data centers very soon. So we'll get to BiCS10 in a bit. But let's first talk about the latest news we talked about BiCS9. So BiCS9 is the first product where we're essentially deploying this hybrid technology, hybrid bonding technology extension. What we have done for BiCS9 is we have taken the BiCS8 cell array -- the mature cell array we have -- and combined it with the next-generation CMOS wafer. And through that, we achieved tremendous performance gains.

And we've done so with minimal incremental capital spending. So we essentially are upgrading our BiCS8 deployment supply bases to the next-generation performance level with minimal capital. And we designed BiCS9 based on the specification from our large hyperscale customers -- they wanted to have incredible performance. They wanted to have a lot of it, and they wanted to have it yesterday. So this allowed us to achieve all those objectives very quickly.

And I'm very happy to report that this product is already as part of our NBMs, and our customers can't wait to have this. So it's going to power our next-generation storage SSDs that Khurram is going to talk about. So really looking forward to seeing this powering your AI very soon. Okay. Let's go to BiCS10 real quickly. I gave you a sneak preview of this BiCS10, the first product in the BiCS10 lineup last year.

This is the 1-terabit TLC die, again, achieving tremendous improvements over BiCS8, which is the best in the industry. Since then, our teams have done a marvelous job with this technology, and it progressed ahead of our expectations. So it allowed us to start sampling this die to our customers as of this month, which we just announced.

The second product in the BiCS10 lineup is the 2-terabit QLC die, QLC being 4-bit-per-cell technology. We're really proud of this technology. It actually is the highest density memory chip in the world. And while delivering this, we have achieved more than 60% density improvements over BiCS8, more than doubled the read and write bandwidth, as well as improved the power intensity and efficiency by 75%. These are truly amazing numbers considering that we're beating the world's best NAND, BiCS8. To give you a better picture of the power of scaling, I wanted to show you BiCS8 and BiCS10 2-terabit dies side by side. So this is what we are able to do with the scaling engine. Each BiCS10 2-terabit wafer has 65% more bits than the BiCS8 2-terabit.

So you have the pictures here, but the wafers are sitting outside. So I will invite all of you to please go out and check in person and experience that scaling. The wafers are sitting out there, and then you can even touch and play with it if you want. So putting this in historical perspective, going back to the data that David showed you earlier, and we're looking at the generations from BiCS5 through BiCS10 and even projecting into what's coming next, BiCS11, we're delivering a 27% CAGR on bit growth per wafer annually -- 27%.

And you already heard that our production plans are based on our long-term demand forecast of about high teens. So this is delivering roughly 50% over that, which means that we have the technology productivity to meet all of our production needs and plans just by scaling the technology alone. As a matter of fact, because of this productivity, our wafer starts have come down over time. And that's another way how we have created value and driven our capital efficiency.

So with that, I thank you all for being here. I will be back to talk to you about HBF at the end of this presentation, and I will invite Eric to be with you to share his intelligence insights in the market. Thank you very much, and see you soon.

Please welcome Senior Director, Market Intelligence at SanDisk, Eric Cherrstrom.

Good morning. I'm excited to be here to give you a brief NAND market update. We expect the flash market to reach 1.2 zettabytes of shipments in 2026. So how do we get there? Well, as you know, this industry started with growth in the consumer and edge. The key drivers were phone scaling to 1.5 billion units of annualized shipments, PCs shifting their storage needs from hard drive to flash, and the emergence of the enterprise SSD. Then, in 2022, ChatGPT was launched. AI propelled the data center segment and increased its share of TAM over time. Data center share of TAM in the early '20s was 20% of bits. Last year, it was 30%. This year, 50% and continues to outpace the market.

So now, when you think about the revenue overlaying the volumes from the prior slide, you can again see 2 distinct periods. Period 1, flash was priced as a commodity. ASPs reduced, offset volume gains. The historical average during this period of time for the industry was $60 billion, and we measured cycles in $20 billion increments. Now, flash is a critical component of a multi-period data center build-out. And this leads us to believe that the flash market is going to grow to over $300 billion in calendar '26 and again grow to nearly $500 billion in 2027.

So what was going on with supply during the same period of time? Well, in the late teens, the industry was targeting over a 30% annualized bit growth rate. The industry had to invest, growing wafer starts all the way up to 1,800,000 wafers per month in 2022. At that peak of wafer capacity, COVID-related inventory digestion dramatically reduced flash industry demand. Supply side had to react, underutilizing 500,000 wafers per month and structurally resetting their capacity to 30% below peak levels.

In spite of that reduction in wafer capacity, the industry was still able to achieve a mid- to high teens production growth rate via nodal migrations. So let's talk cloud. The chart on the left shows U.S. data center CapEx for select hyperscale and Neocloud customers. The chart is showing the projections over time. And as you can see, starting in 2023, we have seen 15 consecutive quarters of upward revisions to this selected CapEx.

Current estimates show that $1.9 trillion of capital will be spent by these companies in 2026 and 2027. And we believe monetization is coming. The most recent Amazon earnings call, they talked about not having enough capacity to support near-term demand and the fact that AWS could reach nearly $1 trillion of annualized revenue. Shifting to the edge. The edge is going through a transition period. The chart on the right shows our expectation of unit decline year-over-year in the mid-teens for both the PC and mobile segment.

And as you can see, all of this reduction is being driven by the low-end devices. Our belief is that OEMs will shift their mix to more premium offerings and ASPs will continue to grow -- ASPs and revenues will continue to grow in '26 and '27. And as you can see from the most recent earnings of major OEMs, year-over-year revenues grew between 13% and 24%.

So flash is entering a new reality, a reality where the data center is the majority proportion of the share, edge continues to mix to premium devices, and on the supply side, bit growth targets are met via nodal migrations. All of this put together gives us the conviction where we see the market opportunity growing over $300 billion in 2026 and can approach $0.5 trillion in 2027.

With that, let me pass it over to Khurram to talk about the era of inference.

[Presentation]

Please welcome Chief Product Officer at SanDisk, Khurram Ismail.

Good morning. I'm Khurram, and I'm here to talk about the infrastructure outlook, specifically as it pertains to flash. The good news is that we are almost at halftime. And since we don't have any breaks for halftime, I get to be your host for the halftime. So let's get into it. So David talked about engineers being at SanDisk for a long time. I'm one of those. I've been in this industry for 27 years, all in memory, and never has been a time more exciting for memory than it is now in AI. I've seen all sorts of peaks and troughs. The pace of innovation that AI is bringing is tremendous, and you all can see that. We see the new frontier models being loaded, right? The system architects are changing the design every 6 months. So the pace of innovation is quite rapid. But the infrastructure required to deploy that innovation is also being deployed at a very unprecedented rate, right?

So what is the role of flash? In my talk for the next 20-odd minutes, I want to leave you with 2 things. First, like David mentioned, our conviction on the critical role that flash plays and the size of the opportunity. The second one, I hope you gain the appreciation of flash is not really a clumpy device sitting at the edge of the infrastructure, but it is actually being proliferated through all the layers of AI infrastructure.

So I want to start with some fun facts as I'll be going over some concepts, and I think it will help us understand those concepts if I draw some analogy to the human brain. And maybe some of you know, I was just doing ChatGPT, Gemini, and I found an interesting fact that each human brain is wired with 2.5 petabytes of memory. Now, you multiply that by entire human intelligence. That's like 20 yottabytes of collective human memory. Now, 1 yottabyte -- I had to look that one up too. I work in zettabytes and exabytes, right? So 1 yottabyte is 1,000 zettabytes.

Now, you look at the right at the cloud infrastructure, which I would characterize as being very early in the innings, it is only hundreds of exabytes to maybe 10 zettabytes, all memory combined, right? One could argue looking at this, as we are going to scale intelligence, the cloud infrastructure can use more storage, right? The interesting thing about human brain is it works with 2 types of memory: the short-term memory and long-term memory. And they both work hand-in-hand utilizing each other to generate intelligence. Turns out the AI intelligence is built on very, very similar concept. You have a transient, short-term working memory that is called ephemeral KV cache, and we'll cover that. And then you have the long-term memory, which is a little bit of more persistence, which is known as persistent KV cache. So there is a similarity. They both work on the same principle of how humans store data and process data versus AI intelligence. Okay.

So Eric talked about the total demand in 2026 to be 1.2 zettabytes for entire flash market, right? Here, I'm only focused on 2030 AI data center TAM, which is equivalent to what we ship as a total output as an industry in 2026. So the opportunity is massive. And I'll come back to this slide again as I go through why that is the case. But it's important to note, as I mentioned, flash is not just a single device sitting at the edge of infrastructure. There are many workloads that are emerging on flash, specifically in AI data center.

What are some of those workloads? Well, you have first: fast data lakes. We talked about it last year. These are the massive data lakes that require massive storage. Second is staging. These are a little bit direct-attached device close to the GPU for training, checkpointing. So that market is having tremendous growth. And last is the KV cache, and that will be the focus of our conversation because that's the fastest-growing segment in AI data center. And when you look at the composition of NAND technology, we see that TLC is dominant technology in 2030, and QLC still has a very good decent-sized share.

So how is the infrastructure being viewed today? And the thinking around infrastructure is changing. We are moving from what used to be total cost of ownership to the total value of ownership. In the past, when you deployed the infrastructure, you sort of prioritized cost running at very large scale. And some of those considerations are listed here. With the total value of ownership, the equation is changing. Infrastructure is no longer being viewed as a cost center, but really a driver of value generation, right? And in this case, the value and the output is intelligence, right? So the race to scale the intelligence is heating up. As you can see, everybody is trying to generate more tokens.

They're trying to generate -- get more users on their systems or on their AI -- but with this scale, there's a lot of challenges that come, right? There's always the challenges of power, right? Where do you store these tokens in volatile media only? There's not enough volatile media. What role does non-volatile media plays, right? So -- and then there is shifting architectures, right? We're moving from training to inference. And turns out flash solves a lot of these problems, and that's where I'll be taking you next. So you saw the video, the era of inference.

The only point I would make here is a lot of focus in the AI data as it flows through the infrastructure is on inference. So I presented this 18 months ago last February in 2025: the 5-stage data cycle. The first 3 we focused a lot last year, which are associated with training the model -- how you store the data, how you prepare the data, and how you present the data to the GPU for training was the focus. And flash did quite well. We had our high-cap QLCs that were used in fast data lakes. We had direct-attached TLC SSDs that provide the active dataset for the model to train. So a lot of infrastructure got built as a result of this.

But now, as the focus is shifting to inference, the question to ask: the infrastructure that was built for training if the same infrastructure relevant? Can that satisfy the need -- growing need of inference? And the answer is no. You can see as we move along, there is disaggregation happening in infrastructure, right? The infrastructure for training is quite different from the infrastructure of inference. So let's look at what's happening inside the inference. So like I said, we have our existing products that go into fast data lake staging checkpointing, but inference something new, and that's where we will hone in on KV cache.

There are 2 interesting trends that are emerging in inference when it comes to NAND flash. The first is data augmentation. And the second is context remembrance, right? A lot of you, I'm sure, use the models. And if you know, if you're using the models, it's becoming more persistent. It remembers who you are, right? So that's the second popular use case. So RAG is one of the most popular techniques that is used to provide external data, so the model can provide you much more relevant and accurate responses.

Second, the users are wanting richer conversations, smarter conversations, longer conversations. And what ends up happening as a result is a KV cache amplification. And the KV cache amplification -- the way to think about is you're having longer context lengths because you want longer conversations. You have longer reasoning chains because you want iterative process. You want the model to know about you. And then there are multimodalities associated with this. All of these are driving the amplification in KV cache.

The easiest way to think about KV cache is if I'm having a conversation with you and you're taking notes as we are having now, so you can refer to the notes rather than listening to my conversation again. And that notebook serves as KV cache. So let's just briefly touch on what is the KV cache, because that's the most important part in inference. Inference has 2 major parts. The first is prefill, where the model is thinking. The way to think about the prefill stage in inference is where the model is thinking when you provide the input. The second is decode, where the model is responding -- is giving you a response. So as a user puts the input prompt, that gets tokenized and all that input gets processed simultaneously.

So one would think, as that process is driving a lot of parallelism, that is compute-bound. You'll hear a lot more people say that prefill is really compute-bound. And then when the prefill happens, a context gets generated. And now, in the implementation, that gets stored in a memory, which is called KV cache. The way to think about KV cache is it's a working memory of inference, or the notebook -- a memory notebook.

It is not important for now for us to discuss that where does that gets stored. I'll take you through how the KV cache hierarchy works, but it's important to know that this context is growing, right? It continues to grow. The most interesting part of inference is the decode process. This is where you generate the response. And as you may know, decode is autoregressive procedure where one token gets generated at a time to generate the response.

Now, to generate the response, the token that gets generated has to know the context of all the previous tokens that were generated. You can imagine if you didn't have the KV cache, that would present a tremendous challenge to the infrastructure, to the power, and computational overhead, which didn't need to happen, right? So that's where you see -- when we see all the memory boats are rising, it is because of this KV cache, because it does make the entire AI process much, much more efficient.

Okay. So we talked about this context is growing. It gets stored in the KV cache. Well, how does a KV cache look from a hierarchy perspective? We presented this last Investor Day, and we talked about the system memory hierarchy in a data center system. And the way to think about this memory hierarchy is around the vectors of performance, power, and capacity, right? And for those of us who are in love with flash, we always made the assertion that flash is the most scalable technology around these vectors.

What it turns out, we were right; inference is a perfect use case, right? Inference is a perfect use case for flash. Why? As we talked about the KV cache amplification, and with the deployment of AI agentic workflows, it's generating a lot of tokens. There's a need to have more pages in your memory notebook, so all these states need to be preserved. The contexts are getting longer. The conversations are getting longer. The reasoning chains are getting longer.

And last, all the data that gets generated as part of your input to the system. So we have very close partnerships with our customers. The NBMs are a testament. We get to learn a lot from our customers. Now, they're all hyper-focused on optimizing this KV cache because it really solves a lot of problems for them, right? But the way they go about it is different. But there is one common theme that emerges from a KV cache memory hierarchy that is generally applicable to all the AI systems that are getting deployed right now.

At the top -- taking you back to the human brain analogy -- the ephemeral cache, right? There will be a quiz after this. Ephemeral KV cache, right? Those are your HBM and system DRAM. The way to think about this: all the hot context -- the current context that the user needs, right, to get the response -- those are stored here. But again, if you look at from top to down, capacity is at play; HBM and DRAM are generally smaller, right?

Next is the long-term memory, the persistent KV cache. So anything that cannot be stored in the high-tier bandwidth of DRAM and HBM gets stored in flash. And this is what we are calling persistent KV cache. Now, there are multi-layers. Like, remember when I said, the flash is proliferated throughout multiple layers of AI data centers. I just want you to remember there are multiple tiers where flash is deployed, right, because we'll use this later on in the presentation.

But it's important to understand that you cannot scale the intelligence without having this persistent KV cache layer. As you can imagine, in inference, the model gets trained once. As the context grows, the interactions are in billions, right? So you need some kind of persistence in your memory hierarchy. And it provides a nice extendable capacity to the AI systems without having the need to take everything through the volatile ephemeral KV cache.

So now that we have covered the KV cache, and we talked about KV cache is going to be 35% of the market in 2030, how do we plan? How do we size the opportunity? Like, how big can this persistent KV cache can be? If I'm an infrastructure planner, right, I have to think about few things. And this is SanDisk equation of how to think about the size opportunity for the KV cache.

But if I am the planner -- and this, again, this we derive from talking to a lot of our customers because we have a lot of close relationship, people who are actually deploying this at scale. So if I'm planning for this, the first thing that I have to think about is the number of sessions that are going to hit my infrastructure. That's number one. But more importantly, as the sessions hit, how many sessions do I want to retain and for how long, right?

We have customers who tell us, "Well, Khurram, we retain the sessions or the context for only a couple of hours so my friend, Luis, can go have coffee and come back and have this context." Or there are customers who are keeping all the context forever. They want to monetize this somehow, but that's how they are looking at it. But you can think about it, if you retain it forever, there's tremendous opportunity for this KV cache to grow, right? So that's how they are thinking about it: how many sessions are going to hit my infrastructure and how long do I keep them.

The second part is if you are planning a build-out, you obviously have an existing infrastructure that has a set of KV cache pools, right? So you want to only plan for the KV cache that your current sessions may miss, and that's represented in cache miss ratio, right? So that's another important factor. And lastly, we have to figure out how much storage will be required in a session. And that's a function of 2 things: the first one being what is going to be your session length or size. A lot of people talk about context length, but it is a series of tokens, right, that determine.

And each token, by the way, as we talked about the decode provide, generates tremendous overhead on storage needs. So each token is represented in tens of kilobytes to hundreds of kilobytes, depending on which model, which implementation you're using. So as you can see, these variables are what people use to determine how big of a KV cache or persistent KV cache, rather, they need to deploy. And here is our answer. And this is, again, a 1-zettabyte installed base. When you multiply all these things up -- and there are, by the way, these are just 4 variables underneath, there are second-order calculations that come in, right, to make sure that we arrive at the right number, but the SanDisk estimated 2030 installed base is 1 zettabyte.

This is, again, this is the fastest-growing workload. And this we are saying between now until 2030, we'll have 1 zettabyte of installed base. This again, our customers are very dynamic. They're changing things. Architectures are changing. There's a lot of optimizations that are happening around flash. But this gives you a good proxy to think about that, hey, if I want to take this case up and if I take the retention time up, the number will be quite large.

So we feel pretty good about this because, in general, we see the context length growing, right? The average session is growing, the number of users that use AI is growing. So we see it in a positive place. So coming back to the 1.2-zettabyte number, and I would say this again: that the KV cache number that is represented here, because I didn't cover it first time, is 35% of the overall market. So right, the previous chart showed the 1-zettabyte number as an installed base. In 2030, specifically, we see the size of KV cache being 35% of the market, so you can do the math.

And this again, really demonstrates that flash is present in multiple workloads of the AI data center. So let's look at a little bit more physically and how flash sits in data center, right? This is just showing the placement -- various placement of flash in the data center. At the foundation of it, in the gray box on the right, is a sea of large-capacity drives. These are your fast data lakes that contain your training data, that contain your embeddings, your vectors, your RAG repositories -- all the things that enterprise need to store to make the AI work.

For this, from flash point of view, high-capacity QLC drives are perfect because they provide that enormous capacity. Now you may ask, "Okay, you say KV cache. You remember the hierarchy that I showed you and there were multi-layers, right?" Just think of it as a very cold, cold KV cache, the way to think about this. It's great for QLC, and we see that deployed in data lakes.

Now, as you get closer to the GPU, the requirements change. This is the second one. These are our direct-attached SSDs. Now here, you have actually active training datasets, right? You have checkpointing going on, and you have a lot of data orchestration that is happening from GPU. And for this, you need a very high-performance TLC SSD. Now again, going back to that G3 Tier-3 of KV cache, this is what I would characterize as a hot and warmer KV cache. And lastly, what we talked about in persistent KV cache: now there is another rack-scale, network-scale data movement that happens between the GPU complex and something that is closer to it via the network, right?

This is what we call the persistent KV cache. And I would characterize this as lukewarm G3.5 that you saw in that pyramid. So this is showing the AI data center eSSD placement. And a lot of people talk about AI data center and they talk about, well, it's a GPU factory. But the way to think about it -- I hope with these placements that you can see that flash is living simultaneously in many different places.

So I would assert that every AI factory is ultimately a data factory. And it's true when it comes to inference. So I'd be remiss if we didn't talk about our products. Last year, when I was here, we were trying to tell you that we are going to succeed in the data center, as David alluded to. Happy to report that both our TLC eSSD and QLC SSD are qualified at major hyperscalers, major customers, OEMs -- so and we are shipping both of them.

For the TLC eSSD, we're shipping it in PCIe Gen 5 configuration in all form factors, and they are great for staging and KV cache that we discussed on the previous slide. When it comes to QLC -- on the TLC SSD, we also demonstrated our next PCIe Gen 6 drive at Flash Memory Summit. It's going to be an industry-leading, high-performance, great power drive, really going to solve a lot of inference bottlenecks. So we demonstrated that at FMS last week. Similarly, on high-capacity eSSD, if you remember, last year we showed you a roadmap of up to 1 petabyte. We, at the Flash Memory Summit, demonstrated our 256-terabyte drive in E3 form factor, and that was very well received. And you will see a lot of the market shifting towards higher cap drive next year, from 128 to 256.

So we have a great portfolio. We have great platforms that will serve all the needs of the AI data center growth now, but also for future. So we talked about a lot of close collaboration with customers, applying our thinking, understanding how the KV cache looks like, what does the market size opportunity looks like. But we want to become AI practitioners ourselves.

We also want to be the AI practitioners. So we started an initiative at SanDisk. It's called AI Lab at SanDisk, where, you can imagine, we have server-scale, rack-scale type of systems, and we run the workloads, the models the way our customers do because we want to understand -- truly understand -- the bottlenecks, and we want to complement it with what we learn from our customers with our own understanding.

So here, I'm just providing you 2 metrics. The way to think about this is this data was collected in a server-scale application or system with a cluster of GPUs, multiple SSDs, HBM, DRAM -- all the hierarchy that I showed you. We ran hundreds of user sessions, like I explained to you in the persistent KV cache equation. We assumed certain things in that equation. And what we see that a system that has SSD versus a system that only has volatile media like HBM and DDR consume 75% less energy.

And I'll extend it -- this is not published data, but for us to generate 1 million token on a system with SSD versus just volatile media or no SSD, it takes one-fifth of less power. So different metrics, but you get the idea that for SSD, you consume much less power. Secondly, on the same system that had an SSD versus no SSD, we saw 75% higher throughput in tokens per second. We generated more tokens per second than you would with HBM and DDR simply why? Because you don't have enough capacity; you are limited, right, and you're having to go recompute. That's an expensive process, both in power and performance.

So as you can see, that SSD is not something that is just an afterthought. It's actually an essential, and that's where you see the explosion of KV cache workload in the market. So I want to leave you with one last thought. I don't know what I was supposed to say. Let me pull it from my persistent KV cache, right? So okay, the persistent KV cache responded. So today, flash represents the work that is completed, right? And if you look at previous compute cycles, whenever the compute cycle was finished, all intermediate states or the nodes were discarded only to be recomputed whenever the compute needed.

If you think about the millions and billions of scale of AI, that strategy is very inefficient. It won't work, right? So from that perspective, we like to think about flash SSD as a token battery, right? It's storing the energy to be used later on, right? And flash truly represents the accumulated intelligence. We believe you cannot build intelligence without persistence, and flash is great. And in an era where the most valuable output is intelligence, preserving it becomes as important as creating it. So in summary, we have a robust growth outlook. We have conviction in the AI data center market. We have a good understanding of where the customers are headed, where things are headed, how the architectures are working. We have the right product portfolio. We have a strong product portfolio that is good for now and for future.

We also acknowledge that there are going to be efficiencies when it comes to inference. You all heard quantization -- all the optimization that is happening to reduce the store in KV cache -- but that's only going to fuel the Jevons Paradox. There's going to be more use cases that will come out of it. So we feel pretty bullish that these efficiencies are welcome, and they're going to drive more utilization.

And last, flash is not something that is an afterthought. Industry is actually innovating around flash. Why? Because it's the most scalable technology. And like I said, you cannot build intelligence without persistence, and flash technology is a great medium to build intelligence. Thank you very much.

And I would like to now invite my friend, Luis, to talk about financials and business planning.

Please welcome Chief Financial Officer at SanDisk, Luis Visoso.

Good morning, everyone. I thought you may want to look at some numbers. So it's great to be back here after 18 months of launching the company. And frankly, this conversation is about sustainable value creation, sustainable value creation. We're committed to doing that every single year. Our journey, as I said, started in February '25 when we separated from Western Digital. Shortly thereafter, we announced our Q3 '25 results. As you may remember those numbers: $1.7 billion in revenue, we lost $0.30 in non-GAAP EPS, and we generated $220 million in adjusted free cash flow. We've come a long way.

Hopefully, you saw our earnings last week. We reported $9 billion in revenue, non-GAAP EPS of $39.25, and an adjusted free cash flow of $5 billion. And this excludes cash we received from our NBMs as prepayments and deposits. So we've come a long way. Now going forward, what are we going to do? We're committed to creating value for our customers. And as we do that, we're confident that we can create value for our shareholders. So let's look back into the year that we just delivered. These are the metrics that matter the most. We operate in a large, fast-growing market. That market has tripled or will triple in calendar year '26, reaching $300 billion on its way to $500 billion in calendar year 2027. So a very large market.

Now, what's very important is the composition of the market is changing from an edge-centric market to a data center-centric market. That brings very different dynamics, and I'll explain some of that. Our revenue for the year was $20 billion, up 175%. That's twice as high as our prior record that we delivered in 2022. So nice growth. And importantly, that growth -- that revenue improved every single quarter throughout the year.

Gross margin 71.6% -- that's up from 30.3% the year before. Again, our performance improved every single quarter throughout the year. We closed the year with 84.6% gross margin. That enabled our EPS to go to $39.25, up from only $0.29 the year before. So great performance on our financials.

And the metric that matters the most is our free cash flow. We generated $8.7 billion in free cash flow, excluding those new business model prepayments. And that also improved every single quarter. Our run rate $20 billion. That's our run rate of generating free cash flow from this business. So a very good business. Growth is there. The market is growing. We're capturing that value.

You may have a few comments -- a few questions about the new business model. So let's go into that. Very importantly, this is our way of strengthening our relationship with our most strategic customers. Why? Because the new business models deliver a fast-growing, profitable, and less volatile business. Going back to what David just said: fast-growing, very attractive, less volatile business -- isn't that beautiful. So we're building these relationships.

The way this started is very custom agreements with each of our customers that center around supply and demand certainty. The conversations didn't start around pricing. Obviously, we do get to pricing, but they start with supply and demand. Our customers want to make sure they have -- they can get the products they need, and we want to make sure we have somebody on the other side. So that's how the conversation started. There are details by quarter, details by month in most cases. And while they are custom-made, there is a framework that's consistent around all these agreements. They start with a multi-year in most cases. When you have a multi-year, those volumes are growing at a very fast pace, faster than we're growing as a company, and there are fixed and variable components on pricing.

And very importantly, every single one of these agreements has a financial guarantee, and I'll talk about that. These conversations go to the highest level of the companies. They require Board approval. We're talking to CFOs. We're talking to treasurers. We're talking to CEOs. This is not like a typical conversation of the past. So let's talk about some of the details. So we have 8 engagements with customers. These are win-win conversations. As Khurram alluded, there is high level, very deep integration from a technical and commercial side.

These 8 customers, by the way -- news to you -- include 3 hyperscalers from the U.S. 3 U.S. hyperscalers are part of these 8 customers. Now, the oldest deal we signed was only in January of this year. And guess what: 2 customers already came back and they said, "Hey, guess what, as I look at my models, they do the math that Khurram was talking about, I need more." So they are already expanding -- either extending the duration of their term or adding more bits to the same contract length.

So we feel very good about these contracts. In terms of duration -- so we're moving from a quarterly price negotiation to large multi-year engagements. Remember, these price negotiations lasted 3 months, sometimes not even 3 months. And over that time, when we were operating in that model, we practically generated no shareholder value and made capital investments super difficult because they were very risky. We did not know for how long our customers were going to take our products. We had no commitments.

Go from there into an average length of our contracts of over 4 years. The longest contract is 5 years now. And we're actively in conversations with several customers to go even further. That is very important. So we're allocating a significant portion of our business to these new business models. Why? Because as I said, they are fast-growing, attractive, and less volatile businesses. We like this business model. We want this to be the predominant way of doing business for our company.

How does pricing work? Well, pricing will be fixed in some of these contracts. Some of them include variable portions. And very importantly, the variable portions include floors and ceilings, and our financials are very attractive even at floor pricing. We talked about around 80% gross margin for the floor pricing. So we believe there is upside to that pricing, and therefore, we feel very good about the financials of the new business models.

Now, the non-business model -- the rest of these bits -- will continue to fluctuate with the market. So if the market continues to go up, obviously, we have an ability to capture that. So let's try to quantify the size of these contracts. So if you look at the $93.9 billion, that's the Total Contract Value, TCV. That's how much we expect to collect in revenue from the beginning to the end of these contracts -- $93.9 billion at an average of 4 years.

Now, some of that revenue has already been recognized. So the remaining performance obligation, the RPO, is $91.1 billion. So a lot of the value, a lot of the revenue is still to come. Both of these numbers reflect the floor pricing, the minimum pricing we expect from these contracts. We believe that there is upside on both of them as prices will be higher than the floors that we have.

So we talked a lot about financial guarantees and ease the risk in these contracts, where we have secured $16.5 billion in financial guarantees from these contracts. There are 2 big buckets of this. The biggest one is financial guarantees held by or provided by third-party financial institutions. The other part, the smaller bucket, exactly $2.9 billion, is deposits and credits from our customers that we have either received or are about to receive. Of the $2.9 billion you will see on this slide, we already have $2.5 billion in our bank.

So the vast majority is financial guarantees provided by or held by third-party financial institutions. Very importantly, our customers will pay for their products in the ordinary course. So this financial guarantee, other than prepayments, will stay constant throughout most of the time. So that is important, and I'll come to that in the next slide.

How do I think about this financial guarantee? How strong of a protection is it? Well, an easy way to think about it is the ratio between your financial guarantee and your remaining performance obligation. You have the numbers. You can do the math, as David said. So if you do that and if you define that ratio at the beginning of the contract -- let's call that the base ratio as you divide the financial guarantee by the TCV, the total contract value -- some of you are questioning, well, how does that ratio evolve over time?

So we looked at our contracts, had multi-year contracts, and we wanted to provide you an illustrative example of how that ratio would evolve over time. So for a 3-year contract 2 years out, in average, you should expect that ratio from beginning to 2 years later to be twice as high. So you guarantee your protection as a percentage of the revenue to come significantly increases as the contract goes on.

So what are we going to do? Well, we're going to execute these NBMs with excellence. We don't want 4-year deals. We don't want 5-year deals. We want these NBMs to last decades, right? And therefore, we're going to execute them with excellence. We're going to have the products with quality on time, just as we agreed with our customers. We want them to fulfill their part of the bargain, and we're going to do the same. And you've seen us do some of that. We're increasing some of our safety stocks. We want to make sure that we have the agreements with our JV partners. We buy our agreements with Nanya to make sure that we have the DRAM. But we want to make sure that we can perform very well on these NBMs.

Again, the goal is to make them even longer. And number two, we're going to be very patient. We're going to be patient as we continue to evaluate new deals. And just as you saw, we only have 8 companies with whom we will sign an NBM. We'll be very selective going forward to make sure we choose the winners that value our products, are willing to pay for them, and want to make commitments which are longer term.

So how do we think about the model going forward, right? So going forward, our financials will be the result of a combination of both models. So we will have a proportion of our business will be the NBMs. That would be the largest portion of our business going forward. Why? Sorry to repeat myself, this is a growing, profitable, and less volatile business. We like this business, and it has reliable volumes. So we're going to keep that NBM, and we know exactly what to expect from that side of the business. And we'll have a portion of the business which will be the non-NBM.

We'll continue to support our customers. David alluded to that. Some customers are just too small to have new business models. Some of them are very strategic -- don't get me wrong -- but this business model may just not be the right solution for them. So when you aggregate all of that, for 2028 through 2030, we expect to grow revenue mid- to high teens, consistent with bit growth. We talked about bit growth in the mid- to high teens, where we expect revenue to grow at about that same rate.

We expect non-GAAP gross margin to be around 80%. We expect non-GAAP operating margin to be 75%. How do we get there? We expect to spend about 5% in OpEx, and we do not expect significant contributions from other income and expense. So you get to that 75%. And then you get to 50% adjusted free cash flow after paying for taxes, working capital, capital spending. And we have for modeling purposes, I would assume mid-single-digit capital intensity as a percent of revenue. That's our gross CapEx.

Now importantly, for '27, we already talked about this as part of our earnings last week. We expect a bit growth to be somewhere in the mid-teens, and we expect sequential prices to be modest throughout the year. So that's the model. Why do we feel confident sharing with you these numbers? Well, our confidence comes from our customer conversations. It comes from our new business models that we signed; based on our conversations, will lead us to believe that more NBMs will come. So we feel very good about our new business model, our conversations, and frankly, the growth of the business overall.

So what are we going to do with the cash, right? So we will continue to invest in the business. This is a great business to have, and it requires cash, and we'll continue to invest in it. What does that mean? Well, we'll continue to invest in OpEx. We'll properly fund the business. We'll invest in CapEx, and we'll continue to do things to strengthen us like the Nanya type of investments, the JV extensions -- those types of things that make us more robust, more sustainable as a company. That's super important for us.

Number two, which we've done very well this year, we'll maintain a strong balance sheet. What does that mean? A healthy cash balance? How much? Well, you've seen us operate in that range over the last few quarters, and we will keep to operating around that range. Now, that doesn't mean it will be exactly the same number. There are payment terms. There are different things that happen, but within the range that you've seen us operate over the last few quarters. We have no debt. We got rid of the TLB. We intend to keep it that way. Our revolver is unused, and we don't intend to use it either. And we'll continue to improve our credit ratings with the agencies over time.

We've made progress this year. We're at BB+ overall, and we intend to continue to make progress. And the rest of the cash -- the excess cash -- is going to go back to you guys. That's why we're here for. Our value as a company is to create value for our shareholders. And the excess cash -- not a portion, 100% -- will go back to you. We've done a lot of work to understand what's the best way to do it. And at the current moment, we believe that the best way to do it is through our share buyback program.

