QUALCOMM Aktie

Okay. Good morning, everybody. Welcome to the Goldman Sachs Communacopia and Technology Conference. My name is Jim Schneider. I'm the semiconductor analyst here at Goldman, and it's my pleasure to welcome you to our first session of the day. We're very happy to have Qualcomm, and CFO and COO, Akash Palkhiwala, with us today.

Welcome, Akash.

Thank you. Thank you for having me here.

Akash, I think at the highest level, Qualcomm is working pretty hard to diversify beyond smartphones over a number of years. I want to get your latest thoughts on the progress starting with data center, but you had a very exciting announcement this morning. So I think a custom silicon agreement with Amazon, multiyear in nature and with some warrants attached to it. Maybe you can kind of unpack that for us, help us dimensionalize the opportunity both in terms of size, type of products and so on to the extent you can.

Sure, sure. So first of all, thank you for having me here. Very exciting deal for us. I think as you rightly noted, the company is transforming. We are -- used to be a smartphone -- primarily a smartphone company. As we look forward, we are very much a highly diversified company with, I'd say, three legs of the stool. We have smartphones, we have auto, IoT and data center. And we set pretty compelling targets for our data center business a couple of months ago at our Investor Day. And so very pleased to announce this strategic transaction with Amazon.

It's a first of many as we make progress towards what we laid out. So you should think of this as really a landmark deal for us in the data center business. It kicks us off there. It includes two components on the product side. So first is customized silicon business with them, multiple generations of that. We are also going to be working with them on optical connectivity solutions, starting with 1.6T and follow-on generations. And the transaction includes a warrant agreement that we've outlined in our 8-K that we filed today. But the way it's structured is we issue warrants against purchases of up to $60 billion from Amazon for data center products over the next 10 years. And there is an upfront vesting of about 15% of those warrants that's associated with the $60 billion, and that is tied to upfront commitments that are being made by Amazon. So the 15% really ties to what they're committing upfront.

If you kind of wind back and look at what we had committed at Investor Day, we said approximately $5 billion in revenue in fiscal '27. Fiscal '27 starts shortly for us, and this makes our number of very high confidence. It also positions us for fiscal '28 to deliver strong year-over-year growth from '27 to '28. We are also similarly proceeding with the other data center customer that we talked about at Investor Day.

So a lot of strong progress. We'll have -- start having revenue with Amazon starting the December quarter. And so we are already in production with them. Very excited about how this deal positions us to execute on our targets going forward.

We also did set a target of $15 billion in fiscal '29 and you should think of this agreement as one of the core components that allows us to execute on that.

Okay. Great. I want to come back to all those things. I'll let you unpack it a little bit more, but maybe zooming out again for a second. From a technology perspective, the company has long-held expertise in processors. And we're seeing a lot of technologies kind of become more relevant to winning formula in AI overall, specifically networking, storage and software. So maybe help investors understand or contextualize for us why these are or not the correct assets to have under the same corporate umbrella? And speak to maybe Qualcomm's right to win in the marketplace against competitors who maybe have a slightly more diversified technology portfolio in some of those areas.

From a data center perspective?

From a data center perspective, yes.

Yes. So let me maybe kind of step back, as you said, and outline the different components of our data center strategy. There are four parts to what we're trying to do. First is custom silicon engagement. And so obviously, in addition to Amazon, we've talked about one other global hyperscaler that we're engaged with. And I think it's very simple. We have a very strong portfolio of technologies, both from a compute and a connectivity perspective. We also do a lot of chips, especially in advanced nodes, tremendous expertise of high-performance, low-power, tremendous expertise of high-yield manufacturing. And so we're bringing all of those things together for our custom silicon engagement. So that's the first part of the business is custom silicon. There's a lot of opportunity in the market. We are obviously a new player, but we have all the technology scale to deliver on what's required there.

The second part of the business is AI accelerator. This is an area where the market is going through a transition. We used to have one uniform solution for all kinds of AI accelerator needs. And what we're seeing in the market is a disaggregation of compute, where there is certain kind of solutions, GPUs would be very good for certain things, XPUs would be very good for other things. As you break the workloads down into training and inference, as you break inference down into prefill and decode, there's a clear desire from hyperscalers to have customized solutions for various components.

One of the challenges in the current set of solutions that are available is memory bandwidth constrained solutions. And so Qualcomm has this innovative technology that we outlined at Investor Day. We call it High-Bandwidth Compute, which is really kind of stacking compute and memory together to deliver extremely high bandwidth that are customer -- that are perfect solution for certain kind of decode workloads within inference. And so we're very excited about the reaction we're getting from customers. And so there's more to come on that. But -- very excited about what that technology brings to the market, something very unique and that the market is looking for. So that's our custom -- sorry, our AI accelerator business.

The third is CPU, where we deliver what we think is the leading CPU at the edge in phones and PCs and automotive and industrial devices, and we're bringing that to the data center. And this is delivering high performance at low power, low cost, and that's what we are bringing to the table. And so we have an agreement with Meta as our first customer for silicon solutions in data center, and so excited about that.

And then finally, connectivity. I think we acquired Alphawave. They had SerDes that is integrated into a lot of these solutions. Even our custom silicon engagement includes Alphawave SerDes. And they have optical connectivity products that are also a part of the Amazon agreement. So I think those four areas, of course, we're a new player. But the fact that we come with so much expertise in chip manufacturing at scale, maybe the broadest technology portfolio on the computing side, and that's what we bring to the table.

Fair enough. Very helpful overview. So maybe also another high-level question. Clearly, the size of the AI opportunity, I think, is a lot -- much bigger than a lot of what investors thought about just a few years ago. One point of controversy, though, in the market has been sort of agentic AI and the scale of token demand that is going to generate. At the same time, I think everyone is also talking about constraints, power supply, other factors, land, power, shell that sort of can measure or limit the pace at which we can generate those tokens. So what's Qualcomm's view on all of this, in general? And at the end of the day, what's your view on whether AI computing stays confined to the data center or moves to the edge? And how you kind of position the company and the products around that?

Yes. Great question. So first of all, we don't think about it as this or this, right? If you think about kind of general compute, not AI compute, normal compute, it's always been distributed in the cloud and in the edge. And we think AI kind of plays the same way. There are certain things that make sense to run on the edge. There are certain things that would make sense to run on the cloud. And that is the way we expect things to evolve. Now with our kind of expanded strategy and presence in data center, we have an opportunity to play on both sides, and I'll address both individually.

To your specific question on data center, our view is you're going to see demand grow very, very significantly in terms of token. So one of my responsibilities is also running IT within Qualcomm. And so as you think about how our engineers use tokens to write software, to design chips, to support customers, we're seeing a massive growth in tokens usage within the company. And of course, that is going to be replicated across every enterprise in the world. And it drives a lot of efficiencies, a lot of time-to-market advantages in terms of being able to launch new products sooner. And so our view is there's going to be very, very significant demand growth for tokens, and there will be a need to satisfy that through different kinds of solutions.

Clearly, there is a constraint on power. There is a constraint on how many wafers are available and just memory is available. And I think some of those things play to our strength. Our strength has always been delivering performance per watt. And so in any power constrained environment, Qualcomm has something unique to offer. And so we're excited that we can bring that to data center.

The second thing I'll say is in terms of wafer scale, maybe Qualcomm is one of the top 5 players in the world. And so us being able to use our wafer scale as a strategic advantage in building our data center business is also very important. And so I think when you think about data center, we bring all those technologies I just outlined. But in addition to that, in a constrained environment, I think Qualcomm is in an advantaged position.

On the edge, we're seeing a very similar transition on the edge happening over the next couple of years as we're seeing in data center, which is build an AI accelerator that sits next to the main chip. And their job is to really run pervasive AI at very, very low power. And so we're building very similar to the HBC technology that we talked about in data center. We're building a similar compute plus memory stack as a coprocessor that would go into all edge devices. So whether, again, it's phones or cars or PCs or industrial devices, robotics, those are going to be areas where that technology will become relevant. And so we see data center as an opportunity. We see edge as an opportunity for AI.

Very good. So I want to come back to the target you talked about earlier from your Investor Day, $15 billion in fiscal '29 for data center revenue. Maybe talk about the underlying assumptions here? And are the existing programs, including the one you announced today, enough to sort of get to that target? Or do you need to announce other customers to come to market to hit that?

Yes. So let me kind of summarize the data points and then build to the $15 billion, right? So we talked about $5 billion in fiscal '27, very high confidence at this point. The year starts in about a month for us, less than a month. And we are already -- we have POs, we're building the chips. So that's very high confidence.

We also talked about strong growth going into '28 based on the engagements that we have. So it allows us to grow significantly beyond the $5 billion. In -- late in '28, going into '29, we'll have the Meta engagement with CPU come on top of that. And then we'll also have -- we've announced HUMAIN as an AI accelerator customer that will be deploying that as well.

So when you add those kind of four, five things that I outlined, you're already very significantly on your way towards $15 billion. And so it doesn't take us a lot more to get to the $15 billion. But don't just think about it like that. I'd say that's the numerical part of the answer. But you should really think about the engagement that we will have will be very broad-based because we're delivering technologies, as I outlined earlier, that the market needs that the customer needs. And so tremendous, I think, very well positioned to take advantage of it.

Fair enough. And then in the longer run, maybe you've talked about this kind of 5% of market share of $1 trillion TAM in all-in. Help us bridge between what you just said and how you get there? And how much is kind of tied to your assumptions around Arm's server CPU penetration versus x86 and maybe your share within the ecosystem more broadly?

Yes. So our longer-term forecast is not just tied to CPU engagement. Of course, it's CPU, AI accelerator, custom chip and connectivity, all four franchises that we're building. Specifically on CPU, when we signed our agreement with Meta, the size of the CPU market was a lot less than it's perceived to be today. And so we're very excited because we have a very strong engagement with one of the top hyperscalers. We are going to deliver what we think is the best performance per watt, performance per area solution in CPUs. And so we have a tremendous opportunity to grow beyond Meta.

The CPU TAM, obviously, has tripled since then. And so we're talking about over $200 billion a year now. And we see a very significant about half of that market transitioning over to Arm architecture where we will have performance leadership. So very well positioned there. But the largest part of the market, obviously, is the AI accelerator market, and we are tackling it two ways. We have the custom chip engagements that we have with various hyperscalers and then these are global hyperscalers, and we're working on their main AI accelerators. This is not kind of ancillary custom chip engagements.

And then the second thing I'll say is we're also -- we have our HBC technology that delivers this significant performance advantage. So overall, very confident that this is a strong portfolio that doesn't rely on one or two customers, doesn't rely on one or two products. This is multi-gen, multiproduct, multi-customer engagement. And that diversity is what gives us the confidence that the opportunity beyond '29 is much, much bigger.

And then just a couple of more on data centers before we close out that section. You just talked a minute ago about inference becoming more disaggregated. Distinct requirements for prefill, decode, other stages of the stack, and it kind of implies a little bit different architecture than we've seen historically. How do you -- how is Qualcomm viewing this shift strategically? And where do you think the biggest opportunities for you to participate are?

Yes. So I think, obviously, when you look at training and inference, we're better suited to inference. Within inference as prefill and decode, we are starting with a very keen focus on solving the memory bandwidth problem in decode. And so our technology stack, HBC is uniquely positioned to that. But that's not the end game. That's where we are starting.

I think our ability to kind of extend into everything inference is very strong. We have the technology assets to go do it. And because of the stacking technology and being able to deliver this performance at very, very low power, and to your earlier point, given the power constraints in data center, we have something that's unique and the customers see it. The customers see the need for disaggregated computing. And I think it just positions us very well in a transforming market where the market is transforming in our favor.

Yes. Very good. And then finally, software. Help us understand how the software ecosystem is changing. You bought Modular this year, announced a number of open source initiatives in general around that. Where do you think the market is going? What are some of the challenges you need to overcome to kind of go head-to-head with software incumbents like NVIDIA's CUDA product? And how is Modular helping you get there?