Now, what are we doing? If you look at last quarter, right, Q4 of '26, we generated $5 billion. How much did we return to you? $4.5 billion, right? So we're living to whatever we're telling you is exactly where we're executing. So the Board authorized a $6 billion program. We executed $4.5 billion. We have $1.5 billion left. So the Board authorized another $14 billion program. So now we have $15.5 billion authorized and not spent yet. So we will give you an update as we go on, but we believe that our role is to return the cash to our shareholders.

So in closing, we're super excited. We're super excited not of the value we have already created -- that's good, don't take me wrong -- but we're very excited about the value we can create going forward. We operate in a very attractive market. It's growing, it's profitable. And frankly, SanDisk is very well positioned to capture that value. You saw our technology. We have leading technology with NAND, leading technology with our products, and we have very close relationships with our customers. Those relationships, those NBMs, are opening doors that had never been opened as wide as they are today. So we feel very good about where we are in the market.

What's our financial model? Super simple: translate bits to revenue, revenue to profit, and profit to cash -- and then the cash flows back to you guys. So that's our model. I hope that you guys find it interesting; we do. So what we're going to do next is we're going to talk about HBF. As David mentioned, HBF is not in the revenue projections. We are funding it. It's part of our OpEx. It's part of our CapEx, but we're not including it in the revenue projections here. So thank you for that.

With that, I'll turn it over to Alper.

Please welcome back to the stage, Chief Technology Officer at SanDisk, Alper Ilkbahar.

Hello, again. Nobody left. That's amazing. Okay. So in the second part of our technology presentation, I get to show you our "innovate to amaze" DNA. And I will talk about 2 technologies that we introduced last year on this very stage. Both of these technologies address the memory wall problem. Memory wall problem is essentially simply DRAM not keeping up with the compute and AI because it just doesn't scale anymore as well as it used to. And to solve that problem, we started working on 2 technologies, both of which are highly scalable and can solve this memory wall problem. So the first technology I'm going to cover is 3D Matrix Memory. So let's just dive into it right away. Okay, I'm back. Quick recap first.

The 3D Matrix Memory -- we started working on this technology back in 2017 in our research fab. And in 2024, we moved the development to a 300 millimeter facility, a modern facility at our development partners, IMEC. And last year, when I was here, we had just delivered a development vehicle that we could just essentially pursued activities at IMEC with, and that's what we had shown you. Since then, we continued making steady progress. We used the development vehicle I showed you, and we started depositing memory layers on top of it. And we delivered 300 millimeter wafers and packaged parts to test and demonstrated multi-gigabit level functional memory arrays.

And our devices are approaching performance levels that are getting pretty close to our product specs that we had. So steady progress. It keeps going, but this definitely is a project that has a longer time horizon, and we'll keep updating you as we make more progress on this.

Okay. With that, let's go on to HBF, High-Bandwidth Flash. Last year, again here, we introduced High-Bandwidth Flash for the very first time. High-Bandwidth Flash delivers the same read bandwidth as HBM, yet with 8x to 16x the capacity. We invented this device with AI inference workloads in mind that actually leverage Mixture-of-Experts type sparse models with long context lengths and large KV caches. That's what we had in mind. And today, when I look at some of the most recent developments in the world of AI and the trends, actually, these do justify the vision we had for HBF 2 years ago.

So on this table here, I have summarized some of the latest frontier models and their characteristics. You're going to see very quickly that certain trends are emerging. First, the parameter size -- the models are growing. Trillion-plus, 2-trillion parameter models are no longer amazing; they're just commonplace. And many of these models actually started utilizing Mixture-of-Experts, sparse models, and they're allowing their users to go up in context length all the way to million-type of tokens.

So this is creating a new paradigm. The large models as well as long context lengths and implied KV cache sizes are driving much higher memory capacities, while the Mixture-of-Experts type sparse models are driving compute needs down. So you're seeing memory needs going up, compute needs coming down. And we call this a new paradigm called memory-centric AI.

And in this memory-centric AI, we think HBF is going to play a very critical role. Before I dive into HBF further, I wanted you to hear from somebody who deals with these LLMs and AI inference on a daily basis at a massive scale. So I'm going to take you back to FMS, which is Future of Memory and Storage Conference in California. It was held last week with about 3,000-plus attendees. And there were several sessions dedicated to HBF during that conference. So I'm going to take you to a panel discussion that we had and going to share with you some of the thoughts from Dr. Xiao Ma, Google DeepMind. So please roll the video real quickly.

[Presentation]

Okay. So Dr. Ma is talking about a memory crisis. So next, let's listen to how he believes we can solve the problem.

[Presentation]

Okay. So with that, obviously, Dr. Ma is one of the many researchers who are actually spending a lot of time thinking about HBF as the latest and most exciting memory technology; it is really becoming increasingly an innovation platform, and researchers are proposing new architectures showing how one could integrate HBF into AI solutions. So I wanted to share with you some of the architectural proposals that have been published recently. So, the first one here is an XPU where we have taken out all of the HBM stacks and chips and replace them 100% with HBF. So this is an HBF-only architecture, very simple.

The second one is where you're sort of mixing and matching depending on workload needs and replacing part of the HBM chips with HBF. So it's a hybrid architecture. The third architecture is also a hybrid architecture. But in this case, the low-capacity HBM chips act as a caching tier in front of the high-capacity HBF.

And the fourth one is a disaggregated architecture. In this disaggregated architecture, we are disaggregating the 2 stages of AI inference -- the prefill and decode -- and optimizing the solutions hardware solutions for these in a disaggregated fashion. The prefill XPU is compute-intensive, but doesn't need a lot of memory bandwidth. So what you can do is couple a performance GPU with just regular DDR DRAM, whereas the decode stage, which is very memory-capacity and bandwidth-intensive, but doesn't require a lot of compute.

You could take a modest GPU and couple it with HBF. So you get the best of 2 worlds and combine to optimize the overall solution. So these are a few of the ideas that are coming out, and there's many more, and results of these have been published. But I want to today double-click on the first, the simple architecture, and share with you some of the workload simulation work that we have been doing on this architecture.

So for the simulation work, what we have done is we've taken a GPU actually that resembles a market-leading GPU today and has 192 gigabytes of HBM on each of them. And then we have created another version of it by replacing all of the HBM chips with HBF, and that gives it about 4 terabytes of HBF memory. So we simulated a benchmark that essentially emulates in a multi-agentic, a multi-turn agentic workload -- it simulates or emulates a code development environment.

So what happens is the AI agents start developing code, spawning more jobs, et cetera, et cetera. And the underlying LLM model here is a 490 billion parameter Qwen3. So let's see how the 2 models are sort of -- the two systems are comparing. And what we're going to measure is the token output. We're going to compare the token output of these systems. So, first off, when we start with the HBM-only system, it turns out that the minimum viable system to run this workload requires use of 8 HBM GPUs. You just cannot fit the model with less than that. So you have to use at least 8 GPUs to start this job. And here's what the workload looks like on our simulator, okay? So 8 GPUs, delivering pretty stable token output. Okay.

So now we are going to show you how an HBF system compares. So, it turns out that we were able to actually run this workload with a single HBF GPU -- 1 GPU alone. And let's look at that. Okay. Obviously, the performance is not as high as 8 GPUs, but if you just wanted to have the minimum capital spend to run this job, all you need is a single GPU. So this is the result. So next, I want to show you what 4 HBF GPUs look like, and here is the result. So with 4 HBF GPUs, we're able to match the performance of 8 HBM GPUs. So we're getting 2x the performance out of our GPUs. So how is this possible? What's happening?

Actually, what happens is as the workload starts running, it quickly starts more and more jobs and runs out of KV cache capacity. It runs out of the high-bandwidth memory capacity. The moment you run out of the capacity, you spill into the system memory, and that spill, and losing that bandwidth essentially costs you roughly half of your performance. Your GPU utilization drops by nearly 50%. And that's why we're able to deliver the same performance with half the number of GPUs.

So out of this work, we had 2 key takeaways. Number one: if you're somebody like, say, a small business or a software developer who doesn't need massive amounts of tokens, but you just want to run this job with minimum CapEx, we can improve your spending by 8x. You get 8x CapEx efficiency using HBF. The second takeaway is that at maximum token output, we are able to deliver you 2x the GPU efficiency, which means your capital will go twice as far, which means you're going to burn half of the energy and all the economics that essentially the benefits that you can gain. This is fundamentally going to change the economics of AI.

This is the crisis -- the memory crisis Dr. Ma talked about -- and this is how we intend to solve it. Obviously, we are very bullish on HBF. But we also think that it's not only for data centers. We only believe that we can dramatically change how AI is run on edge devices with HBF. We are envisioning enabling really sophisticated AI models -- I'm talking about 100 billion-plus parameter sophisticated models -- to run on edge devices and enable an AI experience that I like to call "AI that never forgets."

What I'm talking about here is an AI agent that knows everything about you, that's constantly with you, remembers everything about you in an instant, you don't need to go back and forth many times. Everything is there with you all the time. And we believe that's going to significantly change the way people are experiencing AI in their lives. To that end, we've been working on a second-generation HBF device, which we call HBF for the Edge. And this is what you're seeing. And this has been actively in development with multiple customers.

So this is what's next on the roadmap that we have for HBF. Obviously, we have a pretty big vision for HBF. And having that kind of an vision, you really need to have a vibrant and diverse ecosystem to be successful to realize that vision. We understood that from the beginning on, and when I was here 1.5 years ago, we told you that we intended to create an open ecosystem around HBF. True to our word, last August, we announced a partnership with SK Hynix and talked about our intent to create an open standard around HBF.

We followed up in February of this year, established a consortium under OCP with participation from Google and Tenstorrent -- and last week, we celebrated the release of our first public specification that's going to allow XPU designers to incorporate HBF into their systems and designs. The next step is going to be to expand the membership of this consortium. And I have a piece of news to share with you already: we have Meta joining this consortium. So we're very happy with that. And as they and other participants contribute their feedback and inputs, we're looking forward to incorporate those in the next revisions of the specification over the next 2 years.

One critical element of our ecosystem we view as the advisors that we have -- the Technical Advisory Board that we have built. You may remember I talked about this: Professor David Patterson and Raja Koduri are legendary computer architects. We're very proud to have them on board. And last week, I had the honor of introducing our next Board member, Jim Keller, to our Advisory Board.

Jim is a rockstar chip designer. Over the last 4 decades, he led teams in some of the most consequential processor designs at DEC, AMD, Apple, Tesla, Intel. And I actually started my career as a CPU designer and competed against several of these things, and it's not fun, I tell you. But Jim brings his expertise and guidance into now leading Tenstorrent as the CEO of the company. And today, I have a great pleasure and surprise for you: Jim is here with us, and he's going to join me on the stage for a conversation. So Jim, would you please come on stage. Please welcome Jim Keller. Jim, thank you so much for coming. Please. And let me hand this over to you.

Right?

Great to see you here. You flew yourself?

I did.

Thank you very much.

I had the help of an airplane.

All right. Well, that's great.

Not entirely myself, it's...

Well, thank you for being here. Jim, we talked about all of the great processors and compute projects that you led. But then now you're taking all of that into the world of AI. How has that journey been for you? And please tell us what you and your teams at Tenstorrent have been up to recently.

Yes. So a couple of years ago -- well, it's been obvious for maybe 5 years now, right, that AI is going to take over most of the data center. And there's going to be heterogeneous computing, so AI compute and general-purpose compute, but it's built on the usual fundamentals. So we built Tenstorrent around that premise. So we build high-end AI processors and high-end RISC-V processors, and we have 2 businesses. Business one is we license that IP for a variety of projects: so autonomous driving, robotics, a couple of supercomputer companies, and we're looking at some server projects, right? So we built the IP, but then we put this into our high-end server design.

So we're in production today with Galaxy. Galaxy is a scalable AI computer. And one of the things I think is really interesting, we're going to talk about this, is computing has always been based on the balance of memory, compute, and I/O, like, generally networking. And what happened with AI in the last couple of years came right towards us.

So we built the Blackhole chip that runs AI models, and that's really good for 70-billion parameter models. And we thought we'd build a Galaxy server with 32 chips per server so we can scale up. And in the last 3 years, we went 70 billion, 300 billion, 700 billion, 1.5 trillion, 2.8 trillion. And the scalability of that has been amazing. So we did something interesting in our boxes. We have 56 800-gigabit Ethernet ports per server.

And then we put those together in quad servers and then hook them. Today, 36 of our servers are all hooked together, and we run models on anywhere between -- from a single chip, now up to 20 servers. We're in testing with 36, and it's scaling really well, right? The other thing we did is this is pretty general-purpose AI. It's a combination. We have flash in the host, DRAM in the host, AI, DRAM, SRAM, compute, and networking. And that lets us run a wide variety of models on the same hardware, right?

So we announced in May, DeepSeek at 400 tokens a second. That's batch 32. This is very high throughput but very high token rates. We do prefill and decode on the same hardware. We ran Wan 10x faster than anybody else, it's real-time video. That runs on 4 Galaxy servers. And recently, we just announced 900 tokens a second on Kimi K3, I guess this is 2.6, 3-years up and running in the lab. And then we have a new higher-resolution video model, it's all on the same hardware.

And the reason we really think about this hard is AI is changing so fast. Who here heard about KV caches 2 years ago? Anybody? That's a part of an LLM, the fact that we can cache it. 2 years from now, something different is going to happen. And everything needs to be flexible: compute, memory, and I/O. And what we're going to do with really large memory is amazing because memory is one of the most flexible things, right? You can put programs, models, weights, caches, datasets. There's so much to do with that. So we're pretty excited to be here today.

Great. Thank you. Thank you very much. I can take that if you want or we can just leave it there. Perfect. So Jim, you talked about these super scalable systems that you're building. They go all the way from smaller needs to very large scales. When you look at these scalable systems, like, where do you see some of the bottlenecks?

Well, it depends on how the hardware is built. Like, today's HBM-based models, they're limited by the local size of the DRAM, which you talked about. And they often don't have enough network bandwidth. So one thing we did is we have a terabyte per Galaxy server, 36 Galaxies is 36 terabytes of DRAM. But because the network bandwidth is so high, we can serve memory from one part of the machine to another really flexibly. So I think the 2 biggest bottlenecks today: we're doing pretty good on compute, but memory scalability and then the network scalability, so you can serve the memory everywhere you want. Those are the big ones.

Great. Great. And when you talk about memory, there are a lot of AI architectures that people are talking about, and they're highly differentiated by the way they use memory. Like, we see, obviously, the most commonplace GPUs today with HBM memories. And then we're seeing architectures like from Cerebras or Groq that are relying mostly on SRAM, and now we're talking about HBF. How does this whole thing? How do you think about the variety of these memories? And how do you think about HBF in that context?

Yes. So first of all, GPUs were built for graphics, and they read and write all their data to memory all the time. And that drove them to really push hard on HBM because they don't have enough SRAM on chip. Our processor has 200 megabytes per chip of SRAM, and we can put a large number together. So the balance of SRAM to local DRAM to host DRAM to flash is really important, I think.

So Groq and Cerebras exploited essentially a gap in the GPU roadmap. But the thing that we're going to see is the ratios of compute, what we call KV cache to prefill to decode to prefill today, those ratios were 1:1:1, and then it went to 7:4:1. Now it's 100:10:1. And people are still moving that. So if you build a machine that has specific processors for different pieces and the ratios change, what's going to happen to your compute?

Yes. So, so how do you think about, like, where is like HBF, something like a very high-capacity, high-bandwidth memory, relevant? And can you think of some examples where it would be really useful?

Yes, definitely. So AI is a very high-bandwidth problem: memory bandwidth, network bandwidth, compute bandwidth. And the limitation, flash is great because it's lower cost per bit, much bigger capacity, but it didn't have the bandwidth to really play in that high-bandwidth system. So the really cool thing about HBF is now you brought bandwidth to the table. So that makes it really great. And the other wild thing is, and to be honest, I didn't see this coming, the fungibility of compute and memory is amazing.

Like, who knew compute would be so expensive that we should compute it and save the results of the computation in that big format, right? So people don't realize when you send tokens in, it's a pretty small stream. When you embed that and then compute the KV cache side, it's a very large footprint. And with HBF, it's now effective to save that for a very long amount of time. And that unlocks the ability to balance compute and memory in a new way.

Right, right. I mean, we've talked a lot about data center, but you do also a lot of work outside the data center in the edge. Do you see any applicability of HBF in the edge devices and edge applications?

Oh, definitely. So today, autonomous things look at the world, and they have to process everything, and they have a model that's trained. Having huge augmentation of KV cache for everything you see in flash on-device is going to rapidly change how robotics work. I was joking this morning: how many people here would wish GPS worked in New York City, right? Like it's, imagine you have a device that actually knows everything around you, it knows exactly where you are. And there's going to be so many transformations. And I don't know if that one in particular is going to work out.

But one day I was going: if I had enough data, this problem would be solved. And so there's a really interesting thing about robotics: everything you already know, you don't have to compute. And as we make that memory available in robotics, autonomous driving, so many applications, it's going to be pretty transformational. Memory is way lower power than compute. That's a good trade.

Well, thank you very much. This has been amazing. Thank you for coming and being with us. And I believe you're going to be with us available to answer questions after lunch or during lunch.

Yes, you bet.

Okay. Thank you so much. Okay. Before I finish, I want to probably address one question that I suspect is top of mind for many of you, which is when are we going to see HBF. So here's the latest update. I'm happy to share with you today that we actually taped out our first HBF memory die. You're seeing, actually, you happen to be the very first people in the world outside SanDisk seeing this picture. I apologize. I had to pixelate it because we are still not quite ready to share all the magic that goes into it, but our die has taped out.

But you don't have to wait too long to see the whole thing. A little bit more patience, please, but we're continuing working on this super hard to deliver our first samples to our customers of inference devices with HBF next year. Next year, just a little bit more.

With that, I thank you all very much for being here, and I'm going to invite back our CEO, David Goeckeler, on stage. Thank you, and have a great rest of your day.

I really want to thank Jim for coming all this way to support us and, more importantly, for joining the Advisory Board around HBF. This has really been quite, as I said earlier, if I look back over the last 1.5 years, a lot of really great things have happened at SanDisk. But this one, the ability to make a market and attract people as capable, saying Jim is capable is like a bit of an understatement. But people that are this distinguished in the field to come help drive this technology forward is just really amazing. So we're super happy about where this is. And we will keep you posted on product availability as we continue to drive these milestones forward.

As Alper said, the fact that we now have a die that we have taped out, this product is real. It's going through the fab. We're producing it, and we'll get it back and then we'll put the systems together and get it into customers' hands, in our partners' hands. I think this point that was made about co-development, I mean, that's what it's all about when you're developing new technology. And if you're co-developing with some of the largest customers in the world, that's a really, really fun place to be.

So as we make progress on that, as we get those samples in customers' hands, we'll get a lot more information about what the future of this technology looks like from a market, financial, all of that perspective. So stay tuned as we move through that process next year. All right. We've got, I'm not going to read through all this because you just listened to all of it. But it's a recap of where we are.

I think you can tell we, hopefully, you can tell, we are extremely excited about where we are. Like I said, I've been doing this for now 6.5 years, really trying to unlock the value of this franchise. I feel like we've made a lot of progress. Since we separated the company, we've seen both companies just bloom and really start to get the valuation. But I really do believe we're kind of now, we're entering a very different phase of where we're going to take these franchises.

The ability to really recognize the true value of this technology we've been building for decades as the market changes, we build new customer relationships, and we really get this engine running of, again, turning bits to revenue, revenue to profit, profit to free cash flow, and returning that cash flow back to you. All right. I'm going to bring everybody back, and I don't know, somebody may have a question in this group. It's been my experience that some of you often have a question.

So we're going to open it up for Q&A. We'll bring everybody back. Everybody is fair game, and we'll do our best to answer whatever questions you have. There will be a mic runner. So if you have a question, I guess, raise your hand. Okay, right here. Go ahead, Jim.

Jim Schneider, Goldman Sachs. I have one business question, one technical question. First of all, on the business question, can you maybe talk a little bit about how you're thinking about the diversity of customers you want to include in the NBMs. You talked about the hyperscale component, the data center component. How do you think about the broader mix and having too much risk in any one given end market?

And then maybe secondly, on the technical side, if you think about -- we're hearing a lot of discussions about some GPU customers wanting to reduce the amount of HBM content in their systems today. So before HBF comes to the market, how do you think about the amount of HBF or conventional eSSD content you need to add to an existing GPU configuration to deliver the same performance?

Luis, do you want to start with the customer mix?

Yes, Jim, we've been very thoughtful in which customers we sign NBMs with. We're looking at different markets. It includes data centers, as we talked, it includes edge customers as well. So we'll be looking across. And even when you look at the hyperscalers, their business models are dramatically different. So being very thoughtful in driving that diversity for the reasons you mentioned. We're betting on winning customers. We believe that they're going to be here with us for many, many years to come and the level of integration, both technical and commercial is as strong as it has ever been. Diversity of business models is one of the criteria we look at.

Jim, on the second point, I'll say a few words and then Alper and Khurram can have a point of view on the very specifics. But I think what you're drawing out is what we're seeing in the market. It's definitely what Jim just said. Like this is changing at an incredible pace, right? And that makes it difficult to understand what product to build and how much of it to build, especially with the way the market used to be organized, right, build it and we'll talk about what the price is later.

And so what it says to me is there's just a huge premium on staying very close to your customers because this is going to change. It's going to continue to iterate over and over again. I mean the scaling of inference is incredible. I mean it's one of the most -- I mean, it is the most incredible technology transition I've seen in my career by far, and I've been involved in some pretty big ones. So it's going to change very rapidly. There's going to be constant innovation. As Khurram said, there's going to be -- there's a constant focus on how do you drive the requirements down?

How do I use less power? How do I use less space? How do I make this more efficient? Because the more efficient you make it, you can scale it faster, right, and more economical. And also, if there's 3 different people scaling inference around the globe, if one of them is twice as expensive as the other, that's not going to work very well from the business model. So our customers -- the great thing about where we are from a broad technology point of view is you have these companies now that are just spectacular.

I mean they can scale on a global footprint, something this complex very, very quickly. And staying very close to them and understanding where they're going is, in my opinion, extraordinarily important for where we're going to drive this franchise. And that's another kind of angle on these NBMs. As Luis said, we have NBMs with some of the largest customers in the world. We have their commitment of what products they're going to deploy quarter-by-quarter for the next 3 to 5 years.

That gives us a lot of insight into all of this confusion of what's the product, what's the architecture, how is it going to play out. It gives us incredibly unique insights about how that's going to play out and what are the right products to build and where to put our resources to make sure they're successful. Now, one of you guys want to talk about the specific question?

Yes, sure. So David is right. Our customers provide us with a lot of data, but I'll give you a little bit of a long-winded answer on this. Like I talked about the AI lab at SanDisk Connection. You're absolutely right. The question is the system DDR, how much of it you need as you have the continuum between the volatile and nonvolatile media. It was coincidental, I was sharing the data with David yesterday from our lab where we ran a 1.2 trillion parameter model. And we ran, again, multiple user sessions, prompts. And an interesting thing emerged from that.

When you look at performance and power of the system to execute that workload, what we saw that after -- and we ran both the HBM sweep and DRAM sweep, right? Think of it as a shmoo, right, when you -- and running the shmoo around capacity. So you start with, let's say, and I'm making up the number, like 4 gig of volatile memory like HBM and DDR and then go all the way up to the maximum capacity of that server scale. And it was interesting to see to run that complex workload, having just the HBM SSD was good enough. And you really required a minute amount of DDR to run the batch services.

And that was a revelation to us. By the way, you heard the news that, hey, they're reducing on the Vera Rubin, I think cutting the system DRAM by half, the SOCAMM DRAM. I would contend with the studies that we are doing, you can go even further, right, if you have an eSSD and HBM in the system. And like David mentioned, things are going to change. With respect to the eSSD capacity, again, it's workload dependent, how much you want to run, taking you back to the persistent KV cache pool, right, how big you want to grow it. There's a lot of factors that go in, and those are more use case dependent. But certainly, one thing is emerging with the system DDR with respect to running the inference, it could be reduced. Now could it be reduced by half, one it depends.

Ben, do you want to -- where's the mic? You guys will run the mic.

Ben Reitzes with Melius. It's great to be here. First, I got an observation, which I hope some people find kind of amazing, and I wanted to react to it and then I have a question. I mean, I've been going to tech conferences and Analyst Days since about 1992, and I've never seen a company guide for a year, 3 years out and be trading at less than 3x that number.

And I just -- it's pretty amazing. I just wanted to react to that, trading at less than 3x your FY '30 number. And second, HBF. So it's not in the model, but you're only growing bits mid- to high teens. I mean, how do we put it in? Like what if this is a hit and we got to add it to the model? What are you getting rid of that we just guided for to make room for it? And how do we calculate upside if bits can only grow a certain amount?

Yes. Well -- so I appreciate your observation. I may come up with a different word than amazing for my reaction to that, but that's a whole different discussion. But look, this is the conversation I think we'll have as we get the product in customers' hands, we really understand what the demand is. Again, I'm going to sound like a broken record. We keep coming back to these NBMs. We're going to follow our customers, right? When our customers tell us they need something -- and this is the way most technology businesses work, right?

You work with the customer, they want to buy something and you build it. You don't like talk about what's the industry supply is like -- who cares? Like we have the ability to produce what we can produce, and it's about getting these incentives aligned. And what I would argue, if I go back a year ago, the incentives in this industry were just completely misaligned, right? And that leads to this -- all this volatility, and that's good for nobody.

But we're clearly walking down a path in 2 quarters, we've gone from 3 months of visibility to 4 years of visibility. So I think it's just a little bit of wait and see what it's like a year from now when we actually have this product in those customers' hands, what their demand is going to be for that. And we'll then figure out what the production plan is behind that.

And remember what I said earlier: we own the whole stack, right, from the NAND IP to the production front end, back end, the whole thing. So we're not ready to go there just yet. But imagine what could be possible in the future. All right. I'll go to the back.

Aaron Rakers with Wells Fargo. I appreciate the day and all the details. I want to go down the path with HBF as well because I think it's a fascinating technology. Alper, one of the evolutions in DRAM that we're seeing, specifically HBM, is this idea of customization, right, driving towards a base die that has certain elements of compute in it.

One, do you see that as a roadmap that you could explore on HBF? And then secondly, I guess back to Ben's question, if it's successful and we start to get into next year, and we start to see design-in, right -- and maybe that means 2028 or even 2029 volume -- should we expect the CapEx discussion to change? What's involved in the production of HBF advanced packaging? Is it a different capital intensity that we should be thinking about?

So thank you. I agree. I'm also in love with HBF. So your first question -- customization, I'm sorry, yes. Customization that we started seeing, I mean, it is a natural flow of actually compute moving towards memory, right, because shuffling data around is just so expensive, so energy-consuming. I mean, it's something that we have been anticipating, in or near-memory compute is going to definitely happen, and we're entirely prepared because we do have all the capabilities.

Khurram showed you all the memory management capabilities we have. We only welcome doing compute next to the memory or in the memory. I mean, it's our wheelhouse, and we'll love to do that. We'll cherish that, and we are highly capable of doing that. And so it's entirely a customer-based discussion -- what part of that pre-compute or the compute itself do they want to perform inside the controller that sits next to that memory. It's that conversation you need to align. But after that, you're going to see us being entirely capable of doing that.

The second question in terms of like how do we think about all the investments: certainly, I mean, one of the reasons I'm in love with HBF is because it entirely leverages our know-how expertise in NAND flash, right? We have the best flash, and that's why we were able to create HBF to start with. And in terms of what does it mean for capital and whatnot, I mean, you have seen us talk about the productivity of our current NAND roadmap generating, actually giving us the potential to easily increase our output if we choose to do so.

And this is a great outlet, right? I mean, the beauty of HBF is that as a business, it's entirely orthogonal to our storage business. So we could definitely entertain that. And as Dave said, as the time comes, we will see. Some of the other capabilities certainly are, again, within our capability range, and we're looking forward on working on all of those. So it's going to be definitely very exciting. But let's wait until next year and see how this will get into the market, and I think those are excellent, high-class problems that we're willing to work on.

Let's go back.

Mark Newman from Bernstein. Actually, I wanted to ask about the NBMs. It seems like great progress that you're sharing here again today. You talked about the target -- sorry, you talked about FY '28 being 66%, two-thirds of volumes on NBMs by FY '28. Do you have any kind of target in mind for what percentage of volume that can get to?

And related to that, like what's the volume in FY '29 and beyond? Is it also a similar two-thirds level of commitment on these NBMs? And then the second question -- this may seem a silly question -- but on the 100% excess cash to shareholders, does that mean 100% of free cash flow? Or do you have a different definition of excess cash? Just wanted to clarify that.

Yes. I thought I was -- I hope I was clear on the free cash flow. Excess cash is defined very simple. The cash we're generating minus whatever we're investing in that pocket number one, investing in the business. There is no trick. It's 100% of the cash -- excess cash -- will go back to shareholders. And if you look at what we just did in Q1, that's exactly what we did. We generated $5 billion, and we spent $4.5 billion, right, so pretty much there.

I think on your question on the percentage of NBMs, we're going to be optimizing over time. We're going to be learning a lot, and we'll define what that optimal number is. We think that -- we like the numbers that we see for FY '27 and the numbers that we see for FY '28. We think we've optimized based on the data that we have so far. But we'll keep on talking to customers, we'll keep on evaluating the situation, and that may evolve over time.

Our goal is to create the maximum shareholder value that we can on a sustainable basis. So that's the filter, and that's what going to guide that percentage. What is the number for '29? Yes, I got the question before the session started. It's consistent with the numbers that you are seeing. It's consistent with the '28 number so far, but that number will keep on evolving. That's why we didn't put it there, because we'll keep on optimizing our number. Okay?

It's Krish Sankar from TD Cowen. I had a question for Khurram. Thanks for your interesting presentation on the KV cache. The 2 pushbacks I heard on NAND for KV cache is: one, the tail latency is much longer than average latency, so that impacts system throughput; and number two, it takes a long time to write KV cache onto NAND and so lowers the lifetime of the device. So I'm kind of curious what your answer to that is.

Yes. Thank you. I saw your post-FMS commentary on this, and I wanted to send you a chart that I showed to David. Look, when we talk about latency, again, you have to look at the mix between the top tier -- the high-bandwidth tier really suffers from capacity. We all have to acknowledge that first, right? And like Jim was talking about, right, I mean, like the need for the space to have all that context cannot be serviced by that. So you have to go to the next tiers.

Now, what we see -- so far, what we see is that the overall latency is actually quite good, and it's actually workload-dependent. Like if you're looking at something that is real-time, obviously you won't go to the next tier, right, of the KV cache. But you don't need to. You can put that in the higher-bandwidth tier, right? So it's not just one or the other. It's like it depends on what you are running. And what was the second part of your question, sorry?

Endurance.

Lifetime.

The lifetime. So look, we are making continuous progress, like Alper keeps giving me great technology with very high endurance metrics. And part of this, we are learning a lot not only on our core technology, but also in HBF that's allowing us to get, like, phenomenal endurance numbers. As you saw on the warm cache that we talked about, it's exactly that. That requires, like, for example, 3 drive writes per day.

So we are increasing the endurance at the same time. But again, it's also workload-dependent where you need to read more like we use our standard NAND. But like for the warm KV cache, to your point, we are offering a 10x much higher drive write per day as you would on our conventional SSD. So you're right, but we have tricks on how to do that.

Joe Moore, Morgan Stanley. I also wanted to ask about the business mix between the segments. If you're a third data center now and by most accounts, data center is going to roughly double in the next 12 months, what does that imply for your mix a year from now? Because if it's 60% data center, you really starve the edge businesses a lot. So just how do you see that mix? And it's just -- it seems clear to me that there's not enough supply. Like, how do you think you balance out that shortage?

Yes, Joe. I mean, this is what we're talking about now. So, like, there's a couple of things, the way we think about this, right? So -- and part of it goes back to Mark's question. There's different -- we want to engage with customers on a longer duration, right? So there's different contracts with different duration now. And that's why it's a little hard to say what '29 and '30 are because you're starting to get into a time where some deals are ending and others may start and those kinds of things.

And we also want to -- we want to keep some flexibility in the system as well. But we'll be very -- as Luis said, we're going to be very judicious about where we go from here. We're going to follow our customers. It's very clear how we want to operate our business. We want more visibility. We think that's better for everybody, right?

It helps us make the best decisions. It's very clear that, that -- this model started in data center. I think that's fair to say. Their interests are different. As I said earlier, we're starting to see now that change pretty dramatically where customers are coming to us and wanting to have these kinds of deals. And what I said earlier, we want to make sure the whole portfolio stays robust. And that means you have a flow of products through edge, you have it through consumer, you have it through data center. So we want to keep all of those alive.

I want to make sure Khurram's engineers stay very busy building the best edge products in the world, just like they're very busy building the best data center products in the world. And of course, they're very busy building the best consumer products in the world. I think it's fair to say our data center mix is going to go up, right? And so we're going to continue to rebalance that to make sure that we keep all of those markets alive and we get the best return possible for all of you, again, over that multi-time horizon that we're looking at.

And so that's a little bit of a squishy answer, but that's -- it's a little bit of we know it when we see it. And the deals that get presented to us, we've got a long roster now of, like, "here's the potential opportunities we have to engage in the next contract." And Luis and I must talk about this 4 or 5 times a day -- like, literally for the last 9 months, we're constantly talking about this. First, it was the structure. How are we going to structure this to make sure it works for both of us.

And we just iterated around that. We, like, came up with some ideas of how we thought it would work. We talked to our customers, and we got that landed. So we got the model landed now. And it was -- some of the first ones were some very big ones, quite frankly, and we got those locked in, and we're moving in around that. And so as we go through this process now, we're going to make sure we do all those things I just said. How do we optimize value creation over all of these time horizons? We want to make sure we keep the full portfolio alive and running. And we want to make sure we mix into more data center. Where is the final target? It depends.