Yes. So software, obviously, is a very, very important problem statement. One of NVIDIA's advantages, obviously, is CUDA. And so what we were missing is a very modern stack that makes it extremely easy to port models onto our silicon, right? And so we met Chris Lattner and Modular, and they were on this mission of building a stack that is -- that disaggregates the silicon and works across silicon. And so once you port a model on top of the Modular stack, it works across whether it's NVIDIA silicon, AMD silicon or our silicon. And then they also support AWS and Apple. So it truly disaggregates the need for specific silicon tied to the stack. And then builds a modern stack that's very easy for developers to work on, but then also allows them to port once and work across different silicon platforms without compromising performance.

And so we just came across this incredible company that is just the right size, right culture for us that was building something that we were missing. And so that will become our stack going forward. We're also taking a very friendly approach to open source. We have -- making the kind of the lower layers of the stack, think of it as an Android-like approach, where we're making the stack available to everyone in the industry to deploy. And so very optimistic about what it brings both as a horizontal industry platform, but then something that cuts across everything that Qualcomm makes, whether it's an edge chip or a data center chip. And it makes our silicon very accessible to developers, which was the goal in the first place. So I think it was a missing piece that we have now resolved and it comes at the right time as we start scaling our data center products.

Excellent. Almost 20 minutes in. We have not talked about smartphones yet, so I want to go there.

Please.

You have a huge installed base across billions of devices and smartphones based on your core processing and wireless technologies. As we move to a world where kind of devices at the edge get more important, especially given the realities of AI we just talked about, what changes should we expect to see in smartphones over the next, say, 3 to 5 years, both in terms of processing, memory and so on?

Yes. So it's interesting. I think this is very much -- people think of it as a mature market, but it's a market that's very much in transition. And so let me go through kind of a few trends around smartphones. First is kind of building smartphones for agentic experiences. And so we've obviously been in this touch world that goes app by app. And there is a clear vision from a lot of hyperscaler players who are looking for endpoints for their clouds as to building something that is agentic first, that is voice first. And that's a device that where you just give work to an agent and the agent is figuring out across the apps or across cloud MCPs of how to take out that -- execute on the task for you.

And so very different device it is becoming. I think it's a given that the endpoint is that. The question is how long does it take and how the journey happens, but it's a device in transition from that perspective.

From a silicon perspective, it's a device in transition because of the disaggregation of AI compute and having this AI accelerator that sits next to the main chip and how it is going to increase the silicon content in a device.

The third thing I'll say is there's also this third device being added to the personal device ecosystem that is sometimes connected to the smartphones, sometimes a stand-alone device. And we think of it as a wearable device, a device that can see what you can see, a device that can hear what you can hear and a device that you can have an agentic AI conversation with.

So I'll say every hyperscaler we know of, every cloud company, model company we know of globally is in the process of building a device like that. And so whether it's the big hyperscalers you know in the U.S., the model companies you know in the U.S., the hyperscalers you know in China, every single OEM globally, they're all using Qualcomm chips to build a third device that adds to the smartphone. And so very excited about kind of all of those transitions happening at the same time. And we are obviously coming off of a low point in the smartphone cycle given the memory industry dynamics. So I think of smartphones as an area that will offer -- can only offer growth to us given all these vectors.

Yes. Tactical question to follow up on that. I want to sort of ask you sort of how do you think your customers are adapting to sort of the new normal in terms of memory prices and the unit dynamics that has on the market. Clearly, you've already talked about you think Android has bottomed as a market today. How are you thinking about sort of the forward in terms of smartphone unit growth for the broader market some of the catalysts that might accelerate growth from here?

Yes. So the catalysts are the ones that I just outlined, but maybe to start with the first part of your question. We've seen maybe volume go down by low double digits on smartphones, low teens over the last year. And the impact is largely on the lowest tier and the second lowest tier, everything below $300. That's where the largest impact has been because in a lot of cases, memory has now become more than half of the BOM of the phone. And obviously, it makes some of those device prices unsustainable. And so we're seeing an impact there.

If you look at the very top of the smartphone market, which is where Qualcomm has the most significant presence, we have actually not seen much of an impact. The prices clearly have gone up. But really kind of when you step back and look at the importance of the phone and the consumers' lives, it's the central device. It's the most important device, right? And so people are -- you're continuing to see people spend the money they need to spend to get the right phone. And that, I think, shows up in when you look at the market data as well. So we don't expect that to change. We think the low end of the market will continue to be under pressure. But again, I think the -- as we are -- as we stand today, we are at the low part of the cycle, and we have all these new vectors in place that improves the opportunity for us going forward.

Okay. So we went through a lot of things about sort of how AI is changing, move to the edge, also the impact of agentic. How does this apply to automobiles in your automotive business? How is AI changing the automotive architecture over time and sort of how you participate in that transition?

Yes. Incredible set of changes, obviously happening in automotive, and we've been at the center of it. I'd say, definitely our most successful diversification initiative. Next year, we expect to be the largest chip supplier to the automotive industry. Very, very kind of well positioned in terms of driving the transition, not just participating in it.

I'll say AI is very deeply integrated into automotive. It started with ADAS. Obviously, AI is the foundation for what happens in autonomous driving. And we are one of the two chip leaders on the autonomous driving side. But what is also happening now is AI is becoming central to the cockpit experience as well. So within the car, we're seeing this -- your hands are already busy on the wheel. There is -- it's difficult for you to go and work the buttons as you're operating the car. And so being able to talk to the car, I think if there are people in the audience who drive a Tesla, you can get very used to talking to Grok while you're driving, right?

And so it is just a great way to use voice, to use AI, to use voice AI really to kind of interact with the car and get information to ask it to do things. And so that is becoming very foundationally getting deployed in automotive. And we're seeing every OEM around the world as kind of the core feature that they're looking to deploy in kind of next-generation cars is around AI.

I'll -- if you allow me, I'll also say that the way we think about robotics is really for us an extension of the automotive opportunity. Because from a silicon perspective, those things look very similar. And in some ways, a robot is a car standing up. So we are extending that portfolio. Everything we're doing in automotive, those products to robotics as well. And AI is obviously foundational to that. So very excited about kind of this broader category of physical AI that starts with automotive, goes into industrial and robotics and us being able to take our automotive platform into those areas.

So you see that robotic opportunity is more or less coincident from a product standpoint for you?

A very large leverage of everything we've built from automotive, at least from a silicon perspective, the stack is different, obviously.

Okay. And then just in terms of content growth in that automotive market, what areas of that business do you see kind of driving the fastest content growth?

Yes. So tremendous content growth for us. What has happened is we were at our Generation 3 of products. We went to Gen 4. Now we are deploying -- beginning to deploy Gen 5. The compute content growth from Gen 3 to Gen 5 is 8x. So a lot more processing, a lot more AI. And we've also transitioned the business from selling chips to selling modules and SIPs. And so we're integrating memory. We're integrating passives, and we are delivering a complete solution to our customers.

And so there's a cost advantage to the customers. There's a performance advantage to the customers. There's obviously content growth for us that comes from that. And this sits on top of the fact that there's a demand for a lot more compute, there's a demand for a lot more AI. And all of those things are coming together.

And that's why when you look at our financials for automotive, we had set a target for $10 billion in revenue in '31, then we pulled it to '29. Now we have pulled it back further because we're getting to those numbers much faster than we thought a couple of years ago because of this trend for content growth and then we're gaining share across the board as well, right? Whether it's the Chinese OEMs or whether it's European or Japanese or Korean or American, we're gaining share across the board in the growing part of the automotive silicon TAM.

I want to talk a little bit about sort of smart glasses and wearables market, which you just brought up a minute ago. What is really needed for this category to move to sort of enthusiast level adoption to broad consumer adoption? And do you see the potential for these kind of smart glasses and other wearables to be used more broadly in things like the enterprise? Or is it more of a kind of a consumer-centric thing in your view?

Yes. So I think it's definitely consumer and enterprise. But the way the device is being designed, and maybe I'll broaden the category beyond smart glasses. It's a personal AI device that, as I said earlier, you can talk to it, it can see what you can see, it can hear what you can hear. And so you have certain OEMs focusing on glasses. You have others building watches, pendants, pins, dongles, all kinds of devices being built, but each device has the same premise is a battery-powered high-performance device that is an AI-first device that becomes the interaction point for a user.

The reason for glasses is a very reasonable form factor is, it's close to your eyes and close to your ears. And so it can see what you can see and hear what you can hear. And so that's a form factor that works very well. Outside of the consumer use cases, there's tremendous use case in enterprise. There's -- clearly, this is a platform that allows you to have lot of efficiencies if you're working on a machine, if you're working in a distribution center in retail, just being able to look at something and get information about it, there's tremendous value around it. So I think we're going to see broad deployment of it. Like automotive, our strategy is also transitioning here. We are transitioning from chips to delivering modules and SIPs.

Because form factor is so important that we are integrating everything into the smallest possible module that would be very difficult for someone else to do. And so when you combine processing, wireless connectivity, AI, along with memory, passives and you put it in the smallest possible module that is performing at very low power, only Qualcomm can do that. So I think this is a device that is for us to own. And as the market takes off, you'll see it come through in our financials.

Great. I think I'd be chastised, but you don't have the CFO and don't ask any financial question.

And we don't...

So maybe as data center kind of becomes a larger part of your product mix and portfolio, how should investors be thinking about the longer-term margin profile for your business?

Yes. So I think at Investor Day, we outlined kind of how we see the margins play out. I think gross margins are going to be somewhere in the range of where we are at today. We think there will be a combination of custom products being lower in margin than our current rate, but merchant products being significantly higher, and there will be some weighted average, and we'll have to see how kind of the two businesses play out relative to each other. So that's our view on gross margins.

The operating margin argument for Qualcomm has always been around scale and being able to leverage R&D across everything we're doing. And so as we grow in these new areas, it just automatically helps our operating margin, and that's the way we are thinking about it.

Yes. In terms of the investment balance versus that growth trajectory, maybe talk about that intensity.

Yes. So you should think of our investments increase in OpEx trailing the growth significantly. And so as we look kind of 3 years out, we've set a target of 30% operating margins. But really kind of as we go from there, there is no reason to invest more. It's really just kind of incremental revenue scale sitting on top of the existing investment base. And so excited about what that combination brings from a financial perspective as well.

Great. Now we've covered a lot of ground today. As you think about kind of Qualcomm's position in AI, compute, data center, et cetera, what do you think investors are still sort of underestimating about your opportunity ahead? And what would you sort of like everybody to take away from this if they only had to take away one thing?

Well, I think, obviously, investors are very focused on data center. We outlined a compelling strategy and a compelling set of products, but there was a little bit of, okay, but I want to see how you execute on it. And today is a key milestone for that. Obviously, very large deal with Amazon. But it's not the last one. It's just the first one. And we have a very strong engagement with another hyperscaler that we've not disclosed, but similar to the transaction. We -- the engagement we have with -- technology engagement we have with Amazon. So I think it's just the beginning. We're executing on everything we said we'd execute on. And hopefully, today's announcement gives people a lot of confidence that, that will be the case.

And then when you look beyond kind of data center and when you look at automotive, IoT, clearly, our position is very strong, and we're very well positioned to really execute faster than any target we've set out there as well. And then finally, on smartphones, the third leg of the stool, I'll say, we're at the low point in the cycle. Things can only go up from here, both from a way the market is evolving, but then also from the impact that we are seeing from memory in the short term kind of resolving through. So it is an opportunity to buy into a well-priced, highly diversified and growing stock.

Fantastic. I think unfortunately, we're out of time. But Akash, thanks so much for being here. Thanks. Appreciate it.

I appreciate it. Thank you.

Thank you, everybody. Hi. I am Amy Sutter, and I'm sitting in for Rob Sanders, our European hardware analyst who unfortunately, who is not able to be here today, and he covers Qualcomm. So I will be running the fireside chat. And I want to introduce Durga Malladi, who is the EVP of Tech Planning Edge Data Center. who's here from Qualcomm. And maybe just to get started, I know you said you've been at Qualcomm -- maybe give us a little bit of your background, kind of what different roles you've had?

Sure. Yes. Yes. So I joined Qualcomm in '98, so it's my 29th year over there. Inside Qualcomm, we have gone through like at least 2 different DNA mutations over the last 2.5 decades or so. So this is the third one. I ran our research and R&D organization for the longest period of time until about like 10 years back or so, -- that's when Christiano, who was -- who is our CEO now, my boss now. So he said, "I want you to come over on this side and help on the business. So for the last few years, it's been more than a few years, now over 8 or 9 years have been running the product planning. .