Vijay, at Mizuho. Just a quick question back on HBF. I think one of the things we saw on DRAM side is when HBM came along, it made conventional DRAM very tight. It took out capacity. Do you see the same thing happening with HBF where conventional NAND becomes very tight? And is your first gen in HBF on BiCS8, BiCS9? If you can give us some idea on what the pricing dynamics would be? And what is the capital intensity, I guess, on HBF versus conventional NAND?

You ready to give out pricing yet, Alper.

I feel it's priceless. At the same time, I'll answer your question that I know something about, which is HBF. We base it on -- again, we are using our ability to mix and match different technologies, but the array technology is mostly based on BiCS8. So that's where the capital and installed base is based on. And we're going to use the most convenient CMOS technology that will drive all of the features and capabilities and performance we need. That's going to be something a little bit more -- more than a little bit higher performance than what we use in BiCS8 itself.

I gave an example of that in BiCS9, like whatever we did, and this is not going to be too different than that. As to, like, how we think about its impact, as I said, our technology right now is giving us annually 27% growth capability per wafer. And we're using only two-thirds of that capability in how much we push into the production.

I think we have a lot of headroom to grow through the innovation that we have, right, and allow HBF to squeeze into that window when we need it. If HBF becomes even bigger than that, as I said, high-class problems. I'm sure we'll deal with that because we know how to manufacture things at a major mega-scale.

Yes, all you guys are kind of asking -- it's kind of getting back to the same thing. Like, where are you going to get capacity for it? Can it scale? I mean, one of the things that I hope you walk away from this with is NAND is a very scalable technology. We have the fundamen -- to Alper's point, we have the fundamental engine for at least another decade of line of sight to scaling, right? And I think that's an important issue, right?

I think that in the end -- Alper has taught me a lot of this, quite frankly -- the most scalable technology wins, right? And so if you get the economics right. Now, our problem is not our ability to scale. The issue is the chart that Eric put up. We have this giant discontinuous period where we end up taking hundreds -- thousands and thousands of wafers out of the system because we didn't get supply, and we don't understand the dynamics of the market because we don't have enough visibility. And that is the -- in my opinion, the answer to all of these questions -- that is the -- that's where the magic is. That's where the answer is: getting more visibility into what demand is going to be.

There's nothing we can do about demand this year or supply this year. The fab plan is cooked, right? And -- but when you start talking about the turn of the century, yes, we can start turning the dials. But the other thing you should take away, like Khurram put up those 4 variables, you multiply them all together and you get KV cache size, okay?

So again, I'll go back to: do the math. But when you have 4 variables and you're multiplying them together, when you start changing any one of them, the outcome can just be wildly different, right? So what's -- you're essentially answering this question: what should I build for 5 or 10 years from now? The people that have the answer to that question are the ones that are investing the billions of dollars to build the infrastructure. And the most important thing is to get as close to those people as possible, understand their roadmap, you help them understand what's possible.

I think that's really the magic of HBF. Like, you can't really expect somebody that is not a world-class memory expert to solve the memory wall problem. I think that would be an unlikely outcome. The most likely outcome is the people that are world-class experts in memory and storage design, which these guys have been doing for decades, are going to be the ones that help unlock the scaling issue.

And again, who do you want to stay close to? The video -- co-development is the answer. This has been -- this is not new for us, like, we've been doing co-development with hyperscalers for a generation of technology now because you're going to get this -- that's where you're going to find out how everything fits together. So I think the answers to all these questions are: we need to extend this visibility into what true demand is really going to be. And if we do that, I think everybody wins.

And like I said, we're 2 quarters down this path of kind of -- I don't want to say inventing a new business model -- but again, this is not the contracts of the past. This is not, "well, just sign this NCNR." That's not what this is at all. So, a completely handcrafted structure of how we're going to work together. Two quarters, we've gone from 3 months to 4 years. Let's see where we are next year, and I think it will help answer all of these questions.

And if I may add real quickly, please take a look at the wafers outside. Look at 8 and 10 next to each other, and you're going to see 65% growth from one to the next one. How many generations will it take DRAM to scale 65%? They're scaling single-digit percentage every generation every 1.5 years too. So you're looking at 10 years of -- or more than 10 years of DRAM scaling in one wafer. So you ask, what are you going to do if there's an explosion? I'll move to BiCS10. I'll get 65%. Not too difficult.

DRAM is a fantastic technology. There's an easy way to get more of it, like spend a lot of money, right? Let's go in the back.

C.J. Muse with Cantor Fitzgerald. I guess, Luis, first question for you and then one for Alper. As you think about the NBMs and you contemplate your annual cost downs, is that something that you will 100% benefit from? Or is that something you share inside the contracts? And then for Alper, you highlighted 4 potential solutions with HBF. Curious what your customers are kind of responding to? Is decode disaggregation really where they're focused? Or are one of the other kind of 3 solutions where you're getting the best feedback?

Yes, C.J., I love the level of detail we give you on the NBMs and you guys always want more. I'm not going to get into pricing in more detail. It just gets very sensitive given the conversations we have with our customers. I think the key message you should have is even at the lowest price, at the floor pricing at any point in time, we don't expect these NBMs to be below 80%. We expect that to be around 80%. So you can model that, and we have confidence of that over time, right? So every single year over the period, we should be able to be, even at floor pricing, around 80%.

And answering your question, C.J., when you looked at some of these architectures that I showed, like the first 2 are almost similar architectures. You don't need to -- you just decide based on your specific needs. And the disaggregated one, you could also think of that as a -- from an HBF perspective, one being very similar to an HBF-only architecture, except for at a system deployment level, you're disaggregating. So it turns out that all those 3 architectures can be served through a single architecture, and that's where we're seeing the concentration of the discussion right now.

Of course, the mobile space is very different, and that's what I was sort of showing you differently. There's a little bit of a difference. But in the data center, I think it's all going to be around those 3 architectures. And there might be some other proprietary ones as well that I'm not able to talk right now. But we haven't quite seen a customer that shows this tiered architecture, if you will -- the cached architecture -- quite yet. But it's also very interesting that the published data on that from the university research was very robust.

Okay. We'll take -- we have lunch waiting, and we've kept you in here a long time. We'll all be around to answer questions. We'll take a couple more here if people have them, and then we'll break.

This is Sam Feldman on for Karl Ackerman at BNP Paribas. David, you indicated that your technology roadmap can support a 27% bit CAGR through the duration of your NBM contracts, well above your projection for mid-teens growth. One of the ways to meet that projection is supply and CapEx discipline through the elongated nodal transitions from historical rates of every 18 months.

So first, how does SanDisk think about balancing share if not every NAND provider slows down nodal transitions? And second, should NAND cost improvement of BiCS9 and future BiCS nodes slow down to 10% or less given the reuse of NAND cell arrays and elongated nodal transitions?

I don't know if I got that second question. I think that's always a problem with 2 questions. I always forget the first question as I'm listening to the second question. I think you're asking about can everybody just speed up nodal transitions to increase supply, right? And I think the answer to that question is, look, these are R&D productivity numbers, but then you have to translate them into an actual fab plan, and that takes years of planning.

So it's not like -- it's not a dial you just turn up and down, right? It's not that simple. Like, you have to plan years in advance. There's people in our business -- and I'm sure all of our peers have all the same people, anybody in the semiconductor business. Here's our fab plan for years in advance and, month by month, what node is running where in what fab, what is it yielding, and then you add all that up and you get a bit output. So you can definitely start to change that, right?

The point we're making on productivity is we have ways to scale that are very CapEx-efficient. And again, I go back -- go back -- I don't want to keep talking about it over and over. We go back to the company split. I remember when we split Western Digital, it was like, oh, HDD is -- and it is a great business, it's 4% to 6% capital intensity. Here we are in NAND -- like, Luis, you just committed to what capital intensity?

Yes, low single digits, right?

Yes. So it's like the scalability is incredible in multiple dimensions, right? This idea that 19 generations of innovation by this incredible team has led to this technology, and it's still got a lot of -- there's still a very, very long roadmap. So that -- understanding that scalability is important to understand the business and how it shows up in different ways. And also, there's a flip side to it, right? Like everything in life, there's -- things can be a double-edged sword. If you get it wrong, you can overproduce very quickly. I think that's a little bit of the basis of your question.

That was the pre-2023 world. If you go to the pre-2023 world, it was 30% and 15%, man. I was saying the same thing: 30% growth, 15% cost downs. It wasn't true. And if you invest to that, it's just not going to work, right? So you have to throttle that productivity, which is fine. You just put a mix plan together that delivers it, make sure you stay focused on delivering it, and realize that when new businesses come in, when new innovation comes, it's likely going to come.

Another way I think about it is the scalability is going to attract TAM, right? It's kind of the way I think of it in my head. Like, the more scalable you are, people are going to start using your technology more and more because it's efficient, it can be produced, all these kinds of issues. As you're seeing that happen, you want to have enough -- there's enough horsepower in the engine that we can go at the level we need to go. We just need to not overrun the market.

And again, it's back to the same thing I keep saying. The answer to all these questions is aligning the investment horizons in this business. And where we were, and I've said this publicly a couple of times, it was just ridiculous. This idea that you're going to make 10-year investments and then you're going to hold a quarterly auction to see who buys your stuff and at what price. It just makes no sense. And I think we're rapidly transitioning to something that's very different.

Believe it or not, right? Like, there's the die-hard cycle people, and like, let me tell you how semiconductors work: this is the way it's been for 20 years. I happen to have a belief you can change the future. You can change things. If you methodically make changes and they compound over time, things can be different in the future. Sitting around waiting for the past to come back, in my view, is like a total waste of time.

Like, let's invent something new that works for everybody, and we got a whole bunch of people that want to go there with us. It's unbelievable. It's incredible, unbelievable. I forgot what the second question was. Probably too long of an answer to the first question.

Yes. It was just on cost downs and the reuse of NAND cell arrays for BiCS9.

Cost downs for BiCS9.

Yes, like cost downs going forward.

One thing I said last time, I'm going to go back, I said we're going to stop talking about cost downs, right? I actually said that. And we stopped talking about cost downs. It was -- I have to give all of you guys a lot of credit. Like every earnings call, it's like, "Dave, what's the cost down? What's the --" We're not talking about cost downs anymore. Everybody stopped asking, mostly. So that's our job. Our job is to drive efficiency. Our cost downs are ours. That's part of the productivity of the company, right?

And when we talk about them all the time, everybody else thinks the cost downs belong to them, right? And they don't. They're ours. They're because of all the great work these guys do, and the market will determine what the price of our product should be, not how much it costs to build, right? And it's just as simple as that.

Although it is cheaper in the future. That's what -- you can tell -- I mean, again, this is why it's such an unbelievable franchise. Am I going to tell you what they all are? Of course not. But are they there? Of course they are. That's the magic of what we're doing, right? That's the magic of these guys generate building 19 generations of NAND. And what I said before, if you think it's easy, give it a try, right? Right there.

Wamsi Mohan, Bank of America. Thanks for all the details around the NBMs. I think a lot of people who've gone through a lot of cycles covering different industries have looked at these NBMs through a more skeptical lens in the sort of, look, we've gone through many eras where when you are in an up-cycle, these NBMs obviously stick.

And a skeptic would say in a down-cycle, these NBMs typically have not held up very well across many cyclical industries, not just in NAND. And so I'd love to get your response on why this is kind of different in the sense of we went through hyperscalers back during COVID who had a demand forecast that they over-forecasted, and we had an inventory correction. What makes the forecast today better than their ability to forecast out to 2030 any better? Because I think some of what you're forecasting is dependent on what they're forecasting effectively. So I would love to get your thoughts around that.

I'll try to be brief. This may be difficult. First of all, in everything in life, the easiest thing is to be skeptical. Why don't we just say no at the starting line and never do anything, right? We'll all just be skeptical all the time, right? Why do anything? Why even try? It's all going to fail, right? And look, I get it, like people -- it's good to be skeptical. I'm skeptical. But you have to have a belief that you can change things and get a better outcome.

This is very important. We look at -- I look very deeply at this and people we hire. Are you willing to change things and have the confidence you can get a better outcome? It's somewhat unusual because most people won't change things because they think they're going to break something. We don't have that problem. It was already broken. Like, hint, hint, right.

My good friend, Luis, whom we've worked together with before. He was -- I was lucky enough to have him join me in this, and he comes at it with a different set of eyes and says, "Why are we doing this, right?" It's like -- there's a different way to do this. Other industries, to your point, do this differently and they get a different outcome. And so we can do things differently and get a different outcome. And you have to have the confidence to do that and then methodically do it and let that compound over time and you can get a different outcome.

So is it going to be different? Maybe not. I'm fully willing to admit that. I mean, you guys are all in the investment business. I think the line is investments are subject to risks and uncertainties, right? But where the most uncertainty is and where the most skepticism is, is also where the biggest returns are and where the biggest opportunities are. That should be the message of SanDisk over the last 18 months, right?

The returns we've driven had more to do with the massive skepticism that existed 18 months ago of where we started, what I said before, compounded growth rates, very sensitive to the endpoints. And when you start with the endpoint extremely low, like way less than even the replacement cost of your franchise, arguably some small percentage of it, of course, you can have very good outcomes. It doesn't mean I'm not proud of what we've done over the last year. I'm very proud of what we've done.

But that's why I say you have to put that behind you and look at where we are today and, like, get all that out of your head and then make a clear-eyed view of what is the most likely outcome? What is most likely going to happen, not what happened in the past, and we're just going to go back to the past. We've changed things, and you have to make your own calculus of if you think that's going to be successful. We've made that calculus.

We're very comfortable with it. We're trying to make that case to you. But we can't -- if you don't believe it and you need to see it, then we just have to wait for it to come, right? That's -- but we're pretty clear, we just put a capital return policy in place that tells you what we think. And we're going to continue to execute that very, very dramatically.

But the future is an uncertain place. But I will -- Luis touched on this. You can't imagine the level -- I think I've said this to some of you before. It's very difficult from your perspective to understand what we're going through because unless you're in this business, you don't understand just radical fundamental change that's going on.

I mean, when I started, I came into this business in 2000, and I had a meeting with one of my peers at one of our customers, right, a colleague that happened to work for the CEO of a hyperscaler. I got a call the next day saying the supply chain team and that customer was unhappy that I met with my colleague and that I needed to clear it with them first before I had a conversation with my former peer. That's how -- now we're talking to the CEO. Now we're talking to the CFO.

It's an incredible -- it's gone from this transactional, "you are somebody where we're just going to negotiate with all the time. Your technology is good, and it's respectful. There's nothing -- any issues, but it's just like this is your place in the world, and we will talk about price constantly" to we're having strategic conversations with the most consequential people in these companies.

Those people could be just pulling the wool over our eyes. They could be just a big head-fake. They could head for the exits later. That's not my calculus of what's the most likely thing to happen, right? Those people are officers of public companies just like I am. They have responsibilities. Board of Directors -- I'm on the Board of Directors of other companies. You don't sit in the boardroom thinking about, "Hey, I'm going to enter a contract so I can exit as soon as possible."

You don't think about things like that. If you think that way, you don't get the job. So it's very different. Is it a guarantee? There's no guarantees. But I think it's the most likely outcome, and we're going to work really hard to make sure it is the outcome. I said I was going to be brief, and I wasn't brief. We'll take one more.

And I won't ask about cost downs, although I've used it in estimating what the floor could be like in '29, 2030. But anyway, just on the physical AI and maybe some of these non-data center demand drivers that are out there, assuming there's still like maybe 20%, 30% of the bits that are still going to be consumed as we look beyond fiscal 2028. Like, what's the feedback?

I mean, we just keep hearing a lot about destocking in some of these edge client devices, a lot of them suffering from a lot of pain because of where memory and storage price points are at this point. Like, just help us understand the conversations that are going on. It seems like some of them are in your NBM footprints, some of these client edge customers. So how are they sort of thinking about their demand outlook beyond just, let's say, calendar '26?

Luis, do you want to talk about that?

Yes. It's a little bit of a tricky question because I don't want to get into customer-by-customer discussions, right? But we are engaged with physical AI customers at a deep level. We understand their needs, and we're working with them very, very closely. So yes, we have those relationships, and they're great, and I just can't get into too much details on what they are. But it's a market we are committed to. We see it growing, and there is a huge potential there for sure.

All right. Look, I want to thank all you guys for hanging in there for like over 3 hours now. I see the clock is at 0, which means we're afternoon. But lunch is here. Thanks for spending time with us. And well, I'm sure we'll be engaging a very detailed level going forward.

Good afternoon, and welcome to Sandisk's Fourth Quarter Fiscal Year 2026 Earnings Conference Call. [Operator Instructions]. Please note, this event is being recorded. I would now like to turn the conference over to Ivan Donaldson, Vice President of Investor Relations. Please go ahead.

Before we begin, please note that today's discussion will contain forward-looking statements based on management's current assumptions and expectations, which are subject to various risks and uncertainties. These forward-looking statements include expectations for our technology and product portfolio, our business plans and performance, our capital allocation priorities, market trends and opportunities and our future financial results. We assume no obligation to update these statements.

Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations. We will also make references to non-GAAP financial measures today. Reconciliations between the non-GAAP and comparable GAAP financial measures are included in the written materials posted in the Investor Relations section of our website. With that, I'll turn the call over to David.

Thanks, Ivan. Good afternoon, and thank you for joining Sandisk's Fiscal Fourth Quarter Earnings Call. As we close fiscal year '26, we believe Sandisk is in a strong strategic position to deliver for our shareholders and customers. The strategic actions we have taken over the past year have established a stronger foundation through technology leadership, longer-term customer partnerships, financial flexibility and operational capabilities, which collectively position us well for the next stage of execution, growth and shareholder returns.

Over the past year, we strengthened our portfolio with BiCS leadership across both TLC and QLC and the continued advancement of high-bandwidth flash. Established data center as a major pillar of growth, deepened customer relationships through multiyear partnerships enabled by our new business models, our NBM with momentum continuing to build during the quarter. We reinforced our supply chain and transformed our business model with a net cash balance sheet and the capital allocation framework designed to generate growing and durable free cash flow to reinvest in the business and return excess capital to shareholders.

The fiscal fourth quarter provided our clearest proof point yet. We delivered record revenue, gross margin and earnings per share, each above the high end of our guidance and repurchased $4.5 billion of company stock. We are encouraged by this progress and believe the long-term earnings power, cash generation and resilience of this business will become increasingly evident as we execute against this new foundation. Underlying our performance is the most important force in our market, the era of inference. AI is fundamentally a memory-centric storage-intensive problem. And it is reshaping the demand equation for NAND. The shift to inference in genic AI is generating data at a scale that is redefining storage requirements.

Every AI interaction creates content that must be stored, retrieved and served at low latency. And each of these steps relies on data storage products. including our high-capacity enterprise SSDs. NAND is the most scalable semiconductor technology in the world, and it has become a critical component of the AI architecture. This demand is anchored in strategic long-term infrastructure investments by the world's largest technology companies, which are increasingly working with suppliers who can scale partner with them and secure supply that ensures performance and reliability years in advance.

These enduring and mutually beneficial partnership give customers greater confidence in long-term supply while giving Sandisk clearer visibility into demand and a stronger foundation for planning, investment and more durable cash flow generation. Our technology leadership is how we are capturing this opportunity. BiCS has become recognized as an industry gold standard for NAND. And this year, we ran BiCS 8 to the majority of our bit production delivering industry-leading performance, density and power efficiency across both TLC and QLC.

BiCS 8 was enabled by innovations like CBA hybrid wafer bonding. And our road map builds on that same fundamental approach with future generations extending performance and cost leadership through continued innovation across multiple dimensions of scaling. Our leadership is translating directly into customer adoption across our various end markets. We scaled our compute focused TLC enterprise SSDs across a broad set of hyperscale and AI infrastructure customers. And this quarter, we began shipping our QLC Stargate platform for revenue, giving us a complete complementary portfolio spanning performance-intensive compute workloads and high-capacity AI data lakes.

A year ago, data center represented roughly 12% of our bids. Exiting fiscal year '26, it represents 38% of our portfolio and is our fastest-growing end market. Our technology leadership also extends well beyond data center. Edge remains a large and strategically important end market for Sandisk, spanning smartphones, PCs, tablets and an expanding set of emerging use cases in the realm of physical AI, including automotive, robotics and on-device agenticAI. Near term, both PCs and smartphones are working through a period of adjustment as demand is shifting towards AI-enabled devices and premium configurations, driving higher storage content, particularly in smartphones.

In the PC market, OEMs are growing revenue and expanding margin on a more profitable mix, reflecting demand for higher-end devices. We expect these markets to return to growth in calendar year '27 hand. And over the longer term, on device AI, richer content and entirely new form factors will continue to expand the role of high-performance flash at the edge. Our ability to deliver high-performance density and power efficiency positions us well as these platforms evolve, and we expect increases in content per device through future refresh cycles.

Sandisk's global consumer presence remains a meaningful differentiator within the industry, giving us a unique connection with end users and channel partners. We continue to invest behind the brand sharpen our go-to-market capabilities and innovate around the products, capabilities and experiences that consumers value. Our ability to innovate at this level is enabled by our operational excellence. Sandisk manages the entire value chain from the design of the NAND die through front-end wafer manufacturing and some of the largest fab complexes in the world with our JV partner to system-level design including our world-class controllers and final back-end assembly and test all the way to the hands of our customers.

This end-to-end integration, combined with our R&D depth proprietary BiCS systems expertise and the market diversity that gives us the optionality to direct our technology where it delivers the most value is what enables us to serve customers at attractive returns. Just as important, we grow supply primarily through nodal transitions rather than wafer additions, delivering mid- to high teens bit growth from the productivity of our technology road map with capital intensity that continues to decline as a percentage of revenue. This is a structural advantage and what makes this franchise such a powerful cash generator.

With that, I'll turn the call over to Luis for an update on our new business models and a deeper dive into our financial performance and guidance.

Thank you, David. Fiscal year 2026 was a transformational year for Sandisk. We exited the year significantly stronger than when we entered it. We believe that we're well positioned to create sustainable value for customers and shareholders. Our new business models, our MBMs reflect the stronger and longer-term partnerships that we're building with our most strategic customers, the value they place on our technology and products and the confidence that they have in their demand. Our revenue growth, margin expansion and asset efficiency enables us to deliver leading free cash flow per share and, therefore, generate an attractive return to shareholders.

Since announcing 5 NBMs during our April earnings call, we have signed 5 additional agreements, 3 NBM with new customers and 2 deals expanding on previously signed. These extensions reflect our customer strengthening demand exceeding their prior estimates. One of the 5 signed deals 3 closed before the end of the fourth quarter and 2 closed after quarter end. In total, we now have NBMs with diverse data center and edge customers and reflect the conviction of our customers having their long-term demand and the value they place on Sandisk.

The length of our NBMs varies extending up to 5 years with a weighted average duration of over 4 years. We expect our NBMs to represent more than 50% of our bids in fiscal year 2027 and approximately 2/3 of our bids in fiscal year 2028. NBMs are quickly becoming our predominant way of doing business. We will continue supporting our non-NBM business with uncommitted supply. Pricing for our NBMs include both fixed and variable elements with a variable portion subject to floors and ceilings. We expect attractive margins even at lower pricing. Pricing of our non-NBM business will fluctuate with the market.

The total expected revenue from all our NBMs we have signed is a minimum of $93.9 billion, assuming floor pricing. We believe actual revenue will be above the minimum. The remaining performance obligation, or RPO, at the end of the quarter was $59.8 billion, and would be $91.1 billion, including the 2 NBM signed after the quarter closed. The difference between the total NBM revenue and the RPO is the revenue that has already been recognized.

Each one of our NBMs include financial guarantees through a combination of cash deposits and financial instruments totaling $16.5 billion, which are intended to protect Sandisk if a customer fails to satisfy its purchase obligations under this agreement. These funds and financial instruments are mostly held by or provided through third-party financial institutions with the remaining in our cash balance. For each of the existing deals, the financial guarantees are released towards the end of the agreement. So the ratio between the financial guarantees and the remaining performance obligation increases over time.

Our NBMs are built on clear and detailed supply and demand agreements defined by year and by quarter. These features provide clear visibility for our operations and additional financial protection. Overall, we're pleased with the 8 customers we have signed as they strengthen our strategic partnerships. We expect these relationships to last for many years and to enable our customers to continue building exceptional products for their end customers. Going forward, we will remain highly selective in evaluating additional NBMs the key characteristics we look for are strategic customers that value our products, duration of around 5 years, growing volume requirements and attractive financials that enable us to invest in the business while generating a sustainable return to our shareholders.

We will be patient in these evaluations. With that update, I will turn to the results for the quarter. Revenue for the fourth quarter was $8,965 million, up 51% sequentially and 372% year-over-year. Revenue came in above the guidance range of $7,750 million to $8,250 million that we provided on our prior earnings call. Sequential revenue growth came approximately 1/3 from higher volumes and 2/3 from higher pricing. Turning to our own markets, during the quarter, data center revenue reached $2,977 million, up 103% sequentially. Edge revenue reached $5,432 million or 48% sequentially. Consumer revenue was $556 million, down 32% quarter-over-quarter.

We're pleased with this evolution of our portfolio and remain committed to serving all 3 end markets to maximize long-term value creation. For the full fiscal year 2026, revenue reached 20,248 million. up 175% year-over-year with bit growth in the mid-teens, in line with our plan. By end market, for the full year, Data center revenue reached $5,153 million, up 437% year-over-year, Edge revenue reached 12,160 million, up 195% year-over-year and consumer revenue was $2,935 million, up 29% year-over-year. Non-GAAP gross margin for the fourth quarter was 84.6%, up from 78.4% in the prior quarter and 26.4% in the prior year. This compares favorably to our guidance of 79% to 81%.

Non-GAAP operating expenses for the fourth quarter were $484 million, representing 5.4% of revenue down from 7.5% of revenue in the prior quarter as we generated additional operating leverage. This compares favorably to our guidance range of $480 million to $500 million. R&D represents close to 65% of our operating expenses. Non-GAAP operating margin was 79.2%, up from 70.9% in the prior quarter. Non-GAAP EPS was $39.25 up from $23.41 in the prior quarter and $0.29 in the prior year. This compares favorably to our guidance range of $30 to $33.

We closed the quarter with 157 million diluted shares outstanding. Key GAAP to non-GAAP reconciliation items include a gain of $807 million from our investment in [ Nanya ] and $67 million in stock-based compensation expense. We also recognized a tax benefit of $175 million from higher stock prices related to the vested employee equity, which was offset by $170 million of taxes recognized on the gain from Nanya.

During the quarter, we repurchased 2,836,000 shares of our common stock for $4.5 billion. Moving on to free cash flow. During the quarter, cash flow from operations came in at $7,126 million, partially offset by $153 million from net cash capital spending gross capital expenditures totaled $562 million, representing 6.3% of revenue. We generated $5,035 million in adjusted free cash flow which represents 56% margin. This excludes $1,938 million in NBM prepayments and deposits, which are included in cash flow from operations. We closed the quarter with $4,762 million in cash and cash equivalents on our balance sheet. With that, let's move on to guidance.

We expect the NAND market to continue growing at an accelerated pace, supported by AI in France as a tailwind. We estimate the NAND market will exceed $300 billion in revenue in calendar year 2026 up 3x year-over-year. Looking further ahead, we estimate that the NAND market will approach $500 billion in revenue in calendar year 2027. Within this time frame, we expect data center share of total TAM to expand from approximately 30% in calendar year 2025 to approximately 50% in calendar year 2026 and to continue outpacing the market in 2027. Demand from our customers is growing faster than our supply.

We, therefore, expect BiCS to remain on allocation beyond calendar year 2027. For the first quarter of fiscal year 2027, we expect revenue between $10.3 billion and $10.8 billion with sequential growth driven by both bit growth and higher pricing. We expect non-GAAP gross margins between 83% and 85%. We expect non-GAAP operating expenses between $520 million and $540 million as we continue to invest in innovation and R&D and have higher expenses related to taxes on employee stock compensation. We expect first quarter non-GAAP EPS between $44 and $46, assuming $155 million fully diluted shares.

Here is some additional perspective for modeling purposes. Consistent with our long-term objective of growing supply mid- to high teens, our capital spending will increase year-over-year primarily as we [indiscernible] yet our investment relative to revenue will come down to approximately 6% for the full year. We plan to operate at higher inventory days consistent with current levels to support our MBMs and account for higher component costs. The higher inventory levels reduced sellable bids to mid-teens for the full year fiscal year 2027.

Moving on to capital allocation. Our priority remains to invest in the business to support long-term growth and durability. We will continue returning cash to shareholders. Sandy's Board of Directors has authorized an additional $14 billion share repurchase program, bringing our total remaining authorization to $15.5 billion. We look forward to seeing many of you at our Investor Day in New York next week, where we plan to review the business in greater detail. We're encouraged by the progress made and remain committed to creating value for customers and shareholders.

With that, let me turn the call back to David.

Thank you, Luis. In summary, fiscal year 2026, was the year Sandisk refined what this franchise can be. We delivered record financial results, established data center as a major pillar of growth, secured our manufacturing and supply base through the next decade and fundamentally restructured how we transact with our largest customers. The value of our technology built on decades of R&D and tens of billions of dollars of cumulative investment, is increasingly being reflected in our financial results.

As we enter fiscal year 2027, we do so with a balanced well-structured portfolio spanning data center, edge and consumer. One that has served customers across every part of the storage market for decades and gives us flexibility to move wherever demand grows next. We have industry-leading NAND technology across TLC and QLC and with a road map of continued innovation, extending our leadership, and we are investing beyond traditional NAND and emerging technologies like high-bandwidth flash that carry the potential to change the AI memory storage hierarchy entirely. Combined with our NBMs, a net cash balance sheet and an active capital return program.

The result is a durable growth model, a valuable franchise and a business built to generate substantial increasing cash flow. We're proud of what the team accomplished this year and believe we are still in the early innings of this opportunity. With that, Ivan, let's open it up for questions.

[Operator Instructions]. The first question today comes from Ben Reitzes with Melius Research.

I wanted to ask maybe two things. Last quarter, you talked about the NBMs were being signed in this 80% margin range. I know there's a lot of puts and takes with regard to pricing. I was wondering if you can talk about whether it's in that ballpark or if it's kind of trending actually more towards the guidance for the next quarter, I would think it's more in the 80 ballpark.

And then I have -- my follow-up is with regard to buybacks. If you -- if we just round up to $5 billion per quarter, I know the authorization isn't this big, but if you commit to buying $5 billion a quarter, that's -- and you do that over the next 4 quarters, that's about 10% of your market cap. I mean is that the kind of buyback magnitude we should be thinking of? Or is it too early to kind of call the pace

Let me take the second one and then Luis can take the first one. So we plan to be -- we're walking into the buyback program, where our second authorization now, you saw we were we had a $6 billion authorization. We executed $4.5 billion in the first quarter of that. Now we've reauthorized another $14 billion. We expect to be very consistent in our execution of this program. We have a lot of confidence in the cash generation of the portfolio.

Ben, to your first question, and Louis will talk about the margins. But we spent a lot of time over the last 2 or 3 quarters really working very, very deeply with our largest customers on committing demand over -- we have over 4 years of visibility now. So we feel very good about where the franchise is -- and I'll just speak personally as somebody that's been managing this franchise since March of 2020. I am thrilled to be at this point. We're recognizing the true value of this franchise and really ramping up the shareholder returns.

But Luis can talk about the margin.

Yes, Ben, we -- not to pick a specific number, but we do expect to be around 80% for the new business models. As you know, there is some upside as prices continue to go up, we'll capture some of that offset as well. So -- but we feel very good about the work we've done. We'll talk about more about the MVMs. I'm sure, but you've seen there are not a lot. There are really meaningful deals that we did with 8 very strategic customers of ours. So we feel very good about that.

The next question comes from Mark Newman with Bernstein.

Your Q4, I don't think you've given obviously Q4 strong results. I don't think you've given the breakdown yet on the volume growth. You've given some numbers for the for the year. So we've got some pretty good estimate. But I wonder if you could give some clarity on what's in there for Q4 for volume and also in the Q1 guide. Particularly for the Q1 guide, it seems a bit light compared to expectations. Share price has been down a bit last time I checked in the post market. I just wondered if that is volume related or if that is because the locking in of pricing and so less price upside or perhaps a bit of both? Or is it a bit of conservatism really appreciate it.

Yes. So in Q4, Mark, what we mentioned is that about 1/3 of the growth came from incremental bits and 2/3 came from pricing. So you have that split there. And obviously, as you compare us versus others, there is a different timing on when price increases were taken. So you have to look at things over several quarters to really understand where we are because as you know, we took significant pricing in the prior quarter. If you look at Q1, we expect both the growth to come from both bid growth and modest price increases. So we expect both of them to be contributing to our revenue growth.

The next question comes from David Gibson with MST.

Two questions. Just wondering if NVIDIA's context memory CMX standard and their plans to storage next opportunities for you with products to come. And then the second, you announced the other day the global standard with SK Hynix for HBF. Just wondering how far out are we talking about for samples of such a product? Are we 1, 2, 3 years away? That's my question.

So David, sure. So look, there's a ton of innovation going on right now in inference memory architectures, which we think is fantastic, very healthy. There's a lot of different ideas. We're going to dive into this a little deeper next week at our Analyst Day, where we'll actually go into how we think about the problem and kind of how you dimensionalize it. But yes, all these are opportunities for us. We think, especially inference is a memory bound problem. Storage is extraordinarily important to the equation.

We showed some stuff in our FMS keynote just a couple of hours ago about how when you use HBF, how we simulate performance and maybe break through some bottlenecks. So we think there's an enormous amount of innovation happening as we scale inference, we think it provides an enormous opportunity the biggest. The way we're thinking about this is staying very close to our customers because they're going to be the ones that define what the architecture is in the future. I mean all of us as suppliers will provide a lot of good ideas to that, we'll work with them closely on which ones are going to be the predominant ones for what they're optimizing for.

And these are really where our NBMs. We think they're extremely important in that we have, we have NBMs with several of the largest hyperscalers in the world. They've given us forecast years into the future that include quarter-by-quarter, month-by-month mix. So we know very clearly what products they're going to deploy and the intensity that they're going to deploy them in. And it keeps us very close to the conversation as their deployment architecture changes. It lets us reflect that in our road map appropriately. So there's a lot of good ideas out there, a lot of flashing green light for innovation we think.

We know we're a super innovative company, and we think that provides nothing but opportunity. Now on your second question, HBF is something that we announced almost 1.5 years ago now. It was very much targeted at this whole idea that inference was going to require a different storage and memory architecture we feel very good about where that product is at. We'll tell you a little more about where we're at specifically next week, but we feel very good about deep conversations with customers this week, you saw additions to our advisory board.

At FMS, you saw some very significant customers talking about the technology and how they could use it. So -- so we'll have more to say as we progress with technology on actually shipping and release dates and all of that. But from where we were 1.5 years ago that this was an idea and a lot of research and some work that we thought it was a great idea to where we are now. We've come an enormous distance and we'll talk more about that next week.

The next question comes from Jim Schneider with Goldman Sachs.

When you think about the composition of NBMs you've already signed and the ones you may intend to sign, how are you thinking about the mix of desired customers you'd like to see across edge hyperscalers, AI data centers and so forth. And maybe can you talk about whether you're pursuing additional agreements with large hyperscalers beyond the ones you've already signed?

Yes. Jim, we're open to signing deals with companies that really meet the criteria that we talked, right? One is they need to be customers that are highly strategic that they really value our products. and they're creating amazing products with those. We want customers that are looking at several years, ideally 5 years of agreement with growth year-over-year growth. That's very important because every year, we're producing more bids as we introduce new nodes. So we want customers that grow with us. And we won attractive financials.

So those -- that's the criteria, but we're not close to any segment. As I mentioned, we signed deals with customers across data center and edge, and we feel great about that. In terms of the hyperscalers, you're talking about specific agreements, I think it's not in our interest to be disclosing specifics about contracts. But we feel great about each of them. They are, as David said, long-term relationships they're getting to a level of detail that we've never had before from a technology, from a commercial point of view, and we believe that these relationships will last many, many years.

So Jim, just a few more comments on that from my perspective. So first of all, we really started this journey just 2-plus quarters ago, here we are with a little bit more than 50% of our supply for fiscal year '27, which we're a month in now. is already committed. We understand the financials. We understand exactly where it's going, it -- we already have POs for the year from those customers in FY '28, that steps up to 2/3, roughly 2/3 of our supply is already committed. We understand what the mix is. We understand what the economics are. So we feel like we've just made incredible progress here on taking -- a year ago, we were talking about visibility in this business in 3 months, and now we're talking over 4 years.

Committed financials and understanding the mix and working with, as Louis said, some of the most enviable companies in the world So we feel really good about where this is at. And we'll continue to talk to customers. I think one of the most interesting dynamics is some of our biggest customers are already coming back and running more, right, from just what they thought they needed 3 months ago. It's a very, very robust demand environment, especially in the data center. So we feel good about where we're at. But we're getting a lot of the portfolio spoken for at this point, and we'll be selective from here about how we add to it.

But we're definitely still in deep conversations with additional customers.

And then could you maybe address capital allocation at a somewhat more holistic level. I mean it's great to see the buybacks you've already executed in terms of the authorization. But how should we be thinking about your ability to do sort of more programmatic buybacks on sort of an ongoing basis? And maybe talk about the idea of a dividend if it's -- if you consider that.

Yes. So our priorities continue to be the same as what we've talked since day 1, which is to first continue to invest in the business. And although that's a generic statement, you've seen us do that, right, through our OpEx, our CapEx, through investments in through our JV expansion. So we continue to make our company more robust and more durable over time in any scenario. So we'll continue to do that, and I've talked about some of the numbers, some of the requirements to continue to invest in the business. Priority #2 that we've talked, which is frankly completed, was to get a good cash balance, which we've done and to get rid of our TRB, which is also done.

So that priority #2, we feel very good of where we are. And then really our role as a company is to return cash to our shareholders. We believe that at this point in time and things may change as time goes on, we believe that the best way to do it is to return cash to investors via share buybacks. So we started that journey very quickly in last quarter, and we continue that journey, and we'll keep you updated. But that's the form we believe it's better for our shareholders, more tax efficient and probably the right thing for us to do at this point in time.

Next question comes from C.J. Muse with Cantor.

I guess two questions. First, with pricing up modestly sequentially, can you help me understand why gross margins are guided lower? And then bigger picture, the challenge for semi-investors is moving less of a focus on margin and EPS revisions, but rather buying into the durability of the cycle. So can you focus on why you're so confident in the securities of NBMs and that today's supply is limited and that will drive a durable and elongated cycle.

Yes. So I'll start and then Luis can add on. CJ, I think you answered your first question with your second question, which is we're focused on multiple things here, not just one. Durability is a big piece of it. We want to get a fair return for our product. I think mid-80s gross margin, I would characterize as a fair return. And then we want to increase visibility and durability of the franchise, right? We want to get this kind of boom and bust out of it. It doesn't work for anybody. In '23, we are having very different earnings calls that we're having now.

And we want to get more consistent and deeper relationships with our customers. So we can plan better and they can plan better. And so we're executing what we said we were going to do last quarter. We're committing supply to customers that are willing to commit for years in advance at the economics we guided to last quarter, which Ben said earlier, we guided about 80% gross margin. And then the rest of the portfolio floats. And then there's all kinds of things that happen in their mix and all different kinds of things influence it. It's a fairly complicated business. So when you put that on top of half the supply is committed for FY '27, you get to kind of the numbers we're talking about, which we're thrilled with, quite frankly, because the amount of the operating leverage. Operating profit and free cash flow that drives is very substantial.

And again, we'll go into why that's the case a little more next week. So why do we have conviction that these customers are going to hang in there with us. There's lots of reasons for that. Number 1 is we put a contractual structure in place where we align our incentives. They're making financial commitments to us that if they don't follow through on their commitments that I think $16.5 billion in aggregate would flow to us our balance sheet. We don't ever expect to see that money quite frankly. It's just to align interests. But if some black swan event happens, companies have to do what's in their interest and maybe something happens.

But more importantly, we're dealing at a very different level with our customers than we were a year ago. I mean, we are literally talking to the CFOs, the CEOs of the largest companies in the world. This used to be, quite frankly, just a supply chain conversation every quarter in price negotiations 7 by 24. The business has turned into a highly strategic. Sandisk has incredible products that allow them our customers to build incredible infrastructure to serve the world with things like inference, which is -- we're just getting started on scaling. So it's -- as somebody that's been in the technology industry for 35 or 40 years now at quite frankly, a very high level the level of customer engagement we have is as high as any technology franchise I've ever been a part of. It's incredible.

And so we believe that we have pivoted to a highly strategic supplier to our customers. their demand continues to grow. I think one of the more interesting things, again, I think I said it before, we've already have customers coming back for the second round of NBMs a quarter in. And so that gives us an enormous amount of conviction that this franchise is set up for the long term. We have over 4 years of visibility now. economics, which are extremely attractive. And we're really looking forward to executing this business over the next several years.

The next question comes from Joe Moore with Morgan Stanley.

You talked a little bit about gross CapEx coming up a little for the BiCS and Big 10 transitions. Can you just talk about how you think about that and the fact that the guys who also make DRAM seem to be focused more there. Any thought of spending more to accelerate those technology transitions? Or just how should we think about your decision-making process on CapEx?

Yes. So Joe, basically, we continue to be committed to grow bids, mid- to high teens, right? And we're spending to that level. The reason why our dollar increases from 1 year to another is because, obviously, as you transition from one node to another, you will first take the ECS right or the cheapest transitions and then you would go on and do the more expensive ones, particularly when we had underutilized the fabs in the prior year, we had some tools that could be reused and we were in a good place on that spending. So we're very good.

We continue to execute the plan that we had. We believe that's the right plan to do it, that allows us to provide the big growth that we believe is sustainable in the market. and we haven't changed that position. Now this year, as I mentioned in my prepared remarks, it will be a little bit lower. I mentioned mid-teens as we continue to build capabilities for the new business models. We need to build a little bit of inventory just to make sure that we're always delivering on time. So that mid- to high teens, which is our ongoing number will be a little bit lower for 2027.

Okay. So if we think about a $500 billion kind of industry number for next year, do you think you would -- between the different puts and takes kind of hold segment share relative to that?

That is correct.

That is our goal, Joe.

Next question comes from Karl Ackerman with BNP Paribas.

I have two related, so I'll ask at the same time. Consumer revenue is down a bit this quarter. Is that allocation choice or demand driven? And how should we think about the dollar commitments of these contracts relative to the perceived consumer customer demand through 2028. In the absence of LTAs, how do we gain greater visibility on end market consumer demand.

I'm not sure if I got the last portion of that. Would you mind repeating that one part?

I'll go on the consumer one first, Karl, then we'll come back to you. Consumer is a business that just -- it doesn't move at a pace that the transactional markets move out. It's kind of an interesting business because in normal times, you can move pricing and consumer around more quickly than some of the other markets. But in these kind of markets where the other ones are moving so fast, it's hard to move the consumer up as much as possible. So there's no doubt we're seeking where is the right equilibrium point for pricing and the amount of products shipped and all these kinds of issues in consumer, and that's a little longer process in some of the enterprise markets. Luis, you want to add anything to that?

Yes. I mean, as you can imagine, prices on the consumer market have also come up and there has been some impact on the TAM itself. We're very happy with that business. It's going super well. We're very happy about our market share within that market, and we continue to be committed to drive that business. Do you mind going through your second question, Karl?

Yes, Luis. I was asking, you gave some very substantial numbers with respect to long-term agreements across your customer base. And I was wondering how we should think about the dollar commitments of these contracts relative to the perceived consumer customer demand of these LTAs. And in absence of LTAs, how do you gain greater visibility on end market demand for consumer?

Yes. So the LTAs or the new business models are not related to the consumer business. If you are questioning whether the new business models we're taking bits away from the consumer market. Is that what you're asking, Karl?

[indiscernible].

Yes. I think we feel very good about the demand that our customers with NBMs have if we -- otherwise, they wouldn't be increasing their demand to us just a few months after they sign their commitment. So if the question is on their demand, I feel very good about their confidence, their conviction on their demand as these numbers. I think they are actually being conservative, and they make convenience to us as they come in and increase their numbers just a few months after signing the deals.

The next question comes from Aaron Rakers with Wells Fargo.

This is Mike [indiscernible] on Aaron's behalf. I wanted to ask, it sounds like you guys have significantly improved your visibility, you keep mentioning greater than 4 years. So I'm curious with that and kind of your thoughts on supply growth. When do you see industry supply-demand converging? And kind of how has that view changed relative to 3 months ago?

Well, I mean convert to every quarter. Yes. I mean industry supply and demand converges all the time, I guess, right I mean it's -- that's the way the market works. I think if you're saying when does is supply going to increase, so everybody gets everything they want at the price they want, that's a much more complicated question. Look, I think this is really -- there's a lot into the question you're asking, and I don't mean to make light of it at all because we -- we take this very seriously. And this is really a big part of the NBMs.

I think the thing that we've struggled with is to understand what is demand 5, 10 years from now. And when you're -- when the business is transacted on a quarterly basis, that's very difficult to do, and it's very difficult to get wrong. And if you get it wrong, the implications are kind of tragic, right? We saw that in '23. Right now, we're seeing -- maybe people didn't anticipate that things were not going to be as available as they thought in the market's reacting to that. I am actually extraordinarily optimistic that the market is reacting at an incredible pace. I mean, again, if you just think 3 quarters ago, 4 quarters ago, this whole market was transacted quarterly.

And now we're sitting here saying we have 4-plus years of visibility, and we have customers signing up for 5 years of demand. That is exactly how supply and demand is going to get matched. It's not going to get matched at "the industry level, like we keep saying industry this industry that it's going to get matched with customers and suppliers having deep discussions about supply and demand and how do we get those aligned. And I think we've taken two very big steps down that path over the last couple of quarters. As I said earlier, the level of strategic engagement with our customers, it is difficult to comprehend how advanced it is versus where it was 2 or 3 quarters ago. So I think we're making incredible progress on this, and we're going to follow our customers.

That's the simplest way to kind of run a business, right? We have customers we now have relationships with 8 customers that are some of -- they're just incredibly large customers. and they're going to give us visibility of what their demand is. As I say sometimes, they need NAND, we build NAND, right? It's a match made in heaven. We just need to get the economic model figured out, and we are making extraordinary progress on doing that, and I expect that that's going to continue, and we're going to continue to have those discussions, and that will inform our investment decisions.

The next question comes from Blayne Curtis with Jefferies.

I had two questions. I wanted to ask on Edge. Obviously, the PC smartphone market are probably going to be down for most people in the back half of the year. How are you thinking about that segment? And I guess, can you just speak to the strength you're seeing? I know it's broader than just those end markets. So what is going well? And how do you think about the markets that will be down over the next few quarters?

Yes. I mean they're great markets. First of all, again, as I've said this before, this is one of the great things about the NAND business. It's got tremendous diversity of markets. smartphones is very, very important. There's no doubt those markets are adjusting this calendar year. And we said it in the script, we expect them to stabilize next year. But we see units down mid-teens for both smartphones and PCs this year. But we still see in smartphones, we see average capacity up significantly, mid-teens and PCs, we see it flat.

But then next year, we see units flattening out in both businesses, both markets and then average capacities going up. So we see the market adjusting this year and then next year return to overall exabyte growth across both. So again, the market is adjusting to kind of the reality of where the way the industry works now, and I think it's happening pretty quickly.

And then I just want to go back to a prior gross margin question. I'm trying to understand the margins are down modestly, but pricing is up. I just want to understand is that the impacts of these new business agreements that's kind of capping, -- leading to the modest growth in ASP. And then even with that, is there any other factors that are contributing to gross margin in September?

Yes. I would not assume that our new business models are a drive on gross margin. They are good margins. If you look at where we come from, right, the last 5 quarters, we've expanded gross margin every single quarter, starting with 27% all the way to 86.5%. So we feel that we're driving gross margin. We believe it's one of the most important metrics to drive the financial health of the company. Now if you look at some of the reasons on the math that you're questioning, right, I mean there is some mix. There is -- the high end of the guide is 85, right? And so it's a little bit of versus what we just printed and we're making some prudent assumptions, if you wish, on component costs and other things. And when we factor all of that in, we believe that the right guidance to give you is somewhere between 83% and 85%. So slightly down, slightly off from what we printed this quarter driven by all these factors.

[Operator Instructions]. The next question comes from Wamsi Mohan with Bank of America.

That was kind of timely to go to one, but stuck with the one question, man. Well, I will ask a 2-part single question. So you have an industry forecast here of the NAND industry growing to $500 billion in '27 from $300 million in '26. And we know that your BiCS growth is really not meaningfully accelerating for the industry. So just is the assumption that there is going to be any price normalization that we expect in 2027 or not. That's like the first part of it. And the second part is in that growth of the industry, is your expectation that you would grow outgrow the industry in line or undergrow the industry as you think about that market growth?

We want to grow with the market, we plan to grow with the market. Look, I mean, Wamsi, we're transitioning our business into a more predictable business where we're working on pricing and predictability and duration. And so we expect to grow with the market as it goes, and increase our visibility. We continue to see very robust demand through the and into '28. I mean, customers -- again, customers are giving us demand signals for all the way out to the end of the decade at this point. So I don't know what more to say about it. We're going to grow with the market. And as we grow with the market, I think we're turning in market-leading profitability. And we've got our bit growth plans.

As Luis said, we're going to be carrying a little more inventory as we go into these NBMs, which will have an impact on growth in the near term. But we continue to see a very robust market and it gives us this opportunity to both continue to drive significant profitability of the business and increase the duration of our visibility very significantly.

Next question comes from Asiya Merchant with Citi.

Great. David, I think in the past, Sandisk has talked about the KV cash opportunity. Can you just remind us like how you're thinking about it? Clearly, Agentic AI is gaining a lot of traction here. I'm here at FMS and seems to be a lot of discussion around there as your views or dialogue changed over the last quarter as it relates to and how you're thinking about this [indiscernible] in opportunity into 2027 and beyond.

Yes, it continues -- as it continues to mature, I would say, and that's what's staying close to the customers because it's use case dependent on how much KV cash is going to be, how it's going to be used, how much NAND is a part of that equation, how do you configure the system. And we continue to do an enormous amount of research in this area. And you'll hear more of that next week, but we just continue to get more optimistic on the requirements for NAND as AI gets more sophisticated, models get bigger, context lengths get longer and then Agentic is just a big multiplier on top of that. So we continue to get more bullish on the requirements for NAND.

And again, this is why I'm going to keep going back to the same theme. This is why it's so important to stay close to our customers because I am a very, very big believer now that the customers are driving the requirements in this market. It's not what the suppliers are coming up with on different architectures and those kinds of things. Now we're talking about scaling inference globally. And that is very, very difficult. And to do that and understand exactly how that system is going to work, you need to really understand the use case. And that's different for each provider. So again, why is the reason why we're staying so close to our customers, why we're increasing our visibility because we're going to need to stay very close to those architectural discussions. They're evolving very quickly. and the demand continues to get stronger.

Again, witnessed by we have customers coming back after only 1 quarter and wanting to increase their demand for the next 3 to 5 years based on what they've learned in the last quarter. So it continues to be. It's quite a complicated calculation. We're going to try and demystify it a little bit next week with how we think about it. and it continues to be a very strong story and getting stronger.

The next question comes from Vijay Rakesh with Mizuho.

David and Luis, just two-part question here, too. When you look at the Big 10 and high-bandwidth flash, wondering if you're seeing any price premium on that -- and how should you look at the price premium versus conventional NAND, I guess? And then you mentioned that 2027 growth, you should be in line with industry. What would drive the upside for you, I guess, versus the industry growth I think it's a little early to talk about pricing on some of those nodes just yet, right, and some of that technology, but we'll keep you updated on that. We're certainly very proud of the technology. [indiscernible] a great node, we just announced it, I think, last week or last couple of weeks, and it's been in development for a long time, and we're very -- it's great.

Again, [indiscernible] is going to show you more about that next week as well. in high-bandwidth flash, we continue to have deep engagements both with cloud customers and device customers about using that technology as an inference platform. Look, we grow -- Vijay, we grow through nodal transitions, right? I mean, our technology is so productive. We can grow in excess of the market rates we're talking about just through nodal transition. So that's the way we grow. If we're going to speed up or slow down nodal transitions, even that still takes a significant amount of time, but that's something we always look at and allows us to kind of always stay in line with where the market is at.

This concludes our question-and-answer session. I would like to turn the conference back over to Ivan Donaldson for any closing remarks.

Yes, I'd just like to say thank you to everyone for joining the call today, and we look forward to speaking with everyone throughout the quarter.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.

All right. Good afternoon, everybody. Thank you for joining us at the 2026 Mizuho Global Tech Conference. It's my distinct absolute pleasure to welcome back CEO, David Goeckeler, and CFO, Luis Felipe Visoso. I hope I got that right.

That was great. Thank you.

Perfect. Thank you for joining at the Mizuho Tech Conference. Again, what an year it's been, David.

It's been a great year. Thanks for having us back.

Last year stock was around $60 at this time.

Is that right?

It's $1800 now, 30x. And I think that's a mic-drop moment, I think the fireside is over. We should we going from here.

I kind of have to drop the mic [indiscernible].

Yes, yes. I think if there was a Times Magazine for the best of the company and best management right here, guys. And what can I say, I think -- thank you for joining us again. Welcome back. We really appreciate it. What did they say, AI is hungry for compute, but survives on storage.

Very nice, very nice.

There you 0go. We got -- my team did some good work on you guys. Trying to get these 1 liners done. So let me start off...

Hey, before we start, if you don't mind, just want to read this. We will be making forward-looking statements in today's discussions, including with respect to our technology, our products, our business plans and performance, market trends and opportunities and our future financial results. These forward-looking statements are subject to risks and uncertainties and actual results may differ materially from expectations. Please refer to the risk factors outlined in our annual report on Forms 10-K and 10-Q and other filings with the SEC for more information. Thank you, Vijay.

Fantastic. I think we got all of it. So I just want to start off with a quick run through the numbers. Last year, fiscal '25 June, you guys are doing $7 billion. This year, fiscal '26 June, $20 billion. Next year, Street consensus is like $45 billion. What's happening here? Just kidding. But I mean, I think I wanted to take a step back and David maybe give us a little bit of the lay of the land. What are you seeing on the supply-demand side, NAND bit growth has been growing low 20s. And how do you see that going forward? Demand is obviously accelerating. But maybe you can take a step back and give us a peek into what you are seeing?

Yes. So your question, what's happening here is, like anything in market, this dynamic, a lot of things are happening at the same time. So it's fun to be here, by the way, to be here with you in a year on. This is also the same room where we actually launched the company. We did our...

Yes. Hopefully, next year when we come back, it's a 10x. So...

We did our Investor Day here. So it's good to be back here. And if you look back to then, we had the same level of conviction we have now. I mean we see a market with sustained mid- to high teens growth. And we always thought that the technology wasn't fully appreciated for the value that we were bringing. And I think over the last year, we've done a good job of figuring that out, right, and kind of understanding what the value of the technology is. Clearly, there's a lot of AI demand that's helping that equation. Data center has now become -- or is quickly becoming the largest market in NAND. And we've had many, many revisions of data center CapEx going up. I think, Luis, you were telling me earlier, it's like 14 revisions now...

14x.

14x, all going up. And a lot of that is as AI moves into inference, I think our customers are figuring out the architecture of how does -- the microphone got way. Our customers are working on what is the architecture for inference. And that's when NAND comes into. NAND becomes a big part of that equation. We've seen it with -- people talk about with KV Cache or REG. NAND has always been the most scalable semiconductor technology. So as you start to scale any architecture, I think it's naturally going to come towards NAND if you need storage. And I think that, that -- it's a very difficult question, I think our customers have, which is what is the right architecture for scaling inference on a global basis? And figuring out what's the right concentration of processing power, what's the right concentration of DRAM, HBM, what's the right concentration of NAND. It's very use case dependent, trying to figure out what the use case is several years in the future is an issue all of its own. And I think that as our customers go through that equation and figuring that out, I come to the conclusion that having longer-term relationships with suppliers of NAND is a very good idea. And I think that's what we're putting together with these business agreements. And it's kind of changing the trajectory of our business, I think, in a very positive way. And it's a very powerful relationship between us and our customers. And it's very much of a win-win conversation. So a lot has changed in the last year. But I really do think we're just getting started. I mean we are -- I think AI is -- there's been a lot said about AI, but extremely fundamental technology shift that we're clearly in the very early innings. I think the productivity that people are seeing out of this technology is nothing short of spectacular. And I think our technology is a big part of delivering that on a global basis.

Yes. I want to get to those LTAs and NBMs a little bit later, so we can get Luis involved in that. But I want to get back to the technology road map. Obviously, your enterprises SSD has grown massively. It's almost 25% of the revenues now. Used to be a much smaller number, might be mid-single digit last year, it's grown 7x year-on-year. How -- and it's just starting off because you guys are just starting to get qualified on enterprises SSD, QLC SSDs, et cetera. How do you see that as you -- if you were to roll this forward 12 months, how do you see that market growing? How do you see your wins? How sticky is this with the CSPs?

I mean so first of all, you're 100% correct, it all starts with the technology, right? If you don't have great products -- I mean we're a technology company. You have to have great products. That's -- let's take the basics. And we do have great products, all the way from the fundamental NAND, which we can talk about into the portfolio we have. And it's been no secret, we've been investing heavily in enterprise SSD. It's a part of the market where we've been underpenetrated. And it's about building the right portfolio to increase our mix in that part of the market. There's really two major categories of products there. There's a performance-based TLC NAND products, used a lot for KV Cache and a number of things. That's been the driver of the portfolio over the last few year. The second part of the portfolio is the storage-based product. It's called Star -- project Stargate. We talked about it again here back in February when we launched this company. This is the first quarter we'll recognize revenue on that product. So you're right, we're just getting started on that leg of the portfolio. So obviously, we've got one side of the portfolio in kind of full bloom and declared across multiple, multiple customers and driving the significant growth you mentioned earlier, and now we've got the second leg of that growth coming. So where is that going to lead to over the next year? I think you're going to see the mix of data center go higher. I'm not going to put a specific number on this because we -- what we like to have is a lot of optionality in our portfolio. Every quarter is different. It's changing a little bit now with the business models, and we'll get into that. But we believe that a robust portfolio where we can cover as much of the market as possible is the best for long-term profitability of our franchise. We're unique in the fact that we have a global consumer business that already gives us a great starting point. We have a great client business, and now we're building out that great enterprise SSD business. So I think it puts the portfolio -- it puts the company in just a fantastic position as far as what options we're going to have in the future to get the best financial return.

Yes, definitely. I think talking about optionality and trying to optimize the business to get the most profitability out of it, obviously, when you're looking at the enterprise SSD market, it gives you good visibility. You have massive orders coming in from the CSPs. On the other side, you have exposure to many of the conventional markets where you have much better pricing power. How do you decide where to allocate?

You want to take that, Luis?

Yes. So we like all our customers, right? And we want them all to be successful. And at the end of the day, we produce a wafer, and we have to make a decision of where we are located. We want to have a portfolio that's balanced across segments, right, because that's more sustainable over time. We don't want to maximize value just for this quarter but over time. But we're constantly making choices on where we allocate bids, and the new business models are a foundation for that.

Got it. So let me flip that question. You see pricing in conventional NAND going up. When you look at these SSD, LTA, so the contracts that you get on the CSP side, are those kind of tethered to where market pricing would be because you don't want to leave money on the table either, right? I mean it's not running on -- a charity, you're running a business model. So how does that -- how do those arguments work out?

Yes. So in general, there are fixed price components within these new business models, which I believe are important. And then there are parts of the agreement are with, let's call it, a floor and a ceiling of pricing. And the reason we did that is because none of us wanted to be unhappy, right? If prices go up, then we would be unhappy because we would not be capturing the upside, if prices go down, our customers would be uncompetitive because their peers would be paying lower prices. So we established this floor and ceiling concept for some of our contracts and for some of the [indiscernible] time within that. Now importantly, as we said in our earnings call, even in the low end -- prices, we like the margins, right? This is how we structure them and margins will be consistent with the margins that we guided in -- for the fourth quarter for our fiscal fourth quarter.

Got it.

So there's kind of a little implicit assumption, I think with your question, I just want to kind of touch on. We're not trading duration for price, right? That's not the value proposition. The value proposition is continuity of supply, and price is price, right? Price is whatever is fair for both of us. You can assume we have somewhat unique insights into price being in the market. I don't like to discuss price. But -- and then the other thing I'll say, look, there's 3 -- I think there's three things we're trying to do. And there's three things, I think, in any technology franchise will want to do. Number one, we want to get a fair price for our technology. We work very hard on it. People are professionals, we want to get a fair price. I think we've kind of gotten to that point, right? You can argue, could we get more? If we can get more, we'll get more. That's our job. We'll continue to have that conversation. Number two, we want to get rid of the volatility on the economics. And I think especially in our industry, this is an issue. There's been so much volatility. It'd be down, cyclicality. It's -- I've said this many times, it's just corrosive on the industry. It makes it difficult to invest in the industry. It makes all the decisions harder. So we want to work on that. And then the third thing we want to do is we want to grow, right? We're a business. We want to grow as a business. Well, I think one of the most interesting things about our technology franchise is we have a built-in growth lever of more volume every year. We're talking about mid- to high teens bit growth at more volume every year. I think that's fairly unusual in a large technology business, just like next year, we know we're going to have more volume. And we have a debate sometime about, well, should you be growing faster? And the issue is all three of these variables are related. If you start growing faster, you oversupply the market and then pricing comes down and volatility goes up. And so it's really about balancing all three of these. And that's really what we're trying to do as a management team. It's got all those three balance, to your point, not leaving anything on the table. But you -- the future is uncertain. So you could get to the future, and you find out it tips in the other direction. And so that's where the new business models are really getting at that second issue. How do we put -- that's why we call them new business model. That was Luis's name...

Very innovative.

How do we put a new business model around this? Our business with our customers in a way that we can achieve all of these things. And I think we're not done yet, but I think we're making good progress on that.

Got it. Going back to the cyclicality point that you just raised, that's -- it's a very interesting point because this industry has been plagued with that whole -- it's kind of a labeled very cyclical, and so it comes with a much lower multiple, I guess. How do you convince the investor base that this time it's different? You have better visibility. There is a technology road map here. The players are more rational, more disciplined. There is more of a focus on making profit than being cyclical supplied. So...

Yes. Well, there's nothing I can do about other players in the market or anything else. I only -- we only really manage our company. And what we can do is be transparent about what we're doing and explain what we're doing and why do we think that the way we're doing it is better than the way it's been done in the past. I think generally, that's what we do all the time, right? We're innovators. We invent things, we do things differently. We have the confidence that we can change things and get a better outcome in the future. We're not [indiscernible] to the past. And believe me, I've been told many, many times about the cyclicality in this industry. Whenever you bring up the word LTA, the first thing the [indiscernible] they won't work.

They start rolling their eyes.

Like there's so much scar tissue, and there's so much history. I get it, like how do I disprove something? It's very hard to disprove something except you just keep putting points on the board, that's what we do. We keep putting the numbers up. I mean, again, we were here back in February '25, and we stood on the stage in this very room, and we made the case for our company. And did people believe us? I don't know if they did or not. They gave us like a $6 billion or $7 billion valuation. Now they have a little more belief. So for the people that believed 1.5 years ago, it's turned out very, very well for them. Everybody has to make their own decisions. We've clearly made our decisions. I mean, I spend every day all day at the company, managing this company, and we're fully invested in it, and we have a lot of confidence in what the future is going to hold. And you're going to keep being transparent, and we're going to keep doing very -- what we think are very smart things. We're going to be open. We're going to keep coming to meeting. We had great meetings all day, where we hear feedback from the people that own our company. We factor that in, and we move forward and make the best decisions we can. And we think we have a tremendous franchise, and we're very focused on getting the most out of it as we can.

That's very great to hear. So I remember last -- since you brought it up, talking about things being done better. Last February, at the Analyst Day, I asked you a question, why did you move from WD to SanDisk. And you said because you saw much better innovation, a much more exciting road map at SanDisk. And I guess you proved it right. So -- but to that point, as you look at your whole TLC, SSDs and QLC SSDs and enterprise SSD side, you guys have lagged the market, but now you are catching up. What changed? What is the differentiation in your SSD road map today that is giving you that design win rate versus your peers that is not there last year?

So it really started more than last year. I mean, these are long design cycles. These are like 3 or 4 years, at least. Building an ASIC is not easy. And then getting qualified at a major customer can be a 2-year process. So this target years and years ago. And if I think of SanDisk and kind of just the big picture, there's all these different episodes of SanDisk, if you will. I mean it was just a tremendous consumer company, a tremendous IP, I think it went into Western Digital and inside of Western Digital, what you would have expected happened, it became a great client business. And it was like the rise of client. And why was that? I mean in my opinion, you were -- and I wasn't there. I didn't witness all of it, but I'm just looking at the history, NAND was replacing the hard drive and client. So you knew the customer, you knew the use case, you knew how to test it, you knew what all the features were, and you could build a tremendous portfolio there and a lot of innovations and systems work, the first DRAM-less client, all these innovations. And all of this expertise on building systems for consumer brought into building systems for clients. And then you had a new management team come in 6 years ago to have more of an enterprise background and started bringing an enterprise bent to this, like how do we get the right engineering teams, how do we get the right projects. And projects are very, very difficult. They go on for years, and it's the culmination of hundreds of decisions that happen every week that eventually lead you to having the right portfolio for the right market. And we've just run that play long enough now with an unbelievable internal team that has all of the expertise in how to build NAND controllers, applying all of that to the enterprise market. And when you do that for long enough, you end up with great products, and we're very, very happy with where we are.

Fantastic. Last one -- last question on the SSD side. And then we move on to something more exciting HBF. But first on the SSD side, as you look at KV Cache demand and some of the CMX racks that NVIDIA talks about, are you seeing that gain a lot of traction. You obviously have agentic AI and all that demand coming in as well, but that's supposedly going a different route. But from the KV Cache side itself, are you seeing a very big pull through from the CSPs?