And at this point in time, I run all of our technology road map across all the businesses, ranging from this lowest end IoT all the way up to a data center. Last year, when we restarted our data center business, I ran the business for 1 year just to kind of make sure that we're in the right direction. And now we have like a full-fledged business unit that's running by itself. So that's my role.

Awesome. Great. I think since the Analyst Day, most common questions, I think, that we've been hearing from investors is how are you going to differentiate in the data center with your newly announced high-bandwidth compute technology -- can you maybe help us understand the genesis of HBC and kind of where it fits into your total strategy?

Yes. It started off. First of all, when we unveiled our HPC product portfolio and the multiple generations that we are working on, -- it's important to understand that it's not something that we just did in 1 year or so. It was something that's been in the works for the longest period of time. But it was last year that I felt like we have reached a point in our R&D where now it's time to commercialize this. But working backwards from there, what is the problem that it is solving. That is something that became quite clear to us by 2021 time frame. When we were observing how the compute in each of these data center racks. And in fact, it also holds true for some of the other places. It's growing quite a bit exponentially, generation over generation, while the memory bandwidth was kind of stagnant or maybe growing more modestly, it's more of a linear growth. And it eventually will lead to a memory wall wherein you can throw as much compute as possible, but it is going to do absolutely nothing when it comes to influence decode portion of it, unless you solve the memory bandwidth as well. And so we invested quite heavily in our R&D. And we reached a point where last year where it's becoming like, yes, now it's feasible. We've scracked a lot of the solutions over there.

Keep in mind that we called it as high-bandwidth compute because it is not really a pure memory technology. It is actually about a combination of how do you have the compute right next to memory? How do you design it, codesign it in the right way and then make sure that it interfaces with the rest of the accelerator. And it was a direct take on an alternative way of doing things as opposed to -- and we feel very confident and comfortable with where things are right now, and that's why we unveiled our multi-generation solutions with the first generation coming out next year and the second generation after that. But that is a very unique proposition. And now it's fully endorsed by a large number of the memory vendors as well who've been our partners for a while, but now they are equally out in the open and talking about something like that.

Yes. Great. Yes. I would not have thought Qualcomm as being in the memory game. But talking about it as a compute technology is very helpful. Maybe just to dig a little deeper, when you talked about Gen 1 and Gen 2 and you -- I think at the Analyst Day, you showcased the bandwidth per watt relative to HBM and then you also have some very high numbers relative to SRAM. Can you talk about the unique selling points versus HBM and SRAM that you're talking to customers about in terms of solving the issue with the memory wall.

Okay. So 2 things over there. First is some of the numbers and the second part is on proof points, and maybe I'll shed a little bit of light in terms of how the discussions are coming up as -- so in terms of numbers, we always, always talk about tokens per watt because the power consumption is something that is very important. And today, with HPM, the power consumption is very high. But it's a funny anecdote on that because we talked about 6x improvement compared to HPM in terms of tokens per second to what. And the second part is HPM today is about, give or take, 21, 22 terabytes per second. We had like hundreds of terabytes per second. And when we compare it against SAM, SAM has got its own restrictions. It is the fastest interface, but it's also like it consumes more area, you will need a larger number of racks to do the same thing. So there's a TCO argument to be made against SRAM.

These are the 3 things that we talked about in -- at Investor Day. But the important point is, okay, these are numbers. What are the proof points? So the nice thing is that, yes, we tapeout the silicon, the silicon is back in the lab. It's coming along very well. stay tuned on that front as we start unveiling this a little bit more on what the numbers are beginning to look like. But we are pretty confident in terms of the first generation of the products. A bit of an anecdote on this. Like right after Investor Day, we spoke about this, we said, "Hey, here's HPM, and there were like these 2 nice animations, 1 which talked about, okay, this is no good with HPM. On the other hand, with HPC, this is great.

Right after that, I was in Korea with the 2 memory vendors, and I fully anticipated them to say, what are you talking about? We have this awesome road map with HPM that's like really great. They're going to -- it's going to be like much better than HPC. That is not the feedback that we received. Instead, what we got is, hey, how can we work with you much closer on this. And in fact, we have been working with them because we need to make sure that we have the right supply as we do the commercialization. But in addition to that, there's been a lot of interest coming in directly from the memory community itself, which has been a very positive and pleasant experience. And that's actually good. It also kind of validates some of the points that we made that we need a different kind of a solution as we move forward in these data center act.

The second generation of the product, so the first generation is slated to be commercial and shipping in 2027. We're already working on the second generation, which is going to be 2. And those numbers are even higher. And 1 of the reasons for that as an in terms of the relative comparisons against what we anticipate with HPM 4. And the reason for that is as we kind of look ahead to 2028, A lot of things are evolving in itself. On 1 hand, AI models themselves are evolving, which means the compute that we end up doing in the compute die or the logic die, which is underneath the DRAM stack. We actually are anticipating what we need to do, and we are throwing in even more aridited. It's harder to scale that with HPM, so the relative gain looks even better with that. That's where we are.

Great. Okay. One of the criticisms of it is like with the multiple layers of LPDDR on top of -- with high-performance logic that there's thermal issues. So maybe can we talk about how you're getting around these with I guess, your packaging technology. And then you mentioned Gen 1 and Gen 2, just maybe some milestones for us as investors to think about it's for us to look for in terms of this. And then I know you're talking about talking with the Korean memory players, but also are you engaging with hyperscalers today? I don't know if that's something you've shared yet.

Okay. Short answer is yes to all of them, but I'll start with proof points. So as I said, silicon is back in the lab, looking good. It's coming up along very nicely. I don't believe we have explicitly put down dates as to when this is going to be unveiled. But in the next quarter, I just stay tuned on that front. I think some of it is also within our marketing domain in terms of how we want to expose it. But we will have numbers. We will actually have the silicon validation coming in. And we are on track to ship both from the samples, the engineering and the commercial samples and this technology validation in itself, I think it's looking good. We don't anticipate any issues for the first generation. What's happened is that as we brought this up, and we started talking to every single hyperscaler. Every hyperscaler, they have a lot of their own in-house solutions. Most of them -- some of them are like, "I have my CPU. Some of them are like, I have my CPU and I have my AI accelerator. But very few of -- they always relied upon HPM as okay, it's 1 of those memory bandwidth things. I will acquire that IP and then I'll build my rack on top of it.