Yes. I think undoubtedly, this is a very big backdrop in the industry. We're, as I said earlier, right, I think AI, there is just a huge focus in the early days, appropriately so on model training, right? You don't really have AI until you have a model. And so an enormous amount of focus on building out the systems for training models. And then I think we've all learned over the past several years that these models are extremely valuable, right, that they can -- you can do amazing things with them. I mean I'm an enterprise software guy for 30-plus years. The amount of change that's happening to the discipline of writing software is almost spectacular of what's happening with these models. You have pieces of software that people have worked on for decades and all of a sudden you can run a model against it and find defects that have been there for years and years. It's just -- as a professional, it's just kind of crazy the amount of productivity. So then once you get that, you have to scale it. You have to scale inference. And I think this is really -- there's been -- there's now a number of spectacular companies in the world that can scale technology globally in a way that's never happened before. Cloud-based computing is unbelievable. I mean it used to be -- to scale technology, you actually had to ship the technology to everything. I mean I'm old enough to understand this, right? You used to have to ship technology to every single person. Now you just point your device in a cloud that has the most sophisticated technology in the world, and you're up and running. So the friction of deploying technology in the world has just almost been completely removed. Except for the people that are doing that work, it's extraordinary work to build a global infrastructure to distribute technology. And I think those companies are very focused on how do I scale inference globally. And to do that, it's got to be -- you got to know the use case you're scaling for. It's got to be done in an economic way and you have to be able to do it in a predictable time frame. And I think as those conversations have been going -- people have been going through those conversations, they've been -- what is the right mix of compute, what's the right mix of DRAM, what's the right mix of HBM, what's the right mix of NAND? And that's a very dynamic question, that changes based on what use case you're assuming, what's the power of the model, how big is it? So this is a very dynamic equation that's kind of happening in real time. And I think the big picture answer to that question is, you need a lot more NAND because NAND is the most scalable semiconductor technology. And if you're going to scale something economically you're going to want to use as much of the most scalable technology that you possibly can. And I think that's led to this whole process of discovering what is the true value of this technology to the world.

It's the capacity versus cost question. How much capacity...

It's a capacity versus cost and -- yes, and NAND has been something where it's been almost always focused on density. And now it's focused more on high performance. It's a much more multidimensional equation.

Talking about high performance, I want to go to this HBF. There's obviously a lot of excitement around it, high-bandwidth flash, which is similar to HBM in the sense that you stack 16 layers of NAND, you have a logic-based CMOS bonded array at the bottom. Maybe you can talk to how you see that technology evolving because it starts to deliver the bandwidth of DRAM at a cost that's 1/10 of DRAM. But it has its challenges, whether it's [ right ] or endurance speed, endurance cetera. So maybe you can talk to where you see your road map. I know you guys have talked about HBF in the second half of '26 and might be a controller hardware in the first half of '27. But maybe you can give us some more color on that.

Yes. So I think the original -- the whole original idea from HBF came out of the conversation we just had, except it happened like years ago by a bunch of very, very smart people inside the company that we're basically -- if you're a designer, for your whole life, you've been told, give me more density, like more density, more density, more density. That was kind of the whole idea of designing new NAND nodes. But now you have this -- I think they saw this -- well, I don't think they know, they saw this like inference use case coming. We have these very large models. And so I need a lot of storage. But for NAND to be -- to work in that inference use case, you had to solve some other problems, like you had to make the -- the one thing about inference is a very deterministic equation, like you know what the model is. It's not like training where you're building model, you know what the model is. You just need -- you need to load it into a CPU as fast as you can. So the team started thinking about how do I re-architect NAND for higher bandwidth, higher right bandwidth. How do I get to some of the endurance questions, right? And an enormous amount of intellectual property was developed over the years. Obviously, we don't talk about that. That's our intellectual property that basically said, "Hey, at some point, the world is going to need more storage inference". And can we design a NAND dye to solve that problem. And that's what's kind of, I think, the genesis of HBF. We are working on it for quite a long time before we launched the company last year. We thought that was the right time to start talking about it because we were launching the company. And I think one of the most satisfying things that happened over the last 1.5 years since we launched the company as I think there was a lot of skepticism when we first started talking about it. And I think that there's a lot less now, right. I think it's like people can see, hey, like it inferences is a memory bound problem as much as [indiscernible] by GPU bound. So how can we bring more memory, more storage inference? And that's what HBF is all about. And so we're in the -- now the challenging side of it is, it's not just a plug compatible replacement for something in the current AI inference architecture. It's a system play. You have to change other parts of the system to get the whole thing to work. And that's the process we're in now is talking to customers about how they're going to design inference, whether it's on a device or whether it's in the cloud and how does our technology plug into that, and that's an iterative process, right? They're trying to figure out what the use case is, what they want to build. We're telling them what the capabilities are and then you're constantly changing both sides of that equation. In the meantime, we're off building the dye, right, which we expect to have later this year. And then we're building the controller that actually controls the dye and actually delivers the product, and we'll have that available sometime next year. And we're going through that process, and we're iterating on that process with our customers to get it to lock in to a specific use case, which then we can commercialize. And so we're in that process, right now.

Got it. So when you're thinking about working with the customers, is that you're getting a lot of interest on the CPSs, enterprise to have customers...

No. We haven't really [indiscernible] customers, and we don't want to talk about that, just so we think it plays across the device all the way to the cloud. I think that's what's really good about the technology. And so as models get bigger, as [indiscernible] links get bigger, as agentic coms mixture of experts models. All of these things are tailwinds for that kind of technology and becoming more relevant.

And this would be more of like an on-dye HBF similar to again like an HPM. Is that fair or that's not the same?

[indiscernible] be a NAND dye, it's like a BiCS8 product, it's a new dye type for that that's optimized for this use case.

Got it. The one other pushback we hear is the bandwidth seems to -- you seem to have solved the bandwidth problem. Might be bandwidth is 1.6 terabit or higher versus similar to an HBM, but it has latency. What's your thoughts around that? The latency is higher than typical NAND?

We need to get into a very technical conversation very quickly...

And that's the last question, we will go on to easier one.

You're in a deterministic read here. So it's the pipeline like you know that you're going to -- the latency is only at the beginning, once you get going, you're going. That's a very high level, very, very simplistic.

Okay. I won't be live with it.

We understand that issue, and that's part of the issue of why we work with them as a system level, right? you're not just going to plug it in for something else that was in the system. Everybody needs to change a little bit to get a much better answer.

Yes. What's technology without the challenge, right?

Well, I mean, that's -- I mean, that's where the value is. If anybody can do it, if you could just like wake up and build this stuff because it takes a long time and a lot of very smart people, and it's very, very valuable. That's why, that's why all these people here want to invest in our company before. We're doing stuff...

Maybe with the next 10x, I guess.

It is -- I just say we can -- it's not just there's an economic benefit for us. That's like the second order issue. The first order issue is it provides to our customers a very, very, very compelling value proposition. That allows them to build a more economic business or allows them to build -- to deliver a better service to their customers. That's the first step. And they have spectacular businesses. That's one you asked me why -- you said earlier about why did I come to the NAND business. I Mean, NAND business is incredible. It's like every device in the world uses NAND at some level. And our customers are just like a who's who of every single spectacular technology company in the world. Those are all the people we get to work with every single day and figure out how we can innovate so they can deliver an incredible product.

Yes. I think it's the satisfaction of solving problem and hoping the next...

That's what you do in business.

Absolutely. I want to quickly pivot to Luis here. I think Luis here eagerly waiting to answer the question. Might be on the NBM side. You obviously announced one -- a couple of engagements there. How do you see that pipeline growing on the NBM side? Maybe give us some broad strokes on how these are structured. I mean, obviously, supply is tight. Everybody is trying to get on the NBM sort of this LTA thing. But maybe you can tell us how do you see that mix of proportional revenues going out as you...

Probably when we closed the quarter, we said we had signed 5 deals, right, and we're very happy about that. And we talked about the financial profile of those deals. And as I just mentioned, it's a win-win relationship with our customers. They are great. They're coming back, and we're in constant negotiations with them. As I also mentioned last quarter, there are several conversations going on, and they are progressing well. And we're talking to customers across all segments, right, from data centers to edge customers, really the consumer -- our consumer business is more transactional so it doesn't really apply there. But we're talking to all other customers. And as long as we're willing to operate in this new business model, we're open for business.

Got it. I want to ask you a slightly tougher question. You have BICS10 coming, it's 300-layer plus, right? it's what's out there. Things are tight now, right? But at some point, you're adding -- you have to add CapEx. You'll add CapEx, industry adds CapEx. How do you decide on that CapEx road map, your partner Kioxia, you have other peers in the industry. How do you think through that dynamic of how you add capacity, how much do you add capacity?

Yes. I would start by saying we're adding capacity all the time. We're growing capacity in the mid- to high teens, right? That's what we said. So we're investing. You see that in our financials. We've also said that we made the most cost efficient transition of the BICS8 transition so far. Next year, we're going to be continuing that transition to BICS8, and then we're going to start building some capacity for BICS10. It will be a little bit more expensive per bit, right? So same growth but a little bit more cost -- higher cost, but it would still be lower percentage of revenue, obviously, as our prices have continued to go up. What we do is we take a long-term view of the market. We talk to our customers. We understand what that growth is and that's where we're planning to. We can't react to one quarter of what's happening today, right? It takes 110 days to produce a wafer. So we really take a long-term view, and we want to make sure that we're there to source profitable and sustainable bits.

Got it. One last question, I got a minute here. So your earnings for next year, consensus is close to $200 earnings per share. That's a massive cash flow. What are you going to do with all that cash?

Yes. I think it's just an easy question. We've been very consistent on that. Number one, we need to invest in the business. The CapEx that's required, we need to continue to fund the business. And we've made some strategic investments to strengthen our supply chain. We invested in NAND here to get more access to DRAM, which is very important to us, particularly as we expand into data center, which consumes more DRAM. We extended the JV agreement for another few years. So that's A great investments in the business. We said we wanted to pay down our debt. So we started with $2 billion. That's all gone. And the third thing to do, which is really the role of our company is to return cash to our shareholders. We announced a $6 billion share buyback program with earnings. As you can imagine, we should be executing on that. We will give you an update at the end of the quarter and then what's next.

Right. What a fantastic story. Any last questions? Any one last question. No? Done, gone, gone. Done. That's it. Fantastic. Appreciate it.

Hi. Good afternoon, everyone. I'm Mark Newman, Bernstein's U.S. IT hardware analyst. And great pleasure today to welcome back again, David Goeckeler, Chairman and CEO of SanDisk, who was also previously CEO of Western Digital, during the spin-off orchestrated the spinoff of SanDisk. Thank you very much David, for coming back again today.

It's wonderful to be here, Mark. Thank you for having us.

Thanks.

Can I get started with safe harbor -- it's like it has to be done.

Sure, go ahead.

I'm the only one that can do it appear I will be making forward-looking statements in today's discussion based on management's current assumptions and including with respect to our technology and product portfolio, our business plans and performance, our capital allocation priorities, market trends and opportunities and our future financial results.

These forward-looking statements are subject to risks and uncertainties. We assume no obligation to update these statements. Please refer to our annual report on Form 10-K, our quarterly reports on Form 10-Q and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations.

We will also be making references to non-GAAP financials and a reconciliation of our GAAP and non-GAAP results can be found in the Investor Relations section of our website.

Thanks very much. Okay. Now we've got that out of the way. I'll start with the I've got a bunch of questions I'm going to ask. And just remind everyone, you should have a pigeon hole link for free to put your own question in there. I've got an iPad up here. I'll look at the questions coming in. I'll try to ask a few audience questions. want to get through some of my own, if that's okay.

So I think I'd like to start off with demand. If we could talk a bit about demand and then we'll talk about some of the other items. But first of all, on demand, can you frame this demand environment you're seeing today, given what's going on with but also looking at other areas, mobile and consumer, like how are you seeing demand changing versus last time we talked about demand, particularly in AI.

I'll say first, like this is one of the reasons I really like this market, and I really like this franchise is there is a lot of demand drivers. I mean NAND is used in like every interesting technology there is in the world. Smart -- the traditional markets, smartphones, PCs, data center, which is obviously now growing significantly, but moving on to IoT devices, auto, robotics.

It just kind of goes on and on. It's a very, very diverse market with a lot of demand drivers. Those demand drivers move at different rates. And I think it really makes it a very, very fun place to build a franchise like we have. So what's going on right now? I don't think it's a mystery to anybody. Data center is really really, really growing aggressively.

We came in to -- if you go back maybe 3 forecast cycles, we were thinking data center this year would grow mid-20s. We upped that to mid-40s. We upped that to mid-60s. And now we've upped that even a little bit for that. If you look at like what data center is going to grow on an exabyte basis in calendar year '26.

So that's happening. It's obviously a big driver of the market, a lot of stuff that's happening. But the other markets are -- there's still a robust market across robust demand across all markets, whether it's PC, smartphones, we're still having great conversations with all those customers, across auto sectors, IoT sectors, I think it's just a very -- it's a very robust demand environment.

I guess, I mean, given how strong AI is, data center growing 60% plus, you have other parts of the market getting slightly crowded out though. That's part of what we're potentially seeing. How do you frame that?

Look, I mean it's a market, right? And I think markets always rationalize supply and demand. They're kind of always in balance. And clearly, there's ways that those clear through price -- and there's just an enormous amount of very, very attractive demand environment being created in this market. That's a very, very exciting thing.

That's always going to have impact on other parts of the market that maybe aren't as attractive from an economic perspective. And that's something that happens in any market at any given time. We happen to be -- it's a big market. It's a very liquid market. We know what the price is all the time.

In fact, it's a market that is kind of used to trading price constantly, even in the contracted part of the market has traditionally been set price every quarter. Right, which is really a lot of volatility. That's one of the things we're trying to move away from, quite frankly, I think this is the dynamic you're going to have in any market where there's always a significant amount of new attractive TAM being created.

And look, we are talking about a little bit about this on the way in. It wasn't that long ago that I was back here in New York, launching the company when we did the separation, and I got on stage and I said, we're going to invest for mid- to high-teens bit growth.

And we think that we had a view that this was early '25. We had a view by the end of '25, the market pricing was going to inflect higher. And the predominant view to the end of last summer was that was the wrong point of view.

And I was reading reports as late as, let's say, late summer that said, Oh, SanDisk is going to miss their numbers in December because pricing is going to be down. So it didn't quite work out that way. I don't think anybody could have anticipated the real intensity at which data center has come on.

But we believe that this has been -- this is a great market. We've been investing for growth in this market. We have to make investment decisions many, many, many years in advance of when the actual supply shows up. We're investing heavily billions of dollars in CapEx, hundreds of millions of dollars in R&D productivity, which, by the way, that's a whole -- there's a whole theme there about R&D productivity and NAND.

It it's very spectacular. We can grow a lot with additional productivity, which each node we deliver. But we've been very comfortable for a long time with this idea that we can grow the market, and we're committing to grow the market a mid- to high teens growth rate.

And just going back to demand specifically. Beyond the headline numbers, what are you looking at for leading indicators, such as order book depth customer forecast revisions, qualification activity? What kind of -- what things are you looking at to give you most confidence in the durability of this demand cycle right now?

Because clearly, demand is far exceeding supply given where pricing is going.

Yes. And what we're looking at is what is that environment going to be for the next I mean again, we're we just -- for example, we just invested $1 billion a number of months back to get -- extend our agreement with Kioxia which is a fantastic agreement from 2030 to 2034. So we're obviously looking very far in the future on what demand is going to be.

So there's many, many different ways we go about that to answer your question. So first of all, we do a lot of bottoms-up work -- so we talk to our customers. We know what they're built. And for example, smartphones, PCs, all these kinds of markets. We have deep relationships with our customers. We know what devices they want to launch in the future.

We have a view of what the mix is going to be. We have all kinds of bottoms-up analysis on the big markets of what kind of bit growth that's going to drive. We're obviously looking at CapEx spending. I mean that's what's driving the data center number up, every earnings cycle. The CapEx number goes up.

We know we have a decent idea of how that relates growth in our part of the technology world. So those are kind of long-range things we're looking at on the demand side. And then we're in the market every single day. I mean we are having conversations with our customers. They're calling us and talking about what they need for the -- currently and in the future, we're obviously having discussions about pricing continuously.

And so it's a culmination of all those things that give us insights into where the market is going to.

And specifically within AI, we have these different stages of AI training, early chatbot influence, more advanced influence and now we're moving into this Agentic era. How do you see those impacting NAND demand over time going as we go from the earlier stage to the more later stages of AI.

So we've always believed that inference is really going to be where it's at on NAND. And so we had to get there. And I would say for the first 2 or 3 years of AI, I would constantly get these questions is development and deployment of AI going to impact the NAND business? And it was always, yes, it's going to, but we got to get there, right?

You got to get the models built, you got to get them deployed, you got to get them rolled out, you got to get users using the technology. There's got to be valuable use cases that drive consumption I think we're past all that. Now we're rapidly moving through all of that. And you're starting to see the impact over the last year.

This has really been the story. You're starting to see the impact of NAND on the inference architectures. And I think as our customers start to build out these architectures and you're trying to figure out how do I scale inference globally right?

Training, you don't really have to scale globally. You're training a lot of very smart people, a lot of infrastructure driving training. But inference you want to drive to the masses, if you will. Billions of people are going to be using inference in some way.

And so when you're going to go through that process and you're going to scale something like that on a global basis, it's got to be economic. Right? Early in any kind of technology, you're naturally going to want to -- when you're a technologist and you're doing things for the first time or you're building markets, you're kind of you're kind of overwhelming your architecture with all the resources you possibly could need.

Give me all the compute, give me all the memory, give me all the power, give me all the networking, give me everything I need and then I'm going to build a system but then as you go to scale that system, you need to really kind of drill in what exactly am I going to scale, and that's got to be economic.

Because if it's not economic, it's going to be -- obviously, it's going to be too expensive. That means you're going to have to charge more for it. It means you're going to open yourself up to somebody else coming in and doing it more economic and putting you out of business. So these are like really, really big, very, very hard decisions.

And I think what's been happening over the last year is the people that are responsible for doing this, like spectacular technology companies that have an enormous amount of expertise of scaling technology on a global basis. I think this is really the story of the last 20 years.

I mean the distribution of technology has become almost completely frictionless, right? You just point your device to a URL and you have the most spectacular technology in the world. It didn't used to be like that 10 or 20 years ago, we had to ship us something or you had to upgrade your software, like there was all this friction in the system.

All that friction has been removed, which means we can deploy technology at scale very rapidly which is spectacular, right? And we're witnessing that happen right now. But the people that do that have a very, very difficult job because it's very expensive, and you need to do it in the most economic way.

So those people have been going through that process of how do I build that architecture. And that's where NAND is becoming more and more into the picture. Why? NAND is very scalable. It's the most scalable semiconductor technology in the world. We can produce the supply, right? And so as models get bigger, as context lengths get bigger.

All these kinds of things are driving you to you have to use more scalable technology if you're going to do this in an economic way? Or if you're just even going to do it, there's just not enough of other things in the world, right? DRAM is spectacular technology. HBM is spectacular technology. It has unbelievable characteristics.

It doesn't have the scale to solve a global inference issue. So I think companies have been figuring out what is this architecture and starting to scale it and how we're going to scale it. That's what's been driving this kind of behind the scenes more demand for NAND, more demand for NAND.

I find as I work on that architecture, and I dial in exactly what I'm building to, then I need more or less manned, and you're coming out with the answer, we need more. And so that's what's driving the market, and that's what's driving those customers to come to us and say, hey, look, we're doing planning for years into the future, that's our business.

We want to understand your plan for supplying us this critical technology years into the future. We don't want to just show up every quarter and try and negotiate the price and figure out if there's enough. We need to know now. Can you supply me in 28? Can you supply me in '29? And this is what's leading to kind of this whole transformation we're going through.

I appreciate that. That's phenomenal demand we're seeing right now. [ Jenson Wang ] earlier this year at CES laid out this KV cash vision. It's something like an incremental 17 terabytes per GPU. Are you seeing that? Is this in your demand numbers? Do you think that's going to have a big impact in an incremental additional impact for NAND demand?

Yes. I mean this is exactly the process I was just talking about, where people are designing systems and they're configuring systems. The KV cash is moving into NAND because it's got to scale. I need scalable storage technology, that's NAND. And so depending on what use case you're building for, like I know people want a real clear, hey, if I do so much of this, I get so much of that. But it's not that simple.

Like you need to figure out what use case you're building for in the future and what you're going to scale to, once you know that, then you can design an architecture to do it. And when you go through that process, there's like a whole bunch of variables in there.

How big is the model you're using, how many tokens, what's the KV cash size? Or you have a cash somewhere. What's the hit rate on that cash? You go through this very complicated equation, and we've done some work on this that we've shared. And out of the bottom comes how much NAND you're going and then you kind of come to us or you come to some of our peers, and you say, how do I go acquire this much NAND over the next -- so we believe very much in that vision.

And I say it's much more than a vision. It's what's happening in reality, and it's been happening in reality for quite some time now as companies need to take this brilliant AI technology, and they need to scale it so we can all can use that?

Yes. I think -- I mean you're saying it's about the density, right? The density of the NAND flag versus DRAM and it's just in terms of how many gigabytes you can get per dollar and how many gigabytes you get per square area, just as much more. It's just -- we can deliver more -- we can just deliver more capacity, yes, more density.

That's exactly right. It's a different technology. It solves a different use case. It's not a substitute and it's not -- doesn't mean one's good, one's bad. That's not the issue at all. You need both.

And you're going to have to use this very scalable storage technology as part of that architecture, and that's why these data center numbers keep going up is because as people iterate through this process, of how to figure out what that architecture is I need to scale, the number keeps going up, and that drives the demand that drives the demand higher.

So just pivoting a bit to pricing, ASPs, not LTAs, but just looking at the pricing environment in the industry. For those that haven't been following SanDisk closely, the ASP last quarter per gigabyte went up approximately 140% Q-on-Q for SanDisk.

But that's my estimates. I don't think you've actually given that exact number, but it's pretty much around that, which is just absolutely phenomenal. My question is how do you characterize the pricing environment right now? I mean, clearly, you can't get 100% Q-on-Q continue, which is not sustainable.

But amongst the different segments, are you seeing strength still continuing, pricing still trend up? Or how do you see it?

Look, I mean, we have a forecast for what we forecast. I'm not going to get in to talk about what future pricing is. Look, I mean, we build -- the most important thing in our business is to build very valuable technology. It starts with the technology. It's always about the technology.

And if you find that you build great products that solve real needs, then we're on this journey of figuring out what the value of that technology is. And that's our job is to do that, and we'll continue to do that.

Got it. Okay. So price too strong. Okay. Got it. Historically, NAND pricing has been quite cyclical, and we're going to talk a bit about LTAs in a minute, what gives you confidence that this is going to be sustainable at this level of pricing right now besides the LTOs because we're going to get into the LGs next.

I mean I think this is really part of -- I mean, this is very much how I think about my job is to make this sustainable. I think the cyclicality has just incredibly -- I mean where I've used a number of times, it's just incredibly corrosive. It's we're either in a situation -- it seems like we're either in a situation where on the supply side, we're like scrambling to survive.

I was in that position in '23 a year ago, we launched the company and everybody gave us a valuation that was -- I thought was incredibly low, turned out that turned out to be true. And we're in a situation where we're having the previous conversation, you just asked me where everybody doesn't get what they want.

And to me, that's a thing where our incentives are not aligned. Our business models are not aligned. And so I think what I'm trying to do and what our team is trying to do is I think if there's kind of 3 big things we need to do is in this technology franchise and really any technology franchise.

And what I'm constantly trying to balance. And number one is always get a fair return for your -- what you built, right? We're very proud of our technology. Very difficult to do. Not only do we invest in all the IP to build NAND. We invest in all the IP to build systems -- we don't have one R&D team. We have 2 R&D teams. We have a team that builds the NAND, and we have the teams that build the SSDs and all the products, manufacturing, Oh, we do that too.

We have to invest all the CapEx to do the manufacturing. Oh, back end, yes, we do that, too. We have a captive back in. So we do everything. The whole process. Obviously, we have a lot of brilliant suppliers that provide a lot of important technology for us to be able to do that. But the #1 thing is get a fair return for that investment that we've made. And we've been making that investment for a very, very long time.

And so that's the first thing. And I would say we're doing okay on that now. For a long time, we didn't do very well on that. quite frankly. I mean, again, you only have to go back a year ago and people were basically telling us you're not doing a very good job on that because we don't want to invest in your company.

And so that's the first thing you have to do -- the second thing that we -- that I'm really focused on is we need to do something about the cyclicality, right? It's just corrosive because it's either -- everybody is just waiting for when the downturn is going to come. You have a good quarter. Oh, you're just one quarter closer to a bad quarter, kind of a crazy psychology. Either people aren't getting what they need or they have too much -- it just is not helpful from my perspective at all.

And so we want to do that, and we're doing that through business practices. That's why we call these things new business models. How can we change the way we the way we engage with our customers. And then the third thing you need to do in any technology franchise, you need to grow, right?

You got the right economics. You get the cyclicality out of it or you deal with the cyclicality differently and then you have to grow. And in every technology franchise I've managed in my career, the third one is the hard one. It's hard to grow, right? I mean it's -- especially large profitable businesses are hard to grow.

But that's one we have taken care of, right? We say we're going to grow mid- to high teens and people say, well, can't you grow faster? I'm like, let me get the first 2 taken care of, and then we'll start talking about that. And so balancing this equation is extremely difficult.

And if you start change -- you can always talk about 1 of the but you have to talk about all 3 of them together because if you start messing with 1 of them, a different 1 goes in an opposite direction. So it's kind of that whole equation is what we're constantly trying to balance -- and we're focused on all 3.

And I would argue that the more difficult 1 to solve the growth one, it's a huge advantage for us, right? People want to debate, should you be growing faster? Like, well, maybe we could be growing faster, but at the expense of the economics, that's not a very good trade-off from a valuation perspective.

Should we get more economic, et more of this and live with more cyclicality, that doesn't seem like a very good trade-off. So you have to do all 3. And we're constantly -- and that's the way we think about it. At least that's the way I think about it. And that's what we're trying to balance. And that's a lot of fun.

I think we're seeing very significant change in the franchise in this environment to really get after those first 2 issues.

That's really helpful. And then just drilling down on the second point, the long-term agreements, what you call new business models. Can you just talk about what you can today for how these agreements look in terms of durability, in terms of volume commitments in terms of pricing structure.

If you could explain what you can like how you think about those agreements, where you are today.

And I know you've said on the last call, over 1/3 of volume in these long-term agreements on new business models, as you call them. Where do you expect that to get to would also be helpful.

Yes. So let's talk about -- you said a little bit earlier, and I think everybody understands it's been traditionally a very volatile business, right? I mean, literally, pricing changes every quarter. That's a hard business to plan, hard business to forecast.

And traditionally, agreements, there's been -- and by the way, we have spectacular customers. I mean we're -- this is one of the things, again, that's so attractive about this franchise. I mean [indiscernible] our customers are the most enviable companies in the world. I mean, they're just -- they do spectacular work whether it's PC, smartphone, data center, whatever it happens to be across the board, it's just incredible what our customers do.

But the traditional view of a long-term agreement was I'll commit volume, then we'll discuss price later. Like, okay, well, that's better than nothing, right? So at least we understand if we agree on price, we understand how much volume we're going to allocate to everybody.

But we want to get out of this volatility. And so how do we think about this differently. So that's why we -- this idea of long -- there's a lot of terms that have been thrown around in the industry, long-term agreement, NCRs, take-or-pay, there's like all these different things.

And when you bring up 1 of those terms, in my experience in the last 2 years, as soon as you bring up one of those terms, the person across the table from you starts telling you all the reasons they won't work. And so we studied that very deeply.

And we said, look, what we want to do is we want our -- we want to get our business model aligned with our customers' business model. right? And more and more, we have customers coming to us saying, especially as we got through these data center qualifications.

So you think about the data center business we've been developing, we build an enterprise SSD. That takes like years to do -- it's a very arduous process. Then you start engaging with a customer and that can take 2 years, understanding what you're building, giving them samples, putting thousands of units in a lab, letting them run for a year to qualify, this is a very, very difficult process.

And so at the end of that process, you get to the point where the customer is, okay, like you've built a great product, right? We've invested a lot in this. We built a great product. I want to buy it. Great, right? And I want to buy it for a long time. I want to buy it for the next 5 years because my -- go back to the first thing where we started. I'm doing all this work, I'm building this new technology.

I have a lot of demand for your product. I don't just want to buy something this quarter. I want you to tell me that you can supply me for the next 5 years. And so, okay, well, show me what your demand is. and then we start the conversation. And this is kind of new, right? Because usually, it's like I'm going to tell you what my demand is for the next 12 months, and we'll talk about price 4 times a year.

Now it's like No, no, no, no. I don't want to talk about just the 12 months. I need to know 2, 3 years from now, 4 years from now, can I get from you what I need? Because when I'm building, what they're building is spectacular, again, incredible technology. So then we get into a conversation, which is how do we align our business models, right? You want to consume NAND, I want to produce NAND.

Now the way I produce NAND happens to be a business model that is probably quite different than yours, right? I have to invest 10 years ahead of time. I have to build this huge fab. -- like you see the thing from space. It's like enormous. And I have to plan years in advance for my capacity.

And the good news is I've done all that. We have the fabs. We have the R&D. We know what our technology is roadmaps going to be for years in the future. But now I've done all that, and I turned the fab on, right. Now I'm investing for growth, right? I'm going to grow mid- to high teens. So now my fab is running and there's more wafers tomorrow than they were yesterday. And that's true every single day. And every day, the wafers come out of the fab.

And I've got to sell them. Can't put them in inventory, you can't let them fall on the floor, somebody's got to take them. That's like kind of an unnatural business model for a typical consumer. They have big businesses that are growing, too, but guys, do I have to buy something every single month, right? Do I have to buy more than last month? And the answer is yes.

So how do we align -- you're going to need this supply? I'm going to produce that supply -- how do we align our business model so that I have confidence that you're going to -- you're going to be a strategic partner of mine, and you have confidence I'm going to deliver to you.

And how do we put a contract around that? And that's where we came up with these new business models. So how did we think about that? Number one, we need partners that are going to consume a significant amount of product, right? Because this is going to be a big contractual arrangement. Number two, we need you to grow your demand as fast or faster, hopefully, faster than our supply.

So if I'm investing for mid- to high teens growth bit growth rate you come to a it the same amount for 4 years in a row, that really doesn't help me, right? You need to consume faster than supplying than your big strategic partner to me.

Now the next thing you need to do is you need to consume predictably. Remember, the fab runs every single day. The wafers are coming out. If you're my strategic partner, you need to consume every week, every -- well, let's say, every month, every quarter, you've got to be predictable in your demand.

And the more insight you can give me to what that demand is, the better off, what's your mix going to be how much of this product, how much of that product, we got to get all that figured out. And then we have to put an incentive structure in place because, look, you're a public company, I'm a public company. Something may happen where you have to exit this contract.

I understand that, right? Stuff happens. The Black Swan event happen the whole economy goes up and down. Let's say we have a global pandemic. Let's just imagine an event that may impact the whole world. So at that moment, I need an incentive structure where you're incented to stay into the contract.

And if you don't stay in the contract, that I get a benefit, all right? So I'm going to ask you to put an amount of money aside upfront. And we're going to let a third-party hold that for us, right? We're not going to -- I'm not going -- we're not going to argue. I'm not going to sue you. That's never going to happen, right?

That's you don't see your customers, we're partners, right? Something happens, you have to exit the contract. So ahead of time, let's have a third party hold an amount of money that you have -- you can use that word if you want. But some third party is going to hold an amount of money. But the easiest thing was you just give me all the money upfront.

That's kind of impractical, right? I mean this is a 5-year relationship, you're not -- that's a big check for anybody. That's not realistic from all -- for all kinds of reasons. So we had to come up with something different. So let's have a third party hold that money and they'll have the contract, and they'll be able to say, you walked away from the contract or you didn't walk away from the contract. And if that happens, that third party is going to release the money to me. It shows up on my balance sheet, and we part friends.

At that moment, the contract is over. I keep everything from that point on, you keep everything you paid for and we all go about life, right? And so we think that aligns our incentive system. You're now incented to stay into the contract. You may think, oh my gosh, I need to exit this contract. Do you really want to exit this contract.

There's going to be some amount of money, you're going to have to forgo billions. And so you better be sure. And if you do, if that has to happen, then I get a bit of a soft landing. I get a bunch -- I get some cash, which helps if it's a black swan event or something, it's let's say, it's -- I want to get rid of the 6 guy, let's say it's a huge down cycle, what do you need to down cycle, you need cash.

Then we've insulated ourselves and we're both -- we both move on down the road, and we're both fine and we can do business again at some point in the future. So that's a rough idea of the contract structure we've put in place is that the $12 billion financial commitment.

Yes. So let's decompose the numbers. So we talked about this. We have RPOs now. that's something you would think about. I've run a lot of software businesses, right? So that's a metric from there. But that's really an accounting metric, right? We didn't like wake up and say, oh, we need to use this metric, like that's what the industry does when you have contracts and they have future obligation.

So we signed 3 contracts for the end of the quarter. So the number we had, the $40 billion, whatever billion, $42 billion number was the remaining purchase obligation is the minimum amount of purchasing obligations on those 3 contracts for the life of those contracts.

And then there was another number we talked about, which was a little bit of -- it was a little different, so it was a little complicated. We signed 2 contracts after the end of the quarter, so they're not in our numbers. But of the 5 contracts we signed, that amount of money that's been set aside in case people walk away was in aggregate $11 billion -- so those are how the 2 numbers kind of all tied together.

Got it. Got it. But honestly, we don't ever expect to see that money. I don't ever want to see that money. I think we have great partners. I think these contracts are going to run to the end I think that our interests are aligned and everything is going to be great.

But we live in the real world. There's got to be some incentive system, and I think we've aligned those incentives. And I think our we have willing partners that are willing to go -- that are -- that want to go down with that path with us because they value the commitment of supply.

And customers, those all hyperscalers or No, we haven't said that. We're not going to say that. What we want it was very smooth. I mean, look, we want a diversity of customers, right?

We want to is the same thing we do about -- I've talked a lot about portfolio.

We want a diverse portfolio with a lot of optionality across our products we sell. Where I started this. Why do I love the NAND mark? Lots of reasons I love the NAND market, a very diverse market. A lot of great customers, a lot of places you can sell your product. But you have to have technology to do that. You don't just sell them raw wafers, you got to build products. If you're selling into the consumer market, you got to have a team of people building that.