It's always been like really good when we sat down with all the hyperscalers and they took a look at this, like this is awesome. How do I actually bring this in? And at Investor Day, we specifically talked about these are 3 different tracks, independent tracks. For example, some hyperscaler might come in and say, I have my CPU and my AI accelerator, I need your HPC to be a stand-alone product that works with this. We have a solution for that. There might be someone else who will say, "You know what, I have my Accelerator. That's good. I like your HPC and you know what, your CPU is looking pretty good, too. Maybe I can take both of them. How does that work with me. We have the answer for that as well. It kind of is now at a point where with hyperscalers, we're not coming in with some sort of -- it's all or nothing. No, you can take individual components, if you want, or you can take all of the above. And then we take it from there.

Got it. One of the things when I think HBC, the concern is maybe when you look at prefill versus decode, and I know memory is more focused on the decode side. That's where the bottleneck is. But can you maybe talk about how HBC does there.

Yes. So -- if you are looking at inference in data center, typically, we split that problem into there's the prefill stage and the decostate. The prefilled stage is fully compute dominated. The more compute you throw it, the better it is. And off late, we are seeing something like a 2.5x to 3x increase in compute generation over generation. So you can always solve the problem with throwing more compute at it, HPC will not do much about that 1 at all. On the other hand, decode is completely memory bank we dominated. You can throw as much compute as you want. It doesn't matter unless you have a solution like this, it's not going to work. And that perhaps is the place where in all these discussions, not just the hyperscalers with a lot of the others as well, we see this, "Hey, I have my compute solution. I can use it for prefill, perhaps for decode, as you're doing HPC, there's additional logic that goes into the bottom die. -- can bring in that logic over there. So there's like even more of a mix and match that's occurring in that space.

That's how we anticipate at least in the short-term things to evolve. Now in addition to all 3 of these, CPU or AI accelerator and HPC, at Investor Day, we also talked about our custom silicon business, now if you think about everything that I've said, you're almost getting to the point of is it like a custom silicon for something that's going into a hyperscaler. I think there is a very thin academic line between these 2 arguments because once you start putting it together, it does go in that direction. But there's a heavy dosage of our IP that goes in along with mix and match with their IP as well.

Got it. Got it. When you were -- like when you -- I think Hynix and SanDisk are also talking about high-bandwidth flash, though it seems like it's a little ways out, at least that's my take. But is there -- is that -- and it could -- could hyperscalers take a pool of HBM and HBF and run the models more cheaply? Like how does your HPC defend against that kind of TCO narrative, if that's something that becomes a way that we're trying to attack the memory as well.

So when we talk to memory vendors and we kind of have gone into this in depth, with them in terms of HPC versus HBM and what's coming with the high bandwidth flash. I would -- this is my opinion over here at this point in time. HP Flash is trying to solve a slightly different problem. It's not quite exactly the same problem because 1 of the things that -- the argument over there is -- let's also use flash in addition to DRAM to actually do something else over here. That's not the same thing as increasing the memory bandwidth to the point that the decode performance becomes really good. But it has its own use case. There's still more work to be done. That's just the perspective that we have right now. But it is 1 of those stacks, which is kind of independent of this. Meanwhile, with HBM, comparison against HPM. The same vendors who have -- they've been working on HPM for a while, they are like, I need to open a parallel track in addition to this because it's not obvious to me that in the near future, at least for the foreseeable future, not just near future, there's anything like HPC that I can offer through my HPM.

Now time may well, maybe 5 years down the stream, it's going to be a different story. But as of today, that's the only solution that actually gets us to this level of performance gains compared to what we have.

What I think has been very interesting for me coming from hot chips earlier this week is just the diversity of solutions that are coming out today in compute, GPUs, CPUs and memory and how every hyperscaler and customer model company is attacking it differently. When I think about HPC, can it be relevant for other end markets outside of data center, maybe auto?

It's a very good question. And we -- when we talked about HBC at Investor Day and of course, at Investor Day, we were talking a lot more in terms of our investments into beyond what we typically have done in edge devices, and that's why we talked a lot about data center. But truth be told, HBC is a generic technology. It goes all the way down into devices. And let me explain why that's the case. And we actually are working, Believe it or not, we are working on HPC on devices that are around us today. We're actually working in terms of what is the technology that makes sense for that. And here's the reason why. In data center, the comparison is against stage PM, and we are talking of what's the best way to solve the decode problem and the memory wall problem over there. There is no equivalent problem in devices. Let's take a smartphone or an agent AI device, but there's a different problem over there, especially with agent AI devices, which are like -- these are devices that are being built as we speak with those who've been in the business of thinking AI first, and I need to build a solution around it.

But it's all about ambient AI. It's constantly running. That's why we call it agent. It's always running in the background and just waiting for you to say something and then it picks it up from there. But if you're always on ambient I always on. That means you're always burning power. Now HBC is computers right next to memory. So there's something else that you get. We've completely collapsed that transport latency between memory and the compute and not just that, the power consumption has come down dramatically. It's perfectly suited for ambient AI in all sorts of agentic AI devices that come in now. So the usage of HPC in devices is in the context of ambient AI, whereas in data center, it's about the memory bandwidth that it provides and the deco performance.

And when you say like edge cases? Are we thinking about like something like robotics, where we're thinking about AI there? Or could you think of it being used in a car or a phone?

Going all the way down to smartphones, tablets, PCs, XR devices and yes, into automotive as well. It comes down all the way to those.

And when do you think that happens?

Discussions are going on as we speak. Typically, we have like a lead time on this. Like if you have a commercial product in 1 year, then usually, it's like 3 years before that is when the commercial discussions begin. That's a question for -- once it shows up, you will realize it, but suffice it to say that we're not that far from the first generation of the commercial products coming in.

Got it. I guess then maybe kind of closing out on HBC a little bit. When do you -- I mean, do you think it's with Gen 2 or Gen 1 where you see customers really clamoring for it? And if like if you -- does it become a technology that you might offer to license to third parties to help them scale? Or do you look at this as like some kind of proprietary moat that Qualcomm can offer?

So there's 2 questions over there. So the first 1 in terms of going beyond like Gen 1 and as we're going into Gen 2, where exactly is the action over there. GEN-1 is next year. like that's like literally next year, there would be a few customers will go in that direction. But then as we talk to a lot of the hyperscalers and so on, -- there will be some separation between where the volume is going to be between Gen 1 and Gen 2. We will certainly see traction on Gen 1. I hope we see a lot more because more and more customers are lining up. So the earliest adopters are the ones who are going to go with Gen 1. But now we have so much of incoming interest coming in. I do anticipate more coming up the Gen2 as well. So that's the way that we see it.

The second part is in terms of how we see this technology. Just keep in mind that there's actually 3 different components to it. We're not a memory vendor. So someone is still doing their DRAM stacking, they're not doing the DRAM stacking ourselves. What we prescribe is this is how things should be done. These are the TSVs that need to occur. This is the design for it. The compute of course, comes from us. We do the all the logic inside that. We still have to work with the likes of TSMC to take it all together. I think they're all in our swim lanes in our own way. But we are kind of like the overall system integrated and putting it together in the right way.

So in that sense, this is our current model. We haven't said anything about how it's expected to evolve, but that's where we are now.

And so if I'm a hyperscaler customer, would I be the 1 who is focusing on the memory like in terms of like because memory is a big bottleneck today. You don't have to worry about that. The hyperscaler worries about that, and you just help get the solution integrated together in any...

So when a hyperscaler is interested in our HPC product, they will have certain volume commitment, some sort of a demand signal that we get, but they talk to us and they will cross check and make sure that, hey, you're all like from a supply perspective, that's great, but they talk to us. And we are like the front end for all of that conversation. And of course, they will be talking to the memory vendors to make sure and to the likes of TSMC to make sure everything else is an in the right order. But this is a product that comes from us to them. They are our customers. So that's how it starts.

Okay. Okay. Great. Switching over a little bit. Let's talk about the drag implies C1000 server CPU. You have some very high claims in terms of performance per watt relative to peers. But it's a second half '28, if I'm correct, right. So you're kind of giving away some stuff -- how do you -- like how do you think about what competition is doing? And does the baseline shift from the other big CPU competitors. Could it be challenged at arrival?

So last year, I remember when we decided to restart this and I was thinking about, okay, what are the differentiations that we have over there. The first one, I was like, okay, we got to go into the memory architecture and memory wall, solve the problem. Our R&D project is good to go. So that's good to go. Denis looking at, okay, where are we with our existing CPU solutions? And what's the right way to scale that. We had something in planning already at that point because over the last 3 or 4 years, we've been investing quite heavily in our custom CPU solutions our Orion family of CPUs. And I was paying very close attention to exactly where each of the data center CP data on center class CPUs are. Starting from -- you can start from Novus we do and take a look at the generation over generation. So what I was looking at is we're not competing against anyone else out there today. I want to compete against where they're going to be in 2028, and that was like the mindset with which we went in, and we need to beat it by a certain generation for it to be compelling.

If it's like 2% better, then nobody is going to take it. compared to their own in-house solution or something else that you get, it has to be really compelling. That's the bar that we set for ourselves. And over time, as we kind of built this up, and we did all the benchmarking typically, this is done with Speck and so on, and we have our projections. One of the hyperscalers -- the response from them was your numbers are too good to be true. I was like, okay, so what I'm hearing is, if you do actually show up a drug, you will like them, it was like, Oh, yes, absolutely. So now, we made our claims. You're absolutely right. We feel very confident about those claims. We have absolutely no any sort of a second guessing on that at all. Now it's execution and crunch time.

We announced Meta as a first customer for C1000. They believe that claims. Qualcomm is a company where if you say this is what is going to happen. Never have we actually fallen short of it. We have always been on par or better. So they went with our reputation and said we are good to go. There are others like let's see your silicon and then we'll river. So I'm really looking forward to that, but I feel extremely confident.

And so the next milestone will be launched with your first customer?

Launch with the first customer, but even before that, that's the commercial launch is in '28, but with silicon back in the lab, all the proof points and so on, hopefully, we'll be able to share a few more.

The way -- the offerings we see in the market today, some of them are more focused on more cores. Some of them are more focused on single core like hyperthreading or if I get it right. So I mean I don't -- I'm kind of curious from your perspective, when you look at CPUs in the data center, what you think is the right solution or maybe there isn't, like, again, because workloads are going to vary so much, but I would love to get your opinion on that.

Until about 2 years back, I would say that, by the way, the CPU landscape is evolving as we speak, and it's becoming like for the longest period of time, including inside Qualcomm, by the way, it's like our XPU team or the GPU and the NPU teams are like, okay, we are in the AI business. And the CPU team is not quite there. It's a little bit out there, but it's not quite like that anymore. All 3 processes are important, but in a different way. But what's happened in the last 2 years is it used to be that you had 2 different kinds of workloads in data center. General-purpose compute, that means you just run, you light up all the core if you have -- at Investor Day, we said that our racks have like 250-plus cores and so on. So Okay. So let's say that you have 256 cores or so, then you light up all of them. It's a general purpose, it's like a work horse. Then there is a different configuration, which is an AI head node, where actually most of the AI workload is being done by some XPU that are sitting out there.

The CPU's job is to do some management and kind of get out of the action and let the GPU or the NPU perform whatever it is that's needed. All the AI inference occurs over there. But that led to an asymmetry in terms of number of CPUs to XPU. It was like 4:1. It used to be 8:1, became 4:1. But with agent, that's changing quite a bit. In fact, nowadays, we are more in the 2:1 configuration for every CPU, there's like 2 XPUs, maybe that's not 4. And I wouldn't be surprised if it actually becomes even less than that, like it becomes more and more symmetric. So as these conversations evolve, we initially started, hey, your course looked really great, excellent for general purpose compute. I think I got it for AI head node. I probably don't need you, but general purpose compute is great. But over time, now we are getting like -- let's actually talk about not just general-purpose compute, but also agentic AI because I would like to use those CPUs over there as well. So that's the evolving landscape with CPUs and the workloads themselves are shifting. They're becoming a mix and match of general-purpose compute and agent AI, not just AI head node, but agent workload, which look pretty different.