You got to have a back end that's creating all that stuff. So you want a diverse the portfolio as possible, and that gives you the most optionality possible. The same thing is true for these new business models. We want a variety of term links, right?

You don't want them all to end on the same day. So you want some that are a year, some that are 3 years, some that are 5 years, and then you want a diversity of customers that ideally will cover as much of your portfolio as possible because that's what keeps the portfolio alive and keep that optionality going so we've made the first step right?

And that's what we announced on our earnings call. Again, go back to the 3 things I talked about, get a fair return, deal with the cyclicality and grow. Fair return. I think we're okay. We can do better, but we're pretty good. We've got -- now got 5 in the middle column of starting to address that. More than 1/3 of the portfolio visibility instead of visibility being 3 months at a time or maybe 12 months at a time.

Now we're talking about visibility 3 -- 2 years, 3 years, 5 years, wildly different. And the growth piece, remember, the grow was always there. That box has always checked. -- right? That's always the one that's like that's what's so great about this market. It's going to grow.

So we got the growth box checked. We got the first box. We're in a good spot. Now we need to keep it. That's the second part.

And that's why we call new business models because it is a different business model of how to do this.

That 33% do you hope that to get to 50% to 70% or? Well, that's unrealistic.

No, that's not unrealized. It's TBD, right? Again, this is -- we're not done yet. We took the first step we took this -- maybe we took the first 5 steps, maybe that's the way to say it. But we're still having more discussions, and it depends on this portfolio thing I said earlier.

Look, there's plenty of -- like I said, we have spectacular customers. just spectacular customers. They're great companies, great people. They build unbelievable technology. Some of them like the business model we have before. They like the quarterly, Hey, we're just -- let's just do a core. Great. Fine. We're good with that.

We're absolutely fine with that. We know how to do that. If that's what they want to do, we're all in for that. So we'll see how --

Are those customers are going to get enough supply, though, if they don't sign up for. I can't run their business. They have to run their business, right?

I'm not the only supplier in the market. But what I want to do is get a portfolio of these agreements that give me the diversity across -- give me the diversity I talked about can cover a fairly wide swath of my portfolio. I don't have to be all of it. It will never be 100% because there's a whole bunch of customers out there that just aren't big enough that are great customers and great business to engage in.

So we'll see -- I think it's a bit TBD, maybe a bit unsatisfying for you right now. But it's a little bit TBD what the final landing point is. But I think if we have the opportunity, we will continue to drive it higher. That's correct. I mean that's a lot of clarity, much more clarity than we've got from your competitors so far in long-term agreements. So I really appreciate that.

I love my -- those are all great companies. They really are. Just pivoting to supply and capacity, just given how strong pricing is and how strong demand is. A lot of your competitors, Samsung, Hynix, Micron, these companies, they don't have space to add capacity because they've given all their fab space to DRAM, which has also been tight.

But Sandisk Kioxia together, you're one of the only ones that actually has some space that you could added capacity. I'm not encouraging you to do that. I'm just asking. I was just asking.

You're asking for a friend.

Like how do you think about that I'm asking how do you think about that considering that you have the potential to add capacity pretty much no 1 else is, except for possibly YMTC in China. Could you add more capacity? Or are you just really trying to optimize pricing at the moment.

Okay. So probably a more complicated and than you think. So first of all, we're always adding capacity. I think that's where we need to start. We are always adding -- remember, we're growing mid- to high teens. This is a very big market, growing volume mid- to high teens.

That's amazing, first of all, right? Number two. So it's normal course to be adding capacity. Just is. That's what we -- that's the business we decided to enter. Number two, we have to make decisions far in advance. Like what demand is next quarter has no impact on my capacity decision.

I had to make that decision 3 years ago. We have a fab plan that's years into the future. It's very complicated to move tools around what nodal -- you're not just running one node in the fab, you're running many, many, many different nodes at the same time, you're transitioning really sophisticated.

So you've got to make those decisions far in advance and also, again, I don't want to get harp on this too much, but reflect back, it was only like 12 months ago when everybody told me that we were investing too much. right? And we were saying, No, no, no. We think mid- to high teens growth rate is the right number and people are saying, well, price that's not the right number.

Less than 12 months.

Yes, less than 12 months ago, right? So we can't whipsaw that much. So how do I think about that? What I think about is we're investing from mid- to high teens growth rate. And we have a great partner in Kioxia. It's a great relationship.

It's gone on for a very long time, for a very good reason because very productive and it's very valuable. And we're good at planning. We're good at planning and making sure that we have what we need at the right time to continue to grow the business.

Now one thing I will say that's very important to understand, very important to understand. We can grow through nodal transition. So what I'll call, R&D productivity. The number of bits per wafer continues to go up faster at a compounded rate faster than the mid- to 1-teens rate I'm talking about.

So if we just -- if we just went from node to node with the same number of wafers, we would oversupply the market. So we are constantly adjusting this equation. Now remember, each node requires more clean room space. Each node, more complicated, more steps, more tools, more clean room space.

But this dynamic is extremely important, right, that we are -- if you look at our CapEx as a percent of revenue, it continues to go down as revenue goes up. because we still have all this R&D productivity. So for all of you that are here that are investors, this is very, very important.

Like what it says is I can get the growth without an enormous amount of incremental CapEx and when you get -- again, go back to the model I had, right economics, get rid of cyclicality and grow what you find at the end of the day, what are we in business to do, we're in the business to generate free cash flow.

And what you'll find is this franchise is very good at that, because we're very efficient with the CapEx we spend, we're very efficient of getting incremental output from that.

So I've got a few questions from the audience. Just one more from me, if I can, before I go to the audience questions. HBF, high-bandwidth flash. Any updates on that, that you can touch on?

We've been very excited about this technology from -- again, when we announced it when we launched the company in February of last year. We've believed for a very long time, that once we got to inference, that NAND was going to be a very important technology.

Like you don't need to convince us that the memory architecture needs to change for inference to scale. That's essentially a little bit around what HBF is about. It doesn't mean HBF is going to take over her enterprise SSD. It doesn't mean that HBF is going to be a substitute for DRAM, any of those kinds of things.

What it says is there's an enormous opportunity for innovation as inference scales. And people that have new ideas, like when I see AI right now and the amount of huge scaling going on, I see a giant green light for innovation. If you got new ideas, bring them, right?

Because the world is trying to figure out how to scale this spectacular technology. And the faster we can do it, what I said earlier, now we can scale technology in a completely frictionless way.

It's amazing how fast technology can be made available to everyone if you get the economics right. And so HBF is a strategy for how we can deliver a lot of density to inference, which is predominantly a read-based activity and a deterministic read-based activity. So we're very excited about the technology. It's new. We're building the NAND die now.

We expect to have that by the end of the year. Sometime next year, we'll have the system. We're building the controller on top of it. We've got a lot of work to do. We're working with customers on how they would integrate that into their architecture, right? Because it's not plug and play.

This is not -- we take our component and plug it in, you take something else out. You got to -- it's got a -- it's a system play. So you got to get your customers to adopt it into what they're building, and we're going through that process. So we'll continue to update as we go.

Okay. That sounds great. So questions from the audience. And first one, will the shift to edge or on device compute, be a benefit or wish to send is growth projections?

No, I think -- I mean, anywhere that you're shifting -- you say the AI, basically, AI and edge meaning in your device, in your smartphone in your this is just more of the same theme that NAND is everywhere. And as you start to scale technology, you're going to bring -- you need more capacity, and we have the scalable technology.

And so we see this as a very -- this is -- this is why we're committed to that mid- to high-teens growth rate than we were even last year is because there's like this evergreen nature to this market.

Like the world is just constantly innovating and thinking of new ways to use our technology, and that's a wonderful thing. I've got a great question here from the audience that says, can I ask all just to put your hands up if you own [indiscernible] could I ask you to put your hands up if you own SanDisk stock?

Okay. All right. Thank you.

We're working very hard on your behalf.

Next question here from the audience, can you expand on lessons learned from prior boom bust, how have management incentives evolved across the industry, similar to oil and gas or not to avoid repeating history.

I don't know about oil and gas. But I mean, I've learned -- look, I've been in the technology business now for probably longer than I should admit, like 40 years, building global technology. I started at [ Bell Labs ] a long time ago. And I managed a lot of different technology franchises, a lot of different technology franchises, hardware, software, SaaS, at very large scale.

I was somewhat surprised when I really took -- came in as CEO of this industry about just the way it works. It's just kind of really this whole idea of, like you said, boom bust I've said it many times, I just think it's corrosive like there's somebody that always feels like they're not getting what they want, either the suppliers are like doing what I was doing in '23 where I'm scrambling to survive. Or we're in a situation where people are saying, I can't get everything I need. And it's -- I think it's because of the way we go about this. Now there's some reasons for that, I said earlier, right?

We have to make a long investment cycle we increased supply as more of a step function, demand is more of a curve. So getting these things aligned is not easy. I don't think it's something we should just give up on I don't think it's something that is like it's our fate or something like that.

And I think that just because it's been this way for a long time, it doesn't mean it needs to be that way in the future. I mean maybe I'm arrogant, but that's what we do. We're innovators. That's what -- I mean we invent new things, and that can apply to business models as well.

That's essentially what we get paid to do, and I think the world is very good at it. And I think if we think about this business model, yes, what do we learn from the boom and bust -- let's not do that again. that would be -- it would be really good if we don't do that again.

And so how do we not do that again? The bust part of it. And how do we get to a point where we get a fair return for our technology that we build. And again, it's very difficult -- it's very difficult technology. It's not easy. It's 3D semiconductor technology, people dedicate their lives to doing this.

It's very expensive to do. It requires an enormous amount of CapEx fabs are very difficult things to build and run. Let's get a fair return for that, and let's put a business model in place where we can smooth this out. And I think that's very possible. And I think we've made a couple of steps down that path and we're going to keep going.

Great. Well, with that, we are out of time. Thanks very much, David. Thank you.

Thank you, everyone.

All right. Good morning, and welcome again to JPMorgan's 54th Annual Technology, Media and Communications Conference. My name is Harlan Sur. I'm the semiconductor and semiconductor capital equipment analyst for the firm. Very pleased to have Dave Goeckeler, Chief Executive Officer and Chairman; Luis Visoso, Executive Vice President and Chief Financial Officer of Sandisk here with us this morning.

As we're all witnessing, storage technology is a critical part of unlocking the full potential of AI, right, especially as the frontier model builders are now focused on monetization, therefore, significant unlock of inferencing-based workloads, which are definitely more memory and storage intensive, right? But in the edge and client devices and consumer applications, more and more intelligence, real-time responsiveness, also translates into more storage demand as well. Sandisk is at the forefront of delivering these high-performance flash-based storage solutions to all of these markets.

And so Dave is going to kick us off with some opening comments, a quick review of the March quarter results, June quarter guidance, and then we'll kick off the Q&A. But gentlemen, thank you for joining us this morning. And Dave, let me turn it over to you.

All right, Harlan, thank you. It's -- we're very happy to be here. I think Luis is going to get us kicked off with a few statements, and we'll go from there.

Thank you for being here. We will be making forward-looking statements in today's discussions based on management's current assumptions and expectations, including with respect to our technology and product portfolio, our business plans and performance our capital allocation priorities, market trends and opportunities and our future financial results.

These forward-looking statements are subject to risks and uncertainties. We assume no obligation to update these statements, please refer to our annual report on Form 10-K, our quarterly reports on Form 10-Q and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations.

We will also be making references to non-GAAP financials and a reconciliation of GAAP to non-GAAP results can be found on the Investor Relations section of our website.

Thanks, Luis. Harlan, again, it's great to be here. Just a few opening comments, look, it's a very dynamic time in the business.

Absolutely.

A lot of fun stuff going on. Look, we're a technology business. I think it always starts with the technology. The portfolio is just in great shape. BICS8 is an unbelievable technology node. We've talked a lot about that in the past.

I think everybody knows we've been very focused on our data center business. We have a great consumer business, great edge business. We've been focused on really driving product strategy and data center. I think this quarter, we saw over 200% sequential growth in our data center business, 25% of our mix.

So I think we're getting -- we still expect that to grow going forward, but I think we're getting a really, really solid balanced portfolio across the board at this point. Obviously, the demand situation is very dynamic. And I think one of the most interesting thing that's going on in the business right now is it gives us the opportunity to change the business model. I think this is a business that's been a very transactional business in the past.

That's led to a lot of dynamics, a volatile business. But we're at a point now where we can start to change those practices with our customers. I think everybody is a winner in that process, and it's a lot of fun to go through it. We made some very big strides in the last quarter that I'm sure we'll talk about here today.

Yes, absolutely. And that's a great segue into my first question because you talked about all of the transformational dynamics that are happening in the flash memory segment of the market. But it reminds me interestingly enough of another segment of the storage market.

And obviously, Dave, you were a part of that as well, right? And I'm going to rewind back a couple of years, because it was under your leadership at Western Digital, powerhouse and hard disk drive-based storage, flash-based storage, which is now the Sandisk franchise as we know it today, right?

But the dynamics that the team was seeing in its HDD franchise, '24, '25 time period is exactly what Sandisk has been seeing over the past 12 months, right? Back then, you were seeing an HDD market that was becoming more dominated by data center and AI compute customers that appreciated the technology and performance leadership in nearline drives and a strong exabyte demand profile, right, 20%, 30% per year type CAGR.

That was accelerating within a structurally constrained supply environment, right? Under your tenure, I mean, you drove the build-to-order model with your HDD customers. You then drove the initial multi-quarter customer agreements, which then more formalized under multiyear agreements, which looks very similar to the agreements you're locking in with your customers today with your flash-based customers.

Your customers get multiyear predictability and assurance of supply and return. Sandisk gets predictability on the demand curve and financials. But is much of what Sandisk and the team is architecting today via your new business model, NBM, multiyear agreements. Was that a page taken out of the HDD playbook, which has proven to be so successful in the HDD industry today and was kind of spearheaded by you and the team back a couple of years ago.

I think we're taking some pages out of a lot of playbooks, right? And I think it's really about really understanding your business, what the value drivers are, how your customers use your technology and really getting business models aligned where it's best for sustained investments, sustained value delivery and customers get -- they get the sustained benefit of your products on an ongoing basis.

And so the markets are similar in the fact that really being focused on how you supply those markets and being careful about really understanding supply dynamics and don't oversupply the market is going to lead to better economics and better kind of consistent business. But the agreement -- the NAND business is a little bit different than the HDD business. HDD business is a great business. I like those years I think the changes we made, I think the companies are doing great.

Absolutely.

I think it's there very well-managed companies with tremendous technology and great futures.

It starts with that like the technology is extraordinarily important, right? I think we build a technology that is extraordinarily important for the world. I think the NAND industry, that is a very broad set of customers across devices, across consumer, across the cloud -- these are all -- like there's a lot of diversity in our customer base, a lot of demand drivers, I think that makes it a really great business.

I think the technology is spectacular. I think in the NAND business, we have this huge advantage in that our technology is extraordinarily scalable. We can deliver enormous amounts of incremental supply through R&D and productivity gains. It's not just adding more capacity all the time.

And so what this is really about is one, getting a fair return for that technology and that investment. And again, it's a big investment. We don't just invest in the IP side of it of building the NAND nodes. We also invest in the CapEx side of it of having the production. We really have everything from defining the NAND itself, the fundamental node to manufacturing that, to then have the systems teams, that are building the SSDs, having the back end, all of it. Like as a franchise, we own all of that and we've built that over the last 20, 25 years.

And so -- in the current environment, it's about -- we're always thinking about how do we align our business model with our customers' business model. And how do we get those aligned, because that's going to lead to better economics for both of us. In the NAND industry, what I'd argue that these models have not been very well aligned.

Our business model, Luis and I constantly are making decisions on investing on like a 10-year time horizon. When you build a fab, you need to get a return from it for a very long time. The market has been traditionally a market where the economics are determined on a quarterly basis, right, a lot of volatility.

And that leads to there's -- somebody is always feels like they're not getting what they deserve in that equation. Either the economics aren't really attractive enough to drive this long-term investments or there's people feel like there's not enough product for me to do everything I want to do to run my business. So we think very deeply about that, like how do we get these business models aligned. And that's really what these new business models are all about. That's why we call them NBM agreements. Is about how do we get our business model, which is an unnatural thing, I think, for our customers.

We invest for the long term. We're investing for growth, right? We're investing for mid- to high teens growth. That means we run a fab, every single day, there's more wafers than the day before. Like there's more wafers tomorrow. There's more bits tomorrow than there was yesterday. And that's true every single day. So we have to sell those every single day. You build this huge asset, you turn it on and the idea is you don't want to turn it off. It's very expensive to turn it off.

That's not necessarily our customers' business model. They don't really think about, "Hey, how do I consume on a very predictable basis? How do I consume where I grow all the time"? How do I -- it's just in the way the market has worked that out in the past to say, well, every quarter, we'll just decide what the price is. And then we'll have this clearing price and then you get all this volatility and then you get all the cyclicality. So what we're trying to do is really three things. I think running a technology franchise. Number one, we're trying to get a fair return for our products, right?

That's like always the #1 thing when you run a technology business, you want to get a fair return for what you've built. Last quarter, we drove 78.4% gross margin. We guided for slightly higher than that. I think we can say we're in an attractive neighborhood as my friend, Luis would say. So that's attractive return.

Number two, we want to get rid of the cyclicality or at least dampen the cyclicality -- or at least when the cyclicality comes have different techniques to deal with it than we have in the past. The traditional way to deal with the cyclicality is I've got to turn my inventory into cash because I need cash to run my business.

We want to get out of that model. So what are we doing? We're strengthening our balance sheet. We're building cash reserves. We're putting contracts in place where when that moment happens, if it happens, there's a financial benefit to us and it's not just we're like wondering what to do in that moment.

So get a fair return for your product deal with the cyclicality. And then the third thing in any scale technology franchise, you need to grow, right? I think all of you here, that's what you want. You want to invest in growing businesses, those are fun. And I think one of the big advantages of our technology -- our business is it's going to grow, right? That's not the debate. The debate is how fast it's going to grow.

Well, the issue is, depending on how fast you grow it, the other two get all out of whack, right? So it's how do you balance these three things. And that's what we're doing with these new business models. So it's a little different than other markets. We take pages out of those markets, right? Both Luis and I have run a lot of software businesses, quite frankly, we run a lot of high-scale recurring revenue software business.

We take lessons out of those. Those are very attractive models, right? They get a slightly different multiple than we do today. So we're taking pages out of all of those in constructing a business model for our business with the technology franchise we have. That technology franchise is spectacular. It's important to almost every other technology that's built in the world. It's very scalable. It's difficult to do. We have a lot of expertise -- and we're putting a business model around that, that aligns better with our customers, so we can get those two things aligned.

And we have more visibility to not just demand but economics of those demand and that demand is growing. And we're doing that in a way where we can balance that equation. We get the right economics, we deal with the cyclicality, which has traditionally been extremely corrosive to the industry, in my perspective. I won't speak for the industry, I'll speak for our business, but extremely corrosive to our business, and then we want to grow.

We want to do all three of those things, and we're balancing that equation. And I think these business models that we're putting in place do exactly that. And it's extremely exciting because I think the place we're going to and we're kind of -- we're on that journey. We haven't been on it for very long, but it's rapidly changing, and that's a lot of fun.

We're in a business that's been around for a very long time. It's a very big business, and we're able to change the business model in a way where our customers are very happy with these agreements, right, because they want predictability of supply. And we're very happy because we want to supply. That's the business we're in. They want to buy NAND and it turns out we want to produce NAND. And so that's a great marriage. And the issue is how do we get the business model in place that allows us both to do that and both to be happy.

And so I think we have a very willing partner that's going down that path with us and defining those models.

That's perfect. And we'll discuss more about the new business model transformation. But I want to rewind back last year to your February Analyst Day. The team, the industry was still in the midst of a somewhat oversupplied environment in NAND. I clearly remember this, you put a stake in the ground at the Analyst Day and articulated a demand profile that would rise above supply, the inflection would happen.

You said the inflection is going to happen in the second half of '25 and continue to be undersupplied through 2026 and kudos to you and the team and the crystal ball for nailing this supply-demand inflection, right? Your team always does a great job of understanding the demand drivers and the industry supply dynamics, integrating some of the new business model, constructs that you just talked about, but putting on that same cap, putting on the crystal ball, leveraging your your team, I mean, how does the team seem to see the supply-demand environment as we move through 2027?

So first of all, thank you for the kind remarks. I mean we do a tremendous amount of work. Like we're in the market every day. Like there's a lot of people that observe this market, and I have a lot of respect for people that are trying to observe this market and draw conclusions. I can say it's very hard to do. But we're in the market. So we think we have unique insights and we talk to our customers and we understand the dynamics, and we build our own supply-demand models.

And you're right. In February, when we got on stage and launched the company, we had a very clear point of view that the market was going to inflect in the second half of the year. You could say there might have been some skepticism of that afterwards, but that was fine. We had our conviction and we went off and executed it. And we still have the same level of conviction. We see this market undersupplied for a long period of time.

It's a very dynamic market now. At that point, we said through the end of '26 because frankly, that's the visibility we had. I mean, at that point, it was February of '25 through the end of '26 was a long time away in this market, but we had conviction through that. And I think we can say through the end of '27, we have that same level of conviction now.

And that, what we're going to do is use that period to, one, produce tremendous products they focused on delivering tremendous products. That's always the #1 thing we need to do. If you don't do that, then you can't do any of the rest of it, but also use that period to kind of get this business models aligned with our customers so that we can predictably drive this business forward for our mutual benefit for a long period of time.

The new business model, multiyear agreements with customers. Clearly, the centerpiece are a strong part of the strategic transformation of the company since spinning out. You've signed five multiyear supply partnerships with what you've described as committed supply for customers, committed financials for the Sandisk team, backed by firm financial guarantees the longest contract is five years.

More importantly, you've said over 1/3 of fiscal '27 bits are now under these -- covered under these firm customer commitments. So first, -- can you just walk us through the conversation that you have that unfolds, right? The practical mechanics, how do these NBM conversations unfold? What's the typical customer journey from first conversation to sign contracts. And typically, like how long does it take to wrap something like this up?

Yes. So the conversations have been very a win-win kind of relationship with our customers. Most cases, our customers are approaching us. And the single first thing they say is, how can you ensure there will be supply. It's not a price. It's first of all, can you give us -- can you guarantee supply? And those conversations, they take time, right, because you're not buying something for a quarter for -- there is a huge amount of money involved.

So we go through a lengthy conversation to really understand their needs and the conversations start with how many years of commitment are you willing to make. Some customers are more comfortable committing to three years, some are more comfortable going to five. Some have less visibility and willing to start small.

Now most of these conversations are, "Hey, we want to go longer. We just don't have that visibility at this point in time." The second conversation is how many bits are you willing to buy? And normally, what we see is the procurement teams are being a little bit more cautious than what they are hearing from their engineering teams. And we go through these negotiations and an important feature we want is we want to bet with strategic and winning customers. And we define that by customers that are growing faster than the bits that we're producing.

So faster than the mid-teens, and therefore, they will be taking a higher percentage of our business. And then the conversation is about price. And we're not talking about current prices. We're talking about what is the price that we think makes sense for us and what price makes sense for them. And as we've talked in the shorter term, we define those prices mostly fixed.

And as you go out into the future, there are some variable components. Think about mostly kind of a floor and ceiling concept in which this floor is attractive for us and the ceiling obviously would be even more attractive. And the reason we do that is because prices will fluctuate. We don't know in which direction they will be in five years. And we know one of us is going to be unhappy if we fix the price. If the price goes higher, we would be unhappy because we would have left money on the table. If the prices go low, our customers would be unhappy.

So we want to make sure that each of us can capture some of that volatility, not all of it because we're signing a contract, and we don't want that 100% volatility. But we are getting -- we can capture some of that value. And the last component is we need to make sure that if the customer, for some reason, walks away, there is a financial guarantee where we are in a very good position.

So we've been negotiating a financial instruments through which it's very clearly defined that if they walk away, the funds will flow through us. The funds are either with us or with a third-party financial institution. And the criteria has been predetermined. Think about as a clean-up, although I hate that concept. But just an easy way to understand it. It just said at the very beginning, law lawsuits, no discussions. It's just very simple procedure.

That makes a lot of sense. Now I appreciate the framework and how the team sort of thinks about it. As of the earnings call in late April, you had 5 NBMs committed to the first 3 that you signed in the March quarter have a total lifetime contracted value, $42 billion. Has the team signed more NBM since earnings? And of the total or cumulative NBMs that you've now signed, what is the total lifetime contracted value and/or percentage of bits that are covered as you think about your shipment profile in fiscal '27.

Yes. We'll give you an update in a few weeks when we report or 1.5 months when we report our earnings. It's a little bit premature at this point. We just had our earnings call. But as we mentioned, we're in active negotiations with additional customers, and we're in conversations across different end markets, and we will give you an update soon.

So if you look across your NBM engagements signed or in discussions, right? And are you seeing an interest level across your data center edge/client business, consumer customers. In other words, breadth of customers by end market. Is that there?

And when you're not signing an NBM with customer who comes to the table? What's the typical reason? And is there any concern that those customers reduced their relationship with Sandisk over time as you preferentially allocate to your NBM partners.

Yes. So if you think about the three end markets, this new business model really does not apply to consumer, to what we call, right? So that's a different type of business. We love that business. We're committed to it. We'll continue to drive it. But these new business models really don't apply there. But we are in active conversations across the other two end markets, and we believe that's the case. Now you're asking me a very difficult question, which is what are the causes where a new business model has failed?

Correct.

And I'll tell you that hasn't been the case. No, we are either -- we closed the deals we've closed, and we're in active conversations in all the conversations we started, and we'll give you an update when we have more news to share.

Let me make a few comments on that. So we think it's very important. We think long-term value creation in this franchise, diversity is a big part, so optionality. So we're not going to exit any markets to favor other markets. So we want customers across our whole portfolio, right? We have a great consumer business. We've talked about that. We have a great edge business, device business, right?

Very large share there. We have tremendous innovation over the years. So we want demand in those markets to keep those products world-class. And we've talked a lot about data center, our products. We now have world-class products and demand is driven for those. So across these agreements, our desire would be to have different time frames because you don't want everything ending at the same time. You want diversity, portfolio theory. You want different end markets because that keeps the engine running inside the company.

And we're going to construct that portfolio, right? And then we'll go back to those customers and extend them or whatever we need to do. Now there are a set of customers that are not ready to opt into this model. That's fine. We have stunning customers. All of our customers are incredible companies. They all do amazing things. They build incredible products, and it's a privilege to work with all of them. If they're not ready for this model or they don't think it's right for them, that's fine.

We're still going to have a percentage of our business that we engage in the typical -- there are some customers just aren't big enough, right? There's a long tail of customers that don't -- they're not driving this much demand that are wonderful customers and they're great businesses to be in.

So as we get there, we'll talk more about it about where we land and what percentage of our business is covered under these agreements. But clearly, the amount of our supply that's left available for this quarterly negotiation is going to go down. And that's a good thing, right? We want -- again, it's just more part of this diversity. We want a broad diversity of how we do business. But what we really want is a major part of the portfolio where we understand the economics years into the future, because that -- again, go back to these three things we're trying to sell.

We want great economics. We want to deal with the cyclicality in a very different way than we've dealt with it in the past, which is just episodic where it's like, "Oh my God", like all of a sudden, we have all these situations where we've got to turn inventory into cash and we've got to do all these kinds of things. We're leaving that world behind, right?

It doesn't mean that the world is still going to have recessions. Things are still going to change. Stuff is going to go up and down. The issue is how we respond to it. We're going to be in a very, very different position in our ability to respond to that and make it as short as possible, way more flexibility -- and then we want to grow. And like I said, we got the growth lever taken care of. Again, to me, as a long-time technology manager, like that's the hardest one is to grow, right?

Because usually, when you get to highly profitable businesses that are large, you run out of market and you stop growing and then you got to go do a whole bunch of other stuff to grow. But we got that one taken care of. I think the debate is, well, could you grow faster? Maybe we could grow faster. But we want to do all three of these things at the same time. So we think that's kind of the way we want to structure the portfolio.

And it's extremely exciting because we think that's within our grasp. And the customers that we need to do that with are coming down that path willingly. They're pulling us down that, right? We're just trying to -- we have to innovate around, again, back to this, how do we get our business model, which is an unnatural thing, right? You build a fab, you turn it on, wafers just come out, you better sell them. With their business model, how do we get those two things aligned in a way it works for both of us. We've made, I think, really good progress on that, and we're continuing to having more discussions.

And as you rightly pointed out, growth is a paramount sort of focus issue -- the great thing about these NBMs, in general, whether customers sign or not, you're getting them to the table. You're having them unveiled what they think is their outlook for storage demand for the next sort of several years, right? As you aggregate all of this, I mean, has this changed your views? And then you have a segment of the market data center, which wasn't that big of a driver a few years ago, which is clearly going to become a very significant driver of the franchise going forward.

But whether it's bringing all these customers to the table, you've got this new demand driver from data center. As you aggregate all of this, has this changed your views on driving mid- to high teens type bit supply growth profile over the next few years?

Yes. I'm going to go back to kind of the way I've been framing it. So there's always a debate -- the debate always seems to be around, could you grow faster or should you grow faster? And what we're trying to do is balance all three of these things. And traditionally, the issue in this market is if you start growing faster, then you -- the other two get messed up very quickly.

And so we're very conscious of that, and it's about getting more visibility. So as we get more visibility into the portfolio, I think the answer to these questions will be easier. But right now, we -- again, we got on stage -- it was only February of last year where you started this question. We got on stage and we said the market is going to grow mid- to high teens, and we think pricing is going to inflect up in the second half.

And everybody told us we were wrong, right? And they gave our company a $7 billion valuation. That was only a little over a year ago. And now it's like, well, you should grow faster and you should do it all this stuff. It's like, "Hey, look, we have a model, and we believe in that model and that we believe that mid- to high teens bit growth is sustainable over time." As our customers come to us and they're willing to make longer commitments, that will give us more visibility into what the future looks like. And that's a good thing for everybody.

And on the economic front, on the economic profile front, given the strong pricing environment and margin expansion as a result of the rather tight supply environment, there hasn't been as much focus on cost downs by the market as it relates to Sandisk's execution. However, as you negotiate these NBMs, I mean, the economic framework is important. And so there are certain assumptions, Luis, that I assume that you're baking in from a cost down perspective, right, as you think about the overall economics to Sandisk, but you have laid out a cost profile -- cost per bit cost down profile of low teens type of annualized rate?

So first of all, is the low teens annualized rate kind of still the way to think about how you're driving your cost per bit profile? And then maybe if you can just level set us year-to-date, like what has been your cost per bit decline trajectory and your outlook for this fiscal year?

Yes. So first of all, we will always focus on cost, right? That's our job, and we'll continue to drive that. The cost downs are less than they used to be. And frankly, it's not where we want to focus the conversations with our customers, right? It's about the value we create, the value they create with our products. And we'll always keep on working on.

But it's a part of the economic -- I mean, your customers don't focus on it. But as you think about your economics, you obviously have to factor that in, right?

Right. When we told you that the deals we signed will be consistent with the gross margins we guided, -- all of that is taking into account the current cost structure or everything we see and that -- so that's factored in, but we don't talk about costs.

Harlan, this is, I think, one of the most interesting things for me in the last 1.5 years. So when we launched the company, I remember, we got on stage and I just said, I'm going to stop talking about cost down. Like every earnings call, what's your cost down this quarter, what's it going to be next quarter, we're going to stop talking about this, right? I've never seen an industry where people tell you what their costs are. Like it's -- the business was managed from the cost side of it. Oh, if we drive our cost down, the economics are work. We're just going to stop talking about that. It's been kind of interesting over the last year. People stopped asking about it. It's been great.

They've stopped asking about it partially because pricing has done well. But from my perspective, cost is important, because it's a reflection of the team's execution. It's a part of the reason why Sandisk is -- has the best-in-class manufacturing technology, right? And that's part of the reason why I'm asking that question.

You're 100% correct. I appreciate the question, and I appreciate the comments. Like, look, one of the things I said at the beginning, one of the things I'm so excited about this business we have an enormous amount of R&D productivity in the TAM, right? We know what the BICS road map is going to be years into the future. And you're right, we can increase productivity. We can drive more bits. We don't need to do -- we don't need to build a whole bunch of greenfield and all this kind of stuff because we get the productivity through R&D. And the consequence of those additional supply is they are at a lower cost, right?

So that is true. It's just not -- that's our side of the equation to keep -- and as Luis said, every business drives more productivity. Every business drives more productivity. That's our job. We have -- to your point of your question, we have a big lever that a lot of people don't have of driving productivity, which also drives growth.

Exactly.

It's an amazing thing for the financial model of this franchise. It's incredible. To Luis' point, we're still very focused on it. It's just not the way we run the business anymore, which it used to be the way we run the business, which is was, "Hey, there's infinite elasticity if we just lower the cost, everybody will make money." It turns out that was never really true. I don't think it was ever -- maybe it was true in the 2D era, but never really true in the 3D era. And now we're getting to a different way of running the business, which is the value -- the intrinsic value of the technology being recognized, supply discipline, matching supply to attractive demand, not just, "Hey, we can drive the cost down, we should supply more into the market." It doesn't mean that it's not still there. It is still there.

Absolutely.

We just don't -- it's not something where we're not really highlighting that number at this point.

Let's focus on the area of performance and differentiation. You've said it many times, Dave. But your customers' willingness to enter into these NBMs, these multiyear agreements is, first and foremost, driven by differentiation, right, performance differentiation.

And I think one of the areas where the team has done an outstanding job relative to the last 20 years plus that I've covered, Sandisk stand-alone, Sandisk is a part of WD, Sandisk stand-alone, again, is the team has made significant improvements in your data center enterprise SSD portfolio, right?

And success in data center solutions, as you guys know, is not just based on having the best-in-class well performance like base NAND manufacturing technology, right? Success in enterprise SSD is also that plus I have to have a great silicon controller technology. And then even more importantly, in some cases, I've got to have the right software and firmware, right? And so with this dramatic growth in your enterprise SSD business, along with share gains, I mean, what has the team done to architect a better controller chip design? What has the team done to enhance the firmware, software, which are all key determinants and very strong success factors for your customers?

So again, I appreciate the kind words. I mean, one of the things I've always appreciated about this business, whether it was part of WD or now as a separate business is the R&D team. We just have incredible R&D capability, incredible existing IP portfolio. And in a world where you have AI now, which is like the most exciting technologies, when you have this unbelievable R&D team, they're going to come up with new ideas and be able to innovate.

And that's leading to all kinds of other projects we probably won't have time to talk about like HBF. But -- so why the progress on enterprise SSD all of a sudden, right? So if you look at -- I kind of look at Sandisk, maybe a couple of very big arcs of Sandisk. Sandisk was a great consumer business. The WD era was really about the client, and that's not surprising because the SSD was controlling the HDD, you knew the customers, you knew the use case, you knew the requirements.

And so that was kind of the rise of that portfolio. And then towards the end of that era, new management team came in with more of an enterprise background, obviously, the rise of the cloud in both franchises, all that technology expertise that had first like done all this incredible controller work in consumer and then done all this incredible controller work in client.

That team was moved on to enterprise over the last three to four years. You run that process for three or four years with focus of people that understand how to be in these markets and you end up with great products. And so we haven't forgotten how to do the other one. So it's an accumulating thing. And now we're at a point where we've got, first of all, great underlying NAND technology.

That's the JV story, right? That's the foundation of that. We just signed another contract. We're through 34, we're partners with Kioxia, we're super happy about that. That gives us unbelievable NAND. So the wafer comes out of the fab, what are you going to do with it? Now we have this unbelievable team. They built a great portfolio in consumer. That team then built a great portfolio in client.

Now that team has built a great portfolio in enterprise. And we got exactly the right products at exactly the right time. Market is exploding, a lot of demand from data center. We've got incredible NAND technology, to your point, it's differentiated, wafer bonding, better performance, all these kinds of issues. And we've built clean sheet architectures that we're just starting to ship of new enterprise -- new class of enterprise SSD right at the moment when that market is really pulling on those products.

That's the story of what's happening right now, and we're wrapping a whole new business model around it so that this franchise looks very different going forward than it looks going backwards from a financial model perspective.

I appreciate the insights, Dave, Luis, thank you very much. Looking forward to continuing to monitor the execution of the team as the year unfolds.

Thanks Harlan. I appreciate being here.

Thank you.

Good afternoon, and welcome to Sandisk's Third Quarter Fiscal Year 2026 Earnings Conference Call. [Operator Instructions] Note this event is being recorded. I would now like to turn the conference over to Ivan Donaldson, Vice President of Investor Relations. Please go ahead.

Before we begin, please note that today's discussion will contain forward-looking statements based on management's current assumptions and expectations, which are subject to various risks and uncertainties. These forward-looking statements, including expectations for our technology and product portfolio, our business plans and performance, our capital allocation priorities, market trends and opportunities and our future financial results. We assume no obligation to update these statements. Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations.

We will also make references to non-GAAP financial measures today. Reconciliations between the non-GAAP and comparable GAAP financial measures are included in the written materials posted in the Investor Relations section of our website.

With that, I'll turn the call over to David.

Thanks, Ivan. Good afternoon, and thank you for joining Sandisk's fiscal third quarter earnings call. We delivered another strong quarter with excellent performance across all key metrics, reflecting the strength of the Sandisk franchise. Before turning to our end markets, I'd like to provide an update on a priority we previously outlined. Last quarter, we were engaged in discussions with customers on multiyear supply partnerships. What we refer to as new business models or NBMs. I am pleased to share that we have successfully advanced those conversations with 5 multiyear partnerships signed so far. These partnerships are structured to lock in committed supply for our customers and committed financials for Sandisk.

Our customers' commitments are backed by firm financial guarantees. These partnerships support durable structurally higher earnings and a significantly more predictable and less cyclical business for Sandisk. We believe this marks a fundamental evolution of our business, which is centered on deeper customer alignment, enhanced visibility and long-term value creation. These MBMs reflect the strategic value of our world-class NAND technology, which is built on decades of innovation. The investment we've made in R&D and manufacturing, including tens of billions of dollars in cumulative CapEx and IP have built the foundation for a powerful new business model in which we manage the full stack from front-end manufacturing through chip and system level design to final back-end assembly and test. Both the extension of our joint venture with Kioxia and the supply agreement for DRAM following our investment in Nanya, further strengthen our supply chain resiliency. This leverage is enabling us to drive stronger customer engagement, allowing long-term conversations with partners who value technology performance and long-term supply assurance. With increased engagement in the optionality across the portfolio, we can optimize our end market mix more effectively. Together, these transformations have resulted in a step change in what we believe to be sustainable gross margins, free cash flow generation and earnings power in a market that we expect to grow in the double digits for the foreseeable future.

Data center is a clear example of this strategy in action with revenue growing 233% sequentially. This milestone reflects years of preparation and our deliberate shift toward what is now the most strategic and fastest-growing end market. While we have made substantial progress, there is significant growth opportunities ahead driven by the fundamental shift in underlying infrastructure requirements of artificial intelligence.

We are witnessing extraordinary growth, not just in model size, but in resulting token generation, the duration and complexity of model runs in the increasing importance of context. As AI models scale from billions to trillions of parameters and deployments advance from simple inference to deep reasoning, and increasingly autonomous agentic systems, NAND has become a critical component of the underlying infrastructure. Inference optimizations such as KV cache, along with workloads like RAG, require substantial high-performance, low-latency flash to deliver real-time responsiveness and quality of user experience. These workloads expand the amount of data that now needs to be stored on low-latency flash, which is well beyond the model itself as systems must retain context, intermediate data and large external data sets. As a result, NAND flash is emerging as the only economically viable solution to deliver that capacity, performance and efficiency required to keep models accessible for real-time inference at scale.

This shift in understanding the critical nature of our technology comes at a time when our product differentiation is strongest. Anchored in what has been recognized as an industry gold standard for NAND technology with BiCS8 and a broad leading portfolio with TLC and QLC offerings. We are confident that our world-class product portfolio and technology leadership will continue to drive data center customers to see Sandisk as a partner of choice over the long term, and we are already seeing that preference translate into results.

Our fiscal third quarter revenue was enhanced by strong demand for our TLC-based enterprise SSD portfolio, which powers performance-intensive compute workloads where speed and latency are paramount. Looking ahead to the fiscal fourth quarter, we expect to begin shipping our QLC Stargate solutions for revenue, adding another layer of revenue growth. Together, TLC and QLC serve distinct but complementary roles, reflecting how we are deliberately architecting our portfolio to meet evolving customer needs with our broad portfolio of AI-focused data center products.

In edge, we are seeing a continued shift towards premium devices across both PC and smartphone markets. These platforms are increasingly incorporating on-device capabilities, which are driving higher storage requirements and greater demand for high-performance solution. As a result, our mix continues to shift to high-value configuration and customers that assign the appropriate value to our technology. Consumer saw a strong year-over-year revenue growth across all key storage categories and regions despite evolving consumer industry dynamics. This performance was supported by our strong brand recognition and channel presence as we focused on the most financially attractive demand. In February, we unveiled our next-generation portable SSD portfolio, designed to support faster, more demanding workflows and AI-enabled content creation. This launch reinforced our innovation and leadership in the SSD category generating meaningful external visibility with coverage across multiple global media outlets.

We also continue to strengthen global consumer engagement through new brand-led go-to-market activities, such as our Space to Hold More campaign, which is driving deeper customer connection by localizing global narratives and engaging diverse communities worldwide. Together, these efforts reflect our focus on our end markets and commitment to driving demand through brand recognition, product innovation, and strong go-to-market execution as we shift our portfolio toward higher-value opportunities and transition away from legacy upsell models. Our broad end market exposure sets us apart and we remain committed to serving customers across these markets. With that, I'll turn the call over to Luis to dive deeper into our financial performance and guidance.

Thank you, David. I will begin with an update on our new business models, or NBMs, which are designed to provide us with demand certainty and provide our customers with supply assurance. We signed 3 agreements in the third quarter an additional 2 so far in the fourth quarter, and we're currently in active negotiations with several other customers. These agreements are tailored to meet the needs of our customers and in aggregate, provide us with demand certainty and financials that we expect will be consistent with our fiscal fourth quarter guidance. The duration of this agreement varies, with the longest contract extending to 5 years. In aggregate, volume commitments increased during the life of the contracts with quarterly commitments and a combination of fixed and variable pricing. This agreement with variable pricing allow us to capture upside if prices rise while allowing our customers some upside if prices decline over time.

As you will see in our 10-Q, the 3 contracts signed during the quarter provide minimum contractual revenue of approximately $42 billion. We will update you as we make more progress. Each contract is secured with financial guarantees that protect us if the purchase obligations are not fully performed by our customers. In aggregate, the 5 agreements signed so far include financial guarantees that exceed $11 billion and include prepayments and other financial instruments, managed by third-party financial institutions. Out of these agreements, $0.4 billion in prepayments are included in our Q3 balance sheet. These 5 new business models account for over 1/3 of our BiCS in fiscal year 2027, which we expect to increase as we conclude additional agreements over the next few months.

We expect these new business models to reshape the historical cyclicality of our business, improving visibility and results in pricing and margins that reflect the value of our technology and investments, ultimately delivering higher, more consistent and durable returns for shareholders.

Moving on to our results for the quarter. Revenue for the third quarter was $5,950 million, up 97% sequentially and up 251% year-over-year. This compares favorably to our guidance of $4,400 million to $4,800 million and was driven by both a mix shift towards higher value customers and higher pricing. Our BiCS shipments were flat year-over-year and down high teens sequentially as we build higher inventory levels, primarily to support strong BiCS8 QLC demand in the fourth quarter Stargate ramp and to prepare for our recently signed new business models. In line with our mid- to high teens growth model, BiCS shipments increased 18% fiscal year-to-date.

Moving on to the end markets. Sequentially, data center revenue grew 233% to $1,467 million. Edge grew 118% to $3,663 million. And Consumer came in at $820 million, down 10%, in line with our historical seasonality. Our portfolio planning strategy focuses on delivering attractive long-term economics with diversification remaining a core strength. We remain committed to serving all 3 end markets to maximize long-term value creation.

Our non-GAAP gross margin for the third quarter was 78.4%, up from 51.1% in the prior quarter. This compares favorably to our guidance of 65% to 67% and was driven by our shift towards higher value mix and the overall pricing environment. Non-GAAP operating expenses for the third quarter were $448 million, and represents 7.5% of revenue as compared to 13.7% of revenue in the prior quarter as we generate additional leverage. This compares favorably to our guidance range of $450 million to $470 million. As a result, non-GAAP operating margin were 70.9%, up from 37.5% in the prior quarter. Non-GAAP EPS was $23.41, up from $6.20 in the prior quarter. This compares favorably to our guidance range of $12 to $14.

Key GAAP to non-GAAP reconciliation items include $20 million in stock-based compensation, net of taxes which represents 0.3% of revenue and $46 million related to the write-off of unamortized issuance fees as a result of our repayment of the remaining $650 million balance in our TLB. We closed the quarter with $3,735 million in cash and cash equivalents on our balance sheet.

Moving on to free cash flow. During the quarter, we generated $2,955 million in adjusted free cash flow, which represents a 49.7% margin. Cash flow from operations came in at $3,038 million, partially offset by $83 million from net cash capital spending. Gross capital expenditures totaled $240 million and represented 4% of revenue. Our capital plan is designed to balance growth opportunities and generate attractive returns while supporting our ongoing BiCS8 transition. We remain highly disciplined in how we evaluate such investments to protect long-term sustainability of our business financials.

Moving on to guidance. For the fourth quarter, we forecast revenue between $7,750 million to $8,250 million from both BiCS growth and higher pricing. Our forecast for non-GAAP gross margin is between 79% and 81%. We expect non-GAAP operating expenses between $480 million and $500 million as we continue to invest in innovation and R&D. We expect non-GAAP interest and other income between $10 million and $30 million and non-GAAP tax expenses between $775 million and $875 million. We forecast non-GAAP EPS between $30 and $33, assuming $158 million fully diluted shares.

Moving to capital allocation. The priorities we outlined in February of last year were to invest in the business, achieve a net cash position and then return cash to shareholders. In line with these priorities, we have taken steps over the last 2 quarters to solidify our supply chain, including extending our JV with Kioxia through December 2034, and investing approximately $1 billion in Nanya to secure long-term DRAM supply. We have also taken actions that put us in a strong net cash position by paying off the remaining balance of our TLB.

Given the strong progress Today, we're announcing that our Board of Directors have authorized a $6 billion share buyback program of outstanding shares of common stock. The repurchase authorization is affecting immediately with no expiration date. With that, I'll turn the call back to David for closing remarks.

Thank you, Luis. In summary, we continue to execute with conviction at a critical inflection point for this business. NAND has always been a foundational technology, empowering the world's best-in-class semiconductor storage solutions required to drive the largest technological movements, including PC, mobile, cloud and now artificial intelligence.

Data center has become our fastest-growing market, and the workloads driving that demand, including inference, reasoning and agentic systems represent a structural and durable shift in how the world's most consequential technology is built and deployed. Our new business models reflect this shift. 5 signed agreements to date, over $11 billion in financial guarantees and over 1/3 of our BiCS in fiscal year 2027 under firm customer commitments represent a fundamental reshaping of our business, providing visibility, pricing protection and more consistent durable returns.

Our technology and product portfolio are intersecting this extraordinary demand at exactly the right moment. Equipped with a complete portfolio that now includes a scaled and rapidly growing enterprise SSD business, we are allocating supply to the highest value opportunities in establishing a new pillar of growth for Sandisk.

This progress has converged in a single moment. We believe our margins are sustainable. We've achieved our net cash target. And we've announced plans to return capital to shareholders through buybacks, all while reinforcing our operational foundation. Combined with our multiyear NBMs and the acceleration in the data center end market, this gives us both financial strength and structural resilience.

The result is a durable growth model, a valuable franchise and a business built to generate substantial sustained cash flow. With that, Ivan, let's see if there's any questions.

[Operator Instructions] The first question comes from Mark Newman with Bernstein.

Congrats on another great quarter. A couple of just quick questions here. So the EPS guidance you've given, $30 to $33, I mean these are all fantastic numbers. It does imply that the raise of price increase is slowing a bit into the the current quarter. And I just wondered if that is either being conservative on your side because obviously, we're still quite early in the quarter? Or is that related to some of these very, very long-term agreements that you've signed. And with regards to the long-term agreements, I believe you mentioned about 1/3 of BiCS for FY '27 in some kind of long-term agreement. I'd like to ask to what degree is price fixed in the coming quarters, just so I can get a kind of sense for that.

Mark, it's good to hear from you, and thanks for the comments. So first on next quarter in pricing, so we don't really guide pricing. But I think you saw in FQ3 rather extraordinary pricing acceleration across the business. So we're very happy about that. And you're right, it's early in the quarter, and it's an extremely dynamic market. So it pays to be a bit conservative when you're going down that path. But we're very confident in the numbers. The second part on the agreements, I'll make a few comments and Luis will have something to say as well. I think you were asking about pricing being fixed. So these agreements are really tailored to individual customers. So they have different elements depending on the customer, it depends on the length of the agreement that really gives us some assurance on consistency of demand, which is really what we need. And again, a lot of this is -- I've talked about this in the past, we need to get our customers' business model and our business model aligned. We run a fab. We have very consistent output. We need very consistent consumption. And I think the primary -- one of the major attributes of these agreements is they give us that. And our customers have -- they understand the dynamic very clearly when we talk it through. It's one of the reasons why these agreements don't just happen overnight. It's not just about prepaying for a couple of quarters' worth of supply. This is about establishing up to a 5-year agreement on supply that's very consistent quarter-over-quarter. And as we said, there's financial instruments in place that if that consumption does not happen on that very predictable timeframe, then there's financial commitments that come to us immediately. So they are backed up and they're very, very strong. The pricing, like I said, it's set up where there's fixed elements, there's variable elements. Maybe I'll let Luis talk about it in a little more detail.

I just want to reinforce what David was saying. These models are here to deliver durable more predictable, more attractive, more consistent financial results. So they're very good. And frankly, it's a win-win for us and for our customers. We provide supply, they provide demand and we have visibility for many years all the way to 5 years. So we're very happy about that. I guess you've never heard about us talk about RPO, or remaining performance obligations in this business, and we started to talk about that, and you will see it in our 10-Q, as we mentioned, about $42 billion of RPO in this business. So we're very happy about that. The pricing, as we mentioned, it's a combination of fixed and variable to address your question directly. The shorter time you are in within the contract, the more fixed it is, the longer out you go out, there would be more components of variable. So you could assume that most of the pricing that we're seeing in the very short term is mostly fixed. And then as you go out, there is a little bit more variable for us to capture upside and for our customers to capture some upside if prices were to go down.

The next question comes from Joe Moore with Morgan Stanley.

Wondering if you could talk about the growth in enterprise SSD that you saw pretty impressive. How much of that is the market? How much of that is you guys kind of put in the product portfolio in a better place?

Yes. Like a lot of things, when you see 233% sequential growth, Joe, there's a lot of elements of that, to your point. So it starts with the portfolio, right? I mean the portfolio is in great shape. Our TLC -- this is almost exclusively our TLC product. We're going to start shipping our Stargate project -- our product for revenue next quarter. So really, really strong performance; very, very strong product; a broadening of qualifications. It takes a little while to get into all these accounts. So a large number of accounts at this point. And then there is a strong market pull. There's no doubt about it. There's a lot of demand in the market for these high performance enterprise SSDs. We got the right product at the right time. And we're really happy to see this part of the portfolio now expand and getting to the levels where we expect it to be. It was 25% of the portfolio this quarter, and we expect that to increase as we go forward.

And my follow-up, I mean, where do you see that in a few years? It seems like hyperscalers everything you can make and then some just how much of the business could be enterprise in the long term?

It could be a significant amount. I mean, I think, Joe, you've known for a long time, we value a balance portfolio. So we're going to -- and we want to mix in the way that gets the best financial return for us, and that changes quarter-over-quarter. But I think the key point is we're in a position where we can mix into this -- into data center in a way that we've never been able to do before. So I expect that number to keep rising over the next several quarters and the next several years, to your point.

The next question comes from Ben Reitzes with Melius Research.

It's great to be on board here. I almost feel like I'm on a software call here. So it's great to be speaking with you. Dave, I wanted to talk about the -- you said 1/3 of BiCS growth next year is contracted. Where do you see this going to based on your conversations? Should we expect that every year, you're getting -- it's 1/3 next year, but like every year, can this be above 50% when you know how much is kind of done going into the year? And then I have a follow-up.

Yes. So first of all, welcome. We're glad you're here. It's good to hear your voice again. So Look, I would say a couple of things to this. One is we're just -- we're still in a lot of conversations about how we're changing this business. And it takes a little bit of a while -- it takes a while, depending on the customers. Some customers are come into the conversation really concerned about multiyear supply agreements. And so it's an easier conversation. Other customers come into the conversation, very used to the way the market has worked in the past, where they commit volume and want to negotiate price every quarter. That's not the kind of agreement we're interested in. We're interested in agreements that give us certainty of economics. And a key point of what Luis said in the script, I want to make sure it's understood. There are fixed and variable elements of these agreements, but we're targeting the 5 agreements we've signed, we're targeting financials that are in line with what we just forecasted to. So this is a very, very attractive business. Now then we're still in active conversations for our supply going forward. That includes next year all the way through the next 5 years. So I expect the number that we said at least 1/3. So we're over 1/3, and I expect that number to go up over the next several quarters. Where can it get to? I definitely think it can get above 50%. And -- but we'll see. And I think we can drive it actually quite high, and we have a desire to drive it quite high.

Great. And then just with regard to margins, Dave. I mean do you feel like with -- when you do these kind of agreements that you can lock in margins. I mean, your stock is not trading as if you're going to stay at 80% anyway. But when you do these kind of things, is there a target margin? And will you be willing to kind of share what range at least it's in. You could argue the stock is trading like your margin is going to go back into the 40s or something or something like that. So do you have a target margin that is in a range that you're comfortable talking about?

Yes. I don't think we're there yet to talk about that, Ben. I mean, when we get a little further along in this, we'll wrap this all up in a new model for everybody. But we're very proud of our technology. Let's put it that way. I think we're finally for the first time in decades in this business, getting to the point where the value of our technology is getting recognized, at least for us. I mean, quite frankly, the value of our technology has been recognized in the market. It's just other people have been collecting that value, and it hasn't been the producers. And I think that now we're getting a more even distribution of those -- of that value. So we're not necessarily interested in trading away that value for certainty. We're interested in getting that value and getting certainty as well. And to your point, look, we understand how the franchise is valued, and we're very, very focused on getting the cyclicality out of this business. It's corrosive. It's corrosive to the way we invest our CapEx. It's corrosive to our customers' ability to get a sufficient amount of product to drive their spectacular businesses. And I think we've taken some very meaningful steps down this path now. We've got very significant commitments from very significant customers. And I think as we can continue to proceed down this path, we will move this entire business to a very different spot to everybody's benefit. Our benefit, our investors benefit, our customers benefit and their customers benefit because they're building just spectacular technology and we're a key part of that. So I think we're on that journey. I think it was very questionable if we could even make that progress. I've had a lot of people tell me in the last year, it's never going to happen. It's happening. But we're still in the early stages. And as we make continued progress, we'll continue to give everybody updates on that progress.

The next question comes from CJ Muse with Cantor Fitzgerald.

I'm just curious to get your thoughts around supply/demand going forward for NAND. It's fairly interesting in the sense that we're getting only limited greenfield, mostly layer count driving growth, where new greenfield is really being prioritized for DRAM. So within that kind of construct as well as the agentic AI kind of incremental growth, how are you thinking about when the industry might get into balance?

Yes. So I think, CJ, you know my point of view on this is the industry is always in balance, right? Markets always balance supply and demand. I think the question is -- implicit in your question of what I hear is if you lower the price, will you meet more demand. And I mean that's just -- we're kind of working around that whole environment. Look, let me start on -- make a couple of points. On the demand side, we continue to see data center accelerate. Before what we saw this week, we would raise even our calendar year '26 data center growth number to the mid-70s from where we were in the 60s just 3 months ago, which is up from the 40s 3 months before that and the 20s 3 months before that. So we continue to see very, very strong growth in the data center. Outside the data center, we're seeing some contraction in the market just because of unit decline. That's to be expected, although we expect to see that bounce back in '27. Now on the supply side, I think this is a major, major benefit of this franchise is that we can increase supply through nodal transitions right? We have a very, very productive R&D pipeline. This is something we've invested in for a very, very long time with our JV partner and the BiCS road map. And so we can continue to drive the BiCS growth we're talking about mid- to high teens through nodal transitions. We don't need -- we need to add some clean room space because each node has more steps and more steps is more tools. So there is some additional CapEx. But it's not like other markets that you referenced, where you actually have to add capacity because you're not getting that much from the nodal transition. So quite frankly, I think this is what makes this franchise such a spectacular cash generator is because the amount of CapEx we need to invest, especially CapEx as a percent of revenue is continuing to go down substantially. The absolute CapEx is still there. I'm just saying that relative to our revenue generation. And we've made all the investments in the nodal transition. So we have years of runway into what our nodes are going to be and what the BiCS growth is going to be from those. And we will continue to invest in those and drive those nodal transitions to grow the market in that mid- to high-teens rate. And that's basically what we see what we see across the NAND players, quite frankly.

Very helpful. And I guess just real quickly in terms of capital structure, you're now no debt, $3.7 billion cash. What do you think you need to retain given your view today and given kind of the new contracts that you're signing? And then how should we think about kind of buybacks from here?

Yes. So we did announce a share buyback program with this call, right? We just announced a $6 billion buyback and we'll keep on tracking our cash flow. We're generating good cash. And as things change and as we execute the share buyback program, we'll keep you updated, C.J.

The next question comes from Jim Schneider with Goldman Sachs.

One more question on the new business models, if I could. Can you maybe talk about whether any of the 5 largest U.S. hyperscalers are included in those contracts thus far? And related to this on a go-forward basis, do you plan on providing any sort of ACV or in your confirmed contract value as part of your normal disclosures?

Yes, we're not going to disclose the names of our customers. But what we have, as David said at the beginning, we have some very meaningful customers who are joining and some more that we're working with, but we can't disclose the name of our customers. I think to your second part of your question, James, we will provide you this RPO metric, which I think is very interesting, which is how much of the business is already contracted. And that's based on minimal prices, right? So we'll continue to give that information every quarter, and you have that visibility as we make progress quarter-over-quarter.

And then maybe as a quick follow-up. Can you maybe talk about given these new business models and your visibility on customer demand, what is the state of your discussions with Kioxia in terms of potentially increasing BiCS supply? And are you contemplating anything above the sort of 20% range of growth that you've outlined previously?

No. We still have the same plans and conversations with Kioxia are always very robust and very ongoing, and the teams are working on this every single day. But we have our BiCS8 transition plan that we've aligned on, and we're executing to it, and it's going extremely well.

The next question comes from Aaron Rakers with Wells Fargo.

This is Jake on for Aaron. Congrats on the great results, guys. Just to start off, looking at Stargate starting to ship for revenue in 4Q. Can you just give maybe some color on how meaningful that ramp could be over the next few quarters?

I mean, look, we have a whole another -- there are two major products in the data center space. I think we've talked about this a lot. There's the compute -- what we consider kind of compute focused enterprise SSD, lower capacities, much higher speeds and then their interface speeds and then there's much, much higher densities. Stargate is -- and the progress we've seen so far in the portfolio is coming off of that compute focused TLC drive. And now we're going to bring the whole QLC product to market, which has been under qualification with some major players for well over a year. So we're not going to forecast a specific market segment, but we're very proud of that product, and we think it's going to do quite well in the market.

Okay. And then maybe as a follow-on with some of the more powerful LLMs released over the past few weeks. I guess, how are you thinking about the KV cache opportunity as we see, Agentic AI grow? Has that meaningfully changed over the last few quarters and maybe how customer discussions have changed there?

Yes. I think that we've advanced our understanding of that a lot over the last quarter or 2, since it became a major part of the conversation. I even think the team did a webinar on that, which would be happy to repeat if folks are interested. And I think when you really start to drill into that opportunity and you try and size it. It obviously gets very complicated very quickly. What are the number of concurrent sessions that are going to be run? What's the average input tokens, what's cash hit ratios, storage durations, so there's a lot of elements to that. And I think what it says is kind of where you were going, which is we need to stay very close to our customers because they are the ones that are going to have all the detail on the infrastructure they're building, the ones that are doing infrastructure at scale are going to have the great insight into how are all those variables put together against the use cases they believe they're building to. And I think this just reinforces this business model question as our customers go through those calculations and understand the significance of NAND that, that could drive that is a good foundation for the conversation about striking deals 2 years, 3 years, 5 years in length that are very, very substantial in the amount of demand. I mean we're talking about 5 deals and more than 1/3 of our portfolio. So it's an extremely, extremely dynamic situation. I think these are all the things that go into kind of understanding where this market is right now and how fast it's moving. It's literally moving every single day. And even for those of us on the inside of the market that see the data points literally hour over hour, it is moving very, very rapidly as people start to -- our customers really start to understand the dynamics of the infrastructure they're building. And I think I feel very good about we've been able to stay very close to them. They're obviously very close to our technology and our products. They're responding very positively to those products. They understand that they're willing to commit years of purchasing with financial model around that, that is very attractive for us, and it gives them a very attractive attribute, which is guaranteed supply. And then quite frankly, they're willing to put a very large financial commitment that basically guarantees that ongoing demand. And I think that's a very big part of what we're talking about. I mean, I think that -- we've talked about these agreements a lot. We've gotten a lot of feedback from a whole lot of people. There's a lot of sometimes talk in the market that they won't hold. They won't have teeth, all these kinds of things. And I can tell you, nothing can be further from the truth. We have customers that are literally putting up billions of dollars of collateral through various financial instruments, that will survive for the life of these contracts. And if they don't meet their obligations on consistent purchasing every quarter, then that financial commitment immediately comes to us as a compensation for that contract not being concluded. So I actually don't probably never expect to collect those because I think our customers are extremely serious about needing this product. And I can tell you the normal case is we signed an agreement. And within weeks, we're having a conversation about how we increase the amount of product we can get to them over that time frame. So a very dynamic market. Things are changing very quickly. It does make it difficult to forecast in the things you see. I mean you see the results we're able to put up, and they're significantly better than we thought they were 3 months before, and that's because the market is just moving very, very quickly. And the pieces change literally day by day. And what we're doing is we want to solve a whole bunch of issues for our business in this. We know we have great technology. We've made enormously substantial investments in intellectual property. I mean we've been building the BiCS road map for decades. We have enormous investments in fabs, some of the largest fab complexes in the world with our JV partner. And we want to leverage all of that, get a fair return for our products and get the cyclicality out of the business. Because like I said, I think it's -- I think, quite frankly, it's corrosive for everybody that's in this industry. And -- well, I shouldn't say that. I can't speak for everybody. But I think what's happening is there are now customer sets that very substantial customers that don't want to play the quarter-by-quarter price game. They have spectacular businesses, and they understand that we provide a very important components to their spectacular business, and they want to make sure that they have the best products, which we believe we provide. We know we provide those, and they want them on a very consistent basis so they can continue to plan their own business. That is opening this opportunity I think to fundamentally change the way this business has worked over the last several decades. And quite frankly, that's a lot of fun for us to do that because customers are very happy with those agreements. We're very happy with them. And like I said, I think everybody wins. I'm trying to figure out who doesn't win in this equation. And so far, every agreement we've signed, the customers have been thrilled to get to the point of actually getting it signed. So anyway, probably maybe Jake a little longer answer than you were looking for, but we feel we're very, very good about where we're at. And we think that we're now starting down a path that is as Ben said, quite frankly, there's lots of other technology industries that understand how to do this. This is not. We're not reinventing the wheel. We're just using techniques that people associate with other businesses that are recurring revenue models. But certainly, everybody that -- well, not everybody, if you run a whole bunch of technology business, you understand how recurring revenue works, and it's a very, very powerful financial model. And we think we can bring it to our franchise.

The next question comes from Asiya Merchant with Citigroup.

A great set of numbers here. David, I think I heard you say some client demand, maybe with PCs or smartphones related snapping back. You sounded optimistic on that into next year. Wonder if you're seeing anything, whether it's edge AI on edge devices that underpins your optimism here. And in that same context, I mean, given that the demand/supply seems more tilted towards meeting hyperscaler demand, the data center demand. What gives you confidence that you can meet some of that client demand if it snaps back? And if I can squeeze one in more for Luis as well. CapEx, at one point, there used to be obviously mid-teens as a percentage of revenues. Obviously, your revenues are exploding here, so we don't expect that same ratio. How should we think about that going forward?

Okay. Let me unpack that a little bit and see if I can help. So when we look at '27, we see definitely PC phones units are down in both now, as you would expect. And we see those flattening out to up slightly in '27 as kind of our internal view of the world. And quite frankly, I think that that's just a reflection of market's ability to adapt, right? I mean, I think that's -- if you're in business. And I think especially the device business, those are spectacular companies, very, very smart people that run them, and they understand how to change their portfolio mix and what they need to do, given the environment they're in to drive their spectacular business forward. And I think there's no doubt there's an adjustment process right now, and it's happening. And I think we'll get through that. And I think we'll get back to a point where we're still going to see content per device increase this year at least on phones, PCs, we've got it flat. And we'll see both of those start to inflect up next year while units are flat to up slightly.

So now what will we supply. So that's an interesting question. So we're going to supply the customers that we have agreements with. That's the way we're starting to look at the market. I think this is the change that we've been talking about. We're not going to wait until next year and see what the market gives us. We're talking to edge customers as well about these new business models. Agreements, they are multiyear agreements with all the characteristics we've talked about in the past. Committed growing demand, the same kind of structure we talked about for these first 5. So those customers understand their businesses extremely well. We're engaged in those kind of conversations. We'll see if we reach the finish line on some of those, I'm sure we will. And that's how we'll have great insight into what their demand is, is because they will have told us and they will put a financial commitment behind it, which will allow us to plan a lot better what demand we're going to serve. So we're kind of navigating out of this market where we just show up and kind of see what demand is and see what the price is and then adjust our mix very rapidly. We know how to do that, but where we rather go is the path we're down, which is customers commit to us what their demand is and they committed in a way that we can really count on it. And quite frankly, they can really count on us.

I think to your -- last question on CapEx. I mean, David just talked about our philosophy. We continue to invest towards kind of a mid-teens capacity growth over time. What this translates to into dollars, obviously, you shouldn't think about it as a percent of revenue, but more in dollars, it's a little bit of an increase into the next several quarters as we did the easier conversions first, and they are the next conversions will be a little bit more expensive on a dollar basis to deliver that same kind of growth. Nothing dramatic, but just as you're modeling things, I would put in a few -- a little bit more higher CapEx per -- as we transition, but not a change in our philosophy, as David mentioned earlier.

The next question comes from Vijay Rakesh with Mizuho.

David and Luis, phenomenal set of results here. Just a quick question on the -- the RPO and the financial guarantees that I see your data center is already at $1.5 billion. So that's like an annualized $6 billion run rate on the data center side. Is all the $11 billion and $42 billion RPO mostly all in data center, is that a fair assumption? And on to the pricing question on these guarantees is probably mark to market, you would assume, right, as you look out 2, 3 years, there has to be some benchmark into the market. Is that fair? I have a follow-up.

Yes. So we're not disclosing the customers. So I'll leave it at that. But if you look at the $42 billion, a lot of companies would be able to do that. So that's our RPO and that's the minimum contractual revenue that you would expect from the 3 deals that we signed before the end of the quarter, right? So if you include the other 5, that would be a larger number, and you will see that number in our next quarter, but it's not part of the $42 billion. If you think about the $11 billion, which I think is the second part of your question. There are different financial instruments that we're using to protect us. There is a portion that is in prepayments. And as I mentioned in my prepared remarks, you will see that in the 10-Q that's somewhere around $400 million that you will see in our balance sheet. And there are other financial instruments. And that's probably as far as I can go, which are managed by third-party financial institutions that are triggered if there is a breach in the contract that doesn't go all the way to the end. So that's kind of how it works, Vijay.

David, on the -- as you look at your NAND SSD roadmap, you have a pretty disruptive technology coming down the pipe in terms of high bandwidth flash. Any thoughts on how that's progressing, how that's going? If you can give us some color.

Yes. We're happy with how it's going. It's kind of steady as she goes. We're having conversations with customers on how they would deploy it or building the technology, the NAND die itself, the controller. So we're still on the time line we talked about earlier of having actual the NAND late this year and look for more of a system with the controller early mid next year.

Next question comes from Blayne Curtis with Jefferies.

Great results. Maybe just following on that, I had a question about you're hearing a lot more discussion about different memory tierings, maybe accelerators using more DRAM. I'm just kind of curious if you had any perspective, has a previous question on KV cache. I think everybody just assume there'll be a lot more hard SSDs to serve that. But I'm just kind of curious, as you look at that future roadmap, in different memory tiering, high bandwidth flash fits into that. I'm just kind of curious if there's any change over the last quarter or so on just the thoughts of where that storage will be?

I don't -- not really -- I mean, I think the kind of tiering architecture that came out maybe a quarter ago is what's being deployed. High bandwidth flash is, again, it's not a necessarily a substitute for an enterprise SSD or something like that, it's a way to bring a lot more density to inference in a little different way. We have to follow up on that with a little more detail. But look, we're seeing -- you can see it in our numbers. We're just seeing an enormous amount of pull on that portfolio of high-performance enterprise SSDs as these architectures get deployed and inference starts to get deployed at scale. As we said in the prepared remarks, I mean NAND is just a big part of that architecture now given the size of models, the size of KV cache, the context length, all these things is NAND is the most scalable semiconductor technology in the world. It's now front and center -- well, it's now a critical component of that architecture. And we're seeing that be pulled through. And I still expect this will be refined as we go forward. Again, this is why -- again, why we're staying very close to our customers. Our customers are the ones that know -- the big customers that are deploying this at scale. There -- I suspect they have -- well, I know they have very, very detailed insight into how they're going to scale this across a global franchise. And so understanding what that is and what that means for demand on our products, I think that there is an enormous amount of work going on there. And again, that's what's driving the demand signals that are years into the future for us and allowing us to align our business model around that demand.

The next question comes from Amit Daryanani with Evercore ISI.

This is Victor Santiago on for Amit. I wanted to ask about the Nanya investment and supply agreement last month. Could you help us better understand the strategic rationale for it? Is that primarily the secure DRAM or access to HBF or memory products going forward?

Yes. As you have seen, our data center business is doing pretty well, and we just posted very interesting growth, and that's just with our TLC product and a lot more to come as we continue to expand our business now with QLC and as we drive growth through our new business models. And one of the key things that we need to have access is DRAM, the partnership with Nanya provides an investment into the company and gives us a preferential treatment of access to DRAM as well. So that's the rationale.

Got it. And as a follow-up, I believe you provided the duration on the longest contract that you signed, but could you give us any idea on what the average duration might be across the 5 deals you signed so far?

Yes. Frankly, we're not -- we cannot get into that level of detail. We're giving -- I think you know the minimums we signed before were more shorter, but we're not giving an average. I apologize for that, Victor.

I think, Victor, one thing, we want a portfolio of deals like 1 year, 2 year, 3 year, 5 year, so that they don't all end at the same time and all those kinds of things, and we don't face cliffs. I mean we have every expectation we'll renew some of these. Some of the deals are for a certain number of years with options for more years. So it's still a little bit early. But we're going about this where it's not just -- there's not a fixed template, right? It depends on the customer. Every deal is kind of customized to specific customers. And as we get a little deeper into this, maybe we can have some of the conversations and some of the numbers you're asking for. But right now, it's a little bit early for that.

The next question comes from Wamsi Mohan with Bank of America.

It's Ruplu filling in for Wamsi. I just had two quick ones, one for Luis. On the long-term agreements, is there any restriction on when you can raise prices? Is it allowing for annual price increases? Or are there certain conditions when you can raise prices? And then one for Dave. How do you see the interest in QLC flash trending? And how do you see the mix of TLC versus QLC trending over the next couple of quarters?

Yes. So sorry, I'm going to frustrate you a little bit. We can't go into pricing indeed for each of the contracts. As I said at the beginning, there are some fixed price components and some variable pricing components, and it is very different depending on each of the agreements that we have. So there is no kind of an overall answer overall on pricing.

So TLC, QLC. I mean, if you look across the whole portfolio, roughly 2/3 TLC, 1/3 QLC. If you looked at data center, obviously, for us, it's predominantly TLC and we'll be launching major QLC products next quarter. But there's a lot of demand for TLC, given these performance -- especially in the enterprise SSD space. Given the inference architectures and some of the comments earlier around KV cache and how important it is and quite frankly, how it can scale dramatically based on your assumptions of the use case you're serving. There's a very, very strong demand on TLC. But that said, we expect our QLC products to do very well.

This concludes our question-and-answer session. I would like to turn the conference back over to Ivan Donaldson for any closing remarks.

Yes. I just want to say thank you, everyone, for joining the call today. Thank you for your support, and we look forward to speaking with you throughout the quarter.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.

Well, excellent. Good afternoon. My name is C.J. Muse. I cover semiconductor, semiconductor equipment with Cantor Fitzgerald. Very pleased to have the team from Sandisk, David Goeckeler, CEO; and Luis Visoso, CFO. Welcome. Good to have you both here.

C.J., great to be here. Thanks for having us.

I think, Luis, you have something you'd like to read.

Yes. Hi, everyone. We'll be making forward-looking statements in today's discussions based on management's current assumptions and expectations, including with respect to our technology and product portfolio, our business plans and performance, market trends and opportunities and our future financial results.

These forward-looking statements are subject to risks and uncertainties. We assume no obligation to update these statements. Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations. We will also be making reference to non-GAAP financials and a reconciliation of our GAAP to non-GAAP financials can be found on our website.

Perfect.

Thank you.

So Dave, congrats on a 1-year plus anniversary.

Thank you.

The world has seismically shifted for NAND in the last 8 months. And so essentially...

Changing, yes.

Inflecting since August of '25. What has changed?

Look, I mean, I think there's a lot that's changed. And I think what you see and there's a lot of discussion of pricing and all that. But what you see is, I think, kind of a second order of a major change that's going on in the industry, and that is a couple of big things. One is, this year, data centers will be the biggest buyers of NAND, brings kind of a new set of buyers with a new use case, a new way they use the technology, a new kind of financial model of the way they're deploying the technology or what it's supporting.

And their use cases, I think, are the furthest thing from a commodity, if you will. People want to kind of put that tag on us and customers want 3 options, 4 options that they can swap in and out. But when you're playing in the data center, you're talking like a year-long qualification cycle. So this is a very, very different product, very, very different use case.

And I think that's driving just a fundamental change in the industry over a short period of time. And I think we thought a year ago, when we were launching the company, we said we thought the market was going to be undersupplied through the end of '26. We heard a lot of feedback on that. I think as late as last summer, people were end of the summer, there was a lot of talk that December pricing would be down and all this kind of stuff, and it didn't quite play out that way.

So -- and to me, that's really a follow-on of what happened in '23. And '23 wasn't just a downturn, if you will. In my book, '23 showed that the way you manage these franchises just have to change. This idea that when the economics get stressed, you can release a new node that lowers your cost, the market is infinitely elastic, it can absorb that supply. I think that way of managing these franchises just ended spectacularly in that really, really kind of meltdown of the industry.

And out of that now has arisen an industry -- I'll just talk about our company. The way we've come back from that is just very different. The way we think about managing the business, we're much more proactive on managing supply. We're much more willing to pull back supply, if we don't see the demand for our products or the economics don't hold up. You put that on top of the buyer has fundamentally changed. The economics of that buyer is very different. The way the product is used is very different.

You add all this up, and that amount of change is being compressed in a couple of quarters. And it looks like this huge, huge transformation or all this turmoil, but it's really the product of all of these structural changes in the industry from the supply side and the demand side kind of all coming together at once.

Yes. I think there are certain investors surprised that NAND is the slowest growing semi equipment industry this year, but I think they neglect to think about the fact that we just went through a 7-year downturn. And so I think the hope was that NAND would consolidate. But given that backdrop of the multiyear downturn plus, as you highlighted, a changing kind of buyer landscape, how, I guess, do you see -- and maybe this is a broader comment beyond Sandisk, the overall industry acting?

I like to spend most of my time talking about our company as opposed to the industry. I'm not -- I only run one company or the 2 of us. But look, I don't think I've ever believed this theory that you have to drive consolidation to get the economics right. I think that it helps certainly. But -- and also in any kind of fixed cost industry like we are, it also -- scale always matters. We get that through the JV, which we think gives us a lot of advantages. But I think there's never been a doubt in my mind that the quality of the product we deliver and the value of the product we deliver is extremely high.

Consumers pay $0.40 to $0.80 a gigabyte for NAND continuously when they buy devices. That's kind of the going price. And so the value has always been there. The question is, can you get the business model right. And I think we're seeing that all happen right now. And it's not -- consolidation is not the answer to that. I mean it maybe it helps in some regard, but what really matters is just running the business in a way, where you get a fair return for the amount of intellectual capital we put into this, enormous amount of intellectual capital, intellectual IP. And on top of that, an enormous amount of CapEx to manufacture. We do both.

I mean there's very few industries where you have the level of intellectual intensity that we have, R&D intensity, 75% of our OpEx is R&D. And on top of that, we're investing billions of dollars to produce. So we're doing all of that in one place. And I think the market is now recognizing -- I think it's fairly clear that what that -- the output of that is very valuable to the world.

Makes sense. I guess maybe moving to customer engagement and how that is evolving given today's shortage environment, how are you seeing your customer engagement changing? Is it evolving particularly just data center? Or are you seeing in other end markets?

Yes. We're talking to customers across all our end segments, and they're very interested in NAND, and we happen to have NAND. So that's a great place to be. We're trying to shift from the old business model to a new business model. And this new business model is one where our customers are secured supply and where we are secure structurally sound financials. It's particularly coming from our data center customers because their willingness and ability to pay is much higher than what we've seen from other customers. We're open to business with anyone.

And I think you've -- I don't know if you've called it a 3-month circus or 3-month auction.

I'm didn't call it a circus. I called it a bizarre. All the world's NAND is auctioned every quarter. I mean that's the business model. It's kind of -- I've never look, I come to this from running a lot of different technology franchises. I come to this one. And it's like, wow, this is an interesting way to do this. But that's what we do. And so we know how to run that business model. It just -- it leads to this kind of like somebody is unhappy, like either the suppliers don't have very good economics, which people remind -- a year ago, when I took this job, everybody reminded me about like what a bad decision that was.

Or on the other side, you have the situation you have now where everybody is saying, well, we can't get what we need. And it's like this is not a hard problem to solve. I mean we know like businesses solve this problem all the time. And the question is just are people willing to do that? And to Luis' point, and why I said earlier, we have a new buyer that's now more than half the market.

I think that buyer is less interested in that business model. right? They're more interested in what Luis said, I want supply. I look down the road, here's what I need not next quarter or this year. Here's what I need in '27, '28, '29, '30 and coming to us and saying, can you step up to deliver this to me?

And we basically say, of course, we can because we have this huge fab in Japan, and we have all this wonderful technology. But if you want us to assure that you're going to get that supply, we need to come up with a different business model about how we do business together. And I think they say, fine, let's have a conversation. And we're having those conversations.

And I know it's early, but what does that framework look like to you?

Yes. As you said, it's early. We signed one of these new business models, and there are several in process. And the way we think about it is there are 4 dimensions, right? Time, are you looking for a 1-, 2-, 3-, 5-year deal. And that's very important to us because then we have certainty of financials and they have this certainty of supply. What is the amount of exabytes that you are willing to consume? And we're betting on customers that are winning customers that need more demand every single year because that's better for us because our fab is very, very efficient, and we produce more wafers every single day.

The third element is the price, right? And we want a price that is attractive for us in any scenario in good times, in bad times. That could be a combination of fixed pricing and variable pricing, where we can capture more upside if prices go up and our customers can be protected if prices go down. But in any scenario, we are protected for strong financials.

And the last thing, which is an element we are not talking too much about, but we're working on, which is how do we ensure that our customers would be there until the end of the contract, right? So we're making a commitment, right, on both sides, and we have to live until the very end of that contract.

I wish you luck with those negotiations.

It's fun.

You need someone in the room. I'd...

We'll let you know.

I appreciate it.

Maybe moving to your data center business. I know you're very excited about the traction you're making here. I would love to hear what's the time line to hit that kind of 20% goal?

So look, we focus on like just improving quarter-over-quarter, and we're very much on the right track at this point. This is a tough market to crack. What I said earlier, you don't just show up with a product. You got to go through a very long process to get into very big customers. We're making very significant progress. First quarter, I think we had 29% sequential growth in data center. Last quarter, it was mid-60s sequential growth in data center, and we said we think we'll see that accelerate throughout the year.

So I think this year is going to be the year, where we really, really establish our position in this market in a very serious way. I'm not going to pick a particular number out there because I don't want to like hairpin us around that. We'll see where we go. But I will say it really is the progress in the data center market that's catalyzing this conversation that we talked about earlier, where it's when we get through that qualification with a big customer and they come and say, okay, we love your product. It's a great product. Here's what we need for the next 3 or 5 years. And we look at those numbers and go, as I said, that's a big number.

Like if you just want to show up every quarter and get that supply, my personal view is that's fairly low likelihood that's going to happen. And as we've talked about in the past, I think we're finding that customers prefer supply assurance over price. It's not that nobody is going to play an infinite price, but that supply assurance is very valuable to our customers. And it's these products that are world-class products.

Again, it's a combination of what we have with BiCS8, right, gold standard in the industry right now, wafer bonding, performance, power efficiency, QLC performance, really, really strong, 2-terabit die on top of that, right? That makes -- you can obviously -- you need half as many die as a 1 terabit die to build a high-capacity enterprise SSD. And then the controller on top of that, you wrap that all up, it's a fantastic product.

And the market is responding to that, and it really is unlocking these conversations about, "hey, we're going to base our data center on this product. We need to make sure we can get it for a long period of time." So we feel very, very good about where we're at. But again, we'll see over the next several quarters as the numbers come out.

Makes sense. I know we're early in kind of the life cycle for KV cache, but obviously, Jensen put out a very bullish presentation at GTC Washington. And this morning, we had an equipment company that gave their best guess on KV cache, which was for every 2 million of accelerator units, it would drive 1 point growth in NAND bit demand. I know you haven't put out a number there. I'm not kind of asking you. But I guess, as you've done your initial work, what are your thoughts on how that might change the demand curve and how that might change the industry perhaps having to add incrementally more supply?

I think -- look, I think everybody -- AI is an incredible thing. I mean I don't need to tell anybody hear that, right? And I think that we're now moving into this massive scaling stage of this technology. And when you're going to scale something, you got to really think about the economics and kind of how this is all going to work. Models are getting bigger, caches are getting bigger, context lengths are getting bigger. And we believe this for a long time that at some point, this architecture is going to have to include the most scalable semiconductor technology, and that's NAND.

If you need to store big things, you need to store a lot of it, you have to eventually get the NAND. And that's nothing against what's going on with HBM. I think it's brilliant technology. I think it's incredible innovation. And we're not a replacement for that at any stretch of the imagination, but there's only so much capacity there. And I think customers are trying to figure this out. I think that culminated in what you talked about at what NVIDIA had to say. And there's no doubt that depending on how this plays out, this will be a major driver in the next several years of NAND.

It's not in our numbers yet. Our initial assessment of this was potentially 75 to 100 exabytes next year. I've seen several people trying to triangulate on it. I haven't heard this one yet from this morning. But all these numbers are kind of coming out the same, but it depends on how the customers are going to go deploy this. But we see this, obviously, is very positive. We kind of said it a little different way last year. We talked about HBF, which is not the same issue, but we -- it was the same kind of strategic statement.

NAND is going to come into this architecture because of the density we can deliver. And it's good to see that being recognized. If I had to pick one -- there's a lot has happened in the first year of our company, right? A lot of things have happened. But the -- where we've gone on HBF from where we were a year ago, when we first even introduced the term to where we are today is a huge progression in I think, the acceptance that this type of technology is required in this architecture. And we're very excited about that.

So maybe sticking with HBF, high-bandwidth flash for the inference market. Could you maybe talk a little bit more about where you are in terms of technologically development and your thoughts around commercialization?

Yes. So this -- a little bit of history. This came out of some -- we have a lot of very clever people in the company. As I said earlier, we're a high R&D intensity company. 75% of our OpEx goes into R&D. Those people -- those individuals in the company have been building NAND for 20, 25 years. It's their life's work. And if you're a NAND designer for your whole life, you've basically been focused on how do you increase density, give me more bits. That's what I'm trying to do. You went 2D, 3D, keep going, change memory hole density, all these kinds of things that people just go on and on, and it's been spectacularly successful.

And I think it was very clever a number -- many years ago now, our R&D team said, well, what if we focused on bandwidth instead of density? Like what if -- because that's not what we're -- we've been told our whole life was give me more density. Now it's like, okay, we got density, good, got that, continue to do that. But now think about could you get more bandwidth out of this technology? Could you get more endurance out of this technology?

And it turns out, you get a bunch of very smart people that have done a lot of seminal work in NAND design. They come up with good ideas to this, and that's kind of where HBF was born. And so we're on the time line that we talked about last year. We expect to have a die, the die itself later this year. And maybe a year from now, we'll have what we would call the device, which is the die plus the controller that we can put in customers' hands and they can play around with it and see how it fits in their architecture.

But the real trick is here is this is not a plug-compatible replacement for anything else in the architecture. This is a new architectural element. So it's got to fit into the architecture. It can't -- we got to work with partners on how they're going to build their -- either their device, can go on a device, can go in the cloud. We got to work with them on how are they going to architect it, how are they going to scale inference in their cloud?

How could this technology integrate with what they're building. Essentially, it's 2 gears, you got to get the mesh. Those conversations have been happening for quite some time, and we're making progress on it. We feel good about it. And we think it's going to be a very important technology in the future.

Maybe sticking with your technology road map. I know you're very excited about BiCS8. What are some of the advances inside there, such as CBA that's driving leadership and confidence you have vis-a-vis the competition?

The JV is really a spectacular partnership. And what it allows us to do, it allows us to -- essentially, a lot of people in the JV, they focus on manufacturing side of it, which is very important. Don't get me wrong. I mean if you go to Yokkaichi or Kitakami, they're like incredible places, and it's a huge amount of work to run that.

We also work together on R&D. So our 2 teams are like 1 team. And you put our 2 teams together, and we're the largest or tied for the largest producer in the market. So that means we can invest more engineers than anybody else. And that's a very important -- when you build products for a living, which is kind of what I've made my whole life doing, it's like when you can invest more people in this, especially over a long period of time, you end up with this accumulation of R&D that is really, really strong.

And that's kind of where we've ended up after 25 years. And that spans a whole bunch of different things, the actual cell we have that actually stores the bit, how that is scaled in the XY dimension, how it's scaled in the -- how it's layered, the memory hole density, materials, all different kinds of things, right? So BiCS8, what does BiCS8 add to all of this?

Well, now it was this whole concept of wafer bonding, which is it used to be -- you build the CMOS and then you start building the NAND stack on top of it. And by the time you were done, you had degraded the CMOS somewhat, right? Because you have different kind of processes, maybe different temperatures, things that are probably beyond my technical acumen.

But at this point now, the team decided -- when they first came to me and told me what they were doing like 4 years ago or 5 years ago, I'm like, are you sure this is going to work? I mean it's like we're going to take -- we're going to build the NAND stack on one wafer. We're going to build the CMOS and then we're going to flip one over and we're going to bond them together. And it turns out, when you do that, the CMOS is like pristine. It's like -- so you get like really fast interfaces.

And you couple that with all this accumulated R&D on QLC performance and then you put on top of it our system-level ability to kind of reach into the NAND and do what we need to do to optimize things, hardware accelerators, you end up with an awesome product. And we're now ramping into that. So back to the data center, I'll tie this back to what you asked earlier, our data center position, why we feel so good about this is like we have all these things arriving at the same time.

We've got this world-class NAND node that gives you really, really strong performance, gives you really strong QLC performance, really great power efficiency, right when the market for data center is exploding for people that -- in a use case that find those things very, very attractive, right? And so -- and then on top of that, we've got kind of this clean sheet architecture for controller. And I think it positions us extremely well.

Maybe, Luis, a couple of financial questions. And I know you're not going to answer this one, but -- I've got you hitting 70%, 75% gross margins in the next handful of quarters. And so if I go back to the earlier part of our conversation, you're spending on R&D and you're spending on CapEx. So you're double stacking cost where fabless foundry. Obviously, they only do one part each. And so what is kind of a normalized kind of gross margin for Sandisk?

Yes. We're targeting to get to an attractive gross margin. And there is no ceiling. We believe that our products are very valuable to our customers, and they're paying a fair price for what they're getting, and we continue to price for them. But we think we need to get an attractive gross margin for them.

How attractive is attractive? and then maybe capital returns. Remarkably, you exited December with nearly $1 billion in net cash. You're going to generate significant free cash flow in the quarters ahead. What kind of is the framework in your mind in terms of sort of the net cash you'd like to have on the balance sheet? And then how do you think about capital returns thereafter?

Yes. At a high level, what you've seen us do and we'll continue to do is to invest in the business. Last quarter, we announced the expansion of our JV contract, which requires some cash or will require some cash over time, which we're very happy with, and we'll continue to strengthen our supply chain that is important to us. And other than that, we'll build good, healthy reserves that are prudent given the industry -- the history of the industry. We don't think we're going to go back to the history, but we want to be very prudent and build some reserves. You saw our TLB come down. We closed last quarter with $650 million from $2 billion that we started.

And as you can assume that we'll continue to reduce that over time. And then the obvious thing to do is we'll consider returning some cash to our shareholders. When and how exactly we'll announce at a future date, but we know that, that would be a natural next steps with the cash we're generating.

So maybe CapEx and supply. $12 billion is a lot of money, as you highlighted to me last night for WFE in 2026.

I don't know, $1 billion, is a lot of money.

However, it is a much smaller growth than every other kind of end market in equipment. And we clearly are in undersupply. So curious what kind of gets the industry to push ahead. And I think everyone is adding layer counts, but obviously, that reduces kind of wafer output given the greater complexity. When do you think the industry will start thinking about meaningful greenfield?

Okay. So first of all, I'm not going to speak for the industry. I'm just saying this, but I think this is a super interesting conversation. Remember, it was only a year ago, right? It was only a year ago, we got on stage and we said we're going to launch this company, and we're going to invest for mid- to high teens bit growth. And everybody told us too much, right? Pricing is going to go down, right? That was the prevailing wisdom until the end of last summer.

In fact, maybe until October. And we think of -- like we just invested an extra $1.1 billion for fab space from 2030 to 2035. So it gives you the idea that we're thinking about investing and being in this business in the long term. So now here we are first week of March and the question is, when are you going to invest more? And it's like, look, I think that we just we are going to invest for mid- to high-teens bit growth, and we are going to stay very close to our customers.

As Luis said earlier, customers want to buy NAND. We want to produce NAND. And we're going to produce NAND and we're going to produce more NAND tomorrow than we did yesterday. And what we need to do is get the alignment of that intention to buy that NAND with the actual production of it. And that's not a trivial thing. It's going back to these -- how are we going to get this business model right with our customers. It's not a natural thing for them to think the way we do, right?

If you want to consume something for 3, 4, 5 years, you can't take 2 quarters off in the middle, like we can't just turn the fab all. So we got to figure out a business model. So when we produce, we know it's going to be consumed. And as we get that figured out, I think the smoke will clear on and all the issues about where is the attractive demand in the market, not where is any demand? Like again, I will debate your term undersupplied. There's a market. Supply and demand are always in balance for that market. If you lower the price of any product, you will increase demand. We all learn that and you don't have to be an economics professor to know that.

So the question is, what is the sustained demand for the attractive markets we want to play in. And we're going to figure that out over the next -- I think we'll figure that out over the next several quarters, and then we can come back and revisit this conversation. But the idea that we're going to put more capital in the business, hoping that people show up every quarter to buy it, -- that's -- again, that's not a model that we're very interested in investing behind that model. We're putting billions of dollars behind that model. We're putting hundreds of millions of dollars of R&D behind that model.

Now we got to just get aligned with the demand side and get that economic model kind of a little more solid for a longer period of time and not -- when I have these meetings today, like everybody, when is it going to end? When is it going to end? That's the predominant question. When is it going to end? It's the cycle. When is the cycle going to end? When does the next shoe drop? Why are you trading at a 6 multiple or something? It's because people don't believe in the sustainability of the model.

And it's very difficult to say we're already putting billions of dollars into this market. We should put more money into this market, when the people investing in our company don't believe in the sustainability of it, now we believe in the sustainability of it, but we need to prove it, and that's what we intend to do. And as we do that, we'll continue this conversation.

So my takeaway is maintain mid- to high teens and make sure you have the appropriate economic model before you would entertain believing in perhaps a sustainably higher growth market?

Yes. I mean that's one way to say it. I mean I think the other way to say it is if somebody wants to buy NAND and they're willing to make a commitment to buy it, hey, we're good at that, right? We know how to do that. We know how to produce. There's no issue there. We've got the R&D lined up. We've got new BiCS nodes we're working on well beyond BiCS8. We can increase the productivity of NAND -- that's not the issue. The issue is not our willingness or ability to produce NAND. The issue is the economics of increasing investment in a market where the business practices don't support it.

Well, I think we've run out of time. Thank you both very much. Appreciate it.

Thank you. Appreciate it, C.J. Thanks for having us.

Thank you.

All right. Welcome back. I'm Joe Moore, Morgan Stanley semiconductor research team. Very happy to have with us the executive team from Sandisk, CEO, Dave Goeckeler; and EVP, CFO, Luis Visoso, Thank you, guys.

Joe, thanks for having us. We appreciate it.

Thank you. And I think, Luis, you want to read a safe harbor first.

Yes. We will be making forward-looking statements in today's discussion based on management's current assumptions and expectations including with respect to our technology and product portfolio, our business plans and performance, market trends and opportunities and our future financial results. These forward-looking statements are subject to risks and uncertainties. We assume no obligation to update these statements. Please refer to our annual report on Form 10-K and other filings with the SEC for more information on the risks and uncertainties that could cause actual results to differ materially from expectations.

We will also be making reference to non-GAAP financials and reconciliations to GAAP to non-GAAP financials can be found on our website.

Great. So you guys saw this coming, right?

Well, I mean, we saw [indiscernible].

I mean really, I'll give you credit. You said this time last year that we would have a stronger second half. You are right in a pretty big way. And I guess, just maybe kind of walk us through what did people miss? How did you guys -- you saw some of this improvement coming. Obviously, AI surpassed all of our expectations.

Yes. I mean we've had conviction for a while that the market I think we've been saying for quite some time that we thought the market would be undersupplied through the end of '26, and that was as far as we were doing modeling time. And we were pretty clear about that last year when we launched the company.

How do we come to that position? I mean we do, obviously, a lot of bottoms-up research ourselves. We have a lot of insights into the market because we play in them all the time. So we do models on demand on all the big 3 markets on smartphones, PCs, data center. We obviously do a lot of modeling of supply. Again, we have some unique insights. We have some people on our team that happen to know a lot about how you design NAND and what different people's road maps may look like and what supply is going to look like. And we would just consistently look at that picture internally, and we would just see that we thought coming out of the '23, which was kind of an industry-defining event that this market was going to get to a very different place.

We have a lot of conviction that the way you have to manage the business is different. You can't just release new nodes because they're available and think they're going to drive your cross-sell and the market will just absorb all that supply. We thought that playbook has completely changed. And that turned out to be true.

And then as we move through the second half of last year, we saw the data center keep raising CapEx spend, I think, which was very well understood. And again, we have our own models of how we translate that into calendar year '26 exabyte demand. And so it's just we went through 3 -- 2 forecast cycles. We went into -- let's say, this time last year late last year, we thought data center would grow mid-20s. Next forecast cycle, we thought it was going to grow mid-40s, and then we came out last -- our last earnings call, we talked about, now we see it growing mid- to high 60s. So yes, I mean, again, we just -- we do a lot of modeling bottoms up on both sides of the equation and take a clear eye to look at it and see where we think the market is going to go.

I mean the data center storage market has been really surprising. I guess we saw huge orders in the fourth quarter for immediate delivery. How much of that do you think is because of hard drives? And I guess, I don't feel like it's a hardware replacement because I think it maybe accelerated some of the solid state that might have happened later, but the overall storage, how does that? What role does that play?

Yes, I'm not a believer in the hard drive substitution question. I mean it's not lost on me that 2 years ago...

You now a little bit about the hard drive.

Well, a little bit. 2.5 years ago, I was answering all these questions. no hard drives are going to be sold in 5 years, right? And we've pivoted 2 years later to 1 million questions about there's not enough hard drives in the world. So I think as a NAND industry, we just need to leave hard drives behind. NAND is a different market. Our drive is great. I love that business too. I know a little bit about it. Great business. Great technology, NAND is the same way, great business, great technology.

I think I've said this very consistently for the last 6 years. These are complementary technologies in the data center. In the device, they're substitutes. But in the data center, they're complementary, and NAND is -- they're both going to grow. NAND is going to grow faster than hard drives. And I think that is borne out to be true. And I think with AI, NAND is now being fully integrated into the AI architecture. That's something that we thought was going to happen years ago. We didn't think it was going to happen years ago, but we were looking at the technology, NAND is the most scalable semiconductor technology out there. When you have an architecture that's scaling like AI is, NAND is going to come into the architecture at some point. That is happening. And that, of course, is accelerating the data center growth rate.

And then on consumer, you've seen shortages as well, and that surprised me a little bit. Is that just because there's demand coming from the enterprise side? Is it just -- what's driving that thing?

Yes. I mean I think shortage is an interesting term. I mean we just have a market, a market result -- markets have supply and they meet every day, right? And that's rationalized through pricing. And as that gets rationalized, if your business model is dependent on buying NAND from 1 supplier and turning it into something else, as your costs go up, it's going to make your business model less attractive. And I think it's natural, you're going to see business models that depend on readily available inexpensive NAND are going to be under pressure. And I think this is kind of the -- I think, still a somewhat misunderstood part of the NAND market. The NAND buyer is just fundamentally changing. This used to be a market that was more of a commodity market. The buyer wanted multiple suppliers for every point in their deployment. And that's being replaced by a buyer now that is the predominant part of the market. The largest part of the market is now a buyer that is not -- their business model is not dependent on the volume of NAND they buy. They're not reselling NAND in a different form factor. Their NAND is part of a much larger architecture that is very profitable. And their consumption continues to go up. And I think that is what's catalyzing a change in the business practices in the market. And I think that's really what's going on. And we're in the middle of that.

And when you're in the middle of something like that, it's sometimes confusing, and it's hard to figure out what that's actually going to happen, but I think it's just incredibly exciting. And it is this technology that we've been building for 25 years, it's extraordinarily important and the fundamental dynamics of the way the market works are changing. And it's going to be fun to see how it plays out.

Yes. It hasn't run so far for sure. From the standpoint of capital spending in NAND, we've actually seen NAND capital equipment sales dropped the last 2 quarters as the market's gotten really good. And I think some of that is just timing, kind of random timing stuff, but there's also a clean room space that people are allocating to DRAM that's maybe taking away? Like what do you think is the supply picture here? And what does it take for you guys to resume kind of a higher level of spending?

Well, I mean -- so first of all, I think it's important to realize we are spending more money all the time to increase supply. So we've been very transparent. We're going to grow supply mid- to high teens. We're putting billions of dollars behind that. We're putting hundreds of millions of dollars of R&D behind building the best NAND technology in the world. So we are going to grow the market and we grow the market every single day. And we think that, that is a good long-term growth rate for the market. And we're trying to sort out as this market changes, where is the attractive demand in that market. And I think that's happening every single day.

It would take a lot to get more conviction. We're not interested in what the growth rate of the market is next year or the year after. We're interested in the growth rate of the market a sustained basis for the next decade. I don't think it should be lost on anybody that we just spent -- we just invested over $1 billion to make sure we have supply from 2030 to 2035. That's the kind of time horizon that we're thinking about, and we continue to invest more in the business to make sure we can continue to grow this business.

I think as these business practices get worked out over the next year or so, maybe faster than that, we'll find out what the real growth rate is of the attractive demand in the market.

Okay. And I guess one of your push is really even predating the spin has been that customers need to give you more visibility into what their demand is, more transparency to the long-term view. This tightness in the market, does that help you achieve that? And I guess maybe we can go into some of the long-term agreements and things like that, that are giving you that visibility now.