Right. Okay. So -- and then also kind of talking a little bit about your road map overall. Like you have -- you're also using your risk 5 cores. And so you kind of have 2 road maps here with ARM and risk 5. Maybe you could talk about like why do that? What -- how -- does that create what challenges does that create or what opportunities does that create?

So a couple of data points on that. The risk 5 conversation is an important 1 because it's usually -- I would argue that whenever 1 talks about risk 5 for the longest period of time, it was seen as something which is academic, which is a little bit out there. It's a good science experiment. If you're a grad student in the university, it's an excellent project that you should be working on, but not really for commercialization. That used to be the narrative till about 1.5 years back, not anymore. At this point in time, there's a lot of not just interest, but pre-commercial engagements that are occurring in terms of where do we take risk 5 cores and -- but I'll get to that because the first question to ask is, why? What's causing this? What exactly is the issue with this?

Historically, what has happened is that if you're in the custom CPU custom arm CPU business, then you inherit a certain architecture and your innovation is within the micro architecture domain. That's where you tend to innovate and you try to differentiate. But you inherit a certain architecture in so much and that's all there is to it, which means your scope of innovation is only below a certain threshold but not everywhere else. Risk 5, on the other hand, is completely open. If you have some ideas, you actually go to the standards process and you actually -- you say this is what I would like to see and that gets done. It's an open standard. So there's more innovation coming over there. It's a clean start. IFS every 3 decades or so, there's a new ISA that comes in and instructions at architecture. And we see a lot of potential in resi as well. So that's where as Qualcomm, we started investing quite heavily, first, making sure that the standards organization is like it's a professional organization that is run with commercial milestones in mind and not just for research purposes.

So Qualcomm has a lot of leadership positions in the risk forums. In addition to that, we work with our hyperscaler partners and say, hey, it can be just us. there's got to be like an entire software ecosystem that needs to come in, and they started participating quite heavily, and we have like very good partners over there, a few hyperscalers, both in U.S. and in China, which brings me to the other point because the other negative that comes up is Risk 5 is it's something that is going to be done in China, but I don't know about the rest of the world, not true. In fact, we will start seeing that in the U.S. as well because there's a lot of incoming interest.

So as a part of this, we have started investing quite a bit in our risk 5 portfolio even towards data center, especially towards data center. And last year, we made an acquisition of Antenna Microsystems. And that team is now fully integrated and we are kind of on our way.

Where -- I mean what is the advantage of risk 5 then? So a couple of things coming?

First of all, it's going to be a parallel track, which is occurring at the same time. And we see the demand is going to be like a mix and match of both over time. We expect to see that just because of the features that are coming in on the risk 5 road map as we look ahead and it's to be seen as to how that actually -- where does this end? Are we going to have 2 tracks. We had x86 annam for the longest period of time. And so one shouldn't be surprised to see even these parallel tracks come in. It's not A or B, it's A and B depending upon their own customer needs and choices.

Why did China -- it seems like China has been the early adopter. Is there a reason for that?

That's a good question for China.

All right, I'll ask them. And then maybe to move on a little bit on the software side of things. Obviously, as we all know, I think NVIDIA dominates with CUDA. And changing those programs to be able to adapt to non-NVIDIA chips, it's hard. And modular, that software, I think it -- how does it make it possible if I'm an NVIDIA to maybe use for example, maybe your CPU or what have you. So how do you -- how does that happen? And then how do you reassure developers over time that these compilers will be open and available. So that they don't get locked in a modular.

Absolutely, yes. By the way, that's -- there's like 2 or 3 questions over there, but I'm going to actually start back with -- it's okay, but it's a very big investment that we've made, and it's kind of an important concept. So I'll start with -- at Investor Day, we talked about our acquisition of Modular. We hadn't closed yet at that point in time. And so Chris talked about Chris Lattner and he actually talked about, okay, where this comes in. And there was this 1 flagship slide. For those of you who might have either attended or seen it on video. And if not, I'll just explain where we put together with Chris and Tim, we were literally on the floor and saying, okay, we should put it like this, like literally put them right next to each other. So there's Mojo and then we put in CUDA on 1 side, that's from NVIDIA. Then we said MAX, that's with the compiler layer. I said this is Triton with the compiler layer over there and modular cloud, and that's animal on top of it.

So it's like this is the CUDA stack, if you will, and this is the modular stack. Underneath that, from a hardware perspective, it's NVIDIA hardware only. And here, anyone -- it can be any third-party hardware and any third-party processor. It's a bold claim. And then right below that, we put in a footnote. We said, when you take modular stack, and you take your existing RAC as an example. And you have your native stack and then you say, okay, instead of that, you use modular stack. The performance is going to be on par or better -- and when we said better, we said up to 50% better. That was a fine print over there. It took some time to actually write that down. And I remember telling okay, Chris, okay, now that we've torn this gauntlet, so we're going to get a lot of questions in, and it's going to be fun. And it is. In fact, 1 of the first things that we did at the moment we did that we got some really good feedback from a lot of the others saying, this is awesome. I would really like to actually try this out and take a look at because others have tried.

Others have tried exactly the same thing, and it's not quite panned out that way. It seems like too good to be true. You have true disaggregation of third-party software that can run on any third-party hardware. And as Qualcomm, we are saying -- here is a software product, tried out and it can run on your hardware as well. And second, it's going to be open. There's quite a few new things from Qualcomm's perspective. This is -- that's why I said it's our third mutation over here. Last week, at MODCON, then we went 1 step further, then we said it's going to be open source for Mojo. And on MAX, by and large, it's open source. It's an Apache 2.0 license, anyone can take it. They will have full source access of this. There's some portion of it that's going to be licensed, and we go from there onwards. And we showed even more numbers.

The best part, and we had a person from AMD, who showed up on stage and saying, this is awesome. I'm going to try this out. That is like a game changer. I mean it seems -- yes, that's okay. But actually, at least for us, we were like it's quite something to have a third party come on stage and say, I'm going to actually start taking your software, and I'm going to try this out. So everyone is interested in now kicking the tires and taking a look at what the performance is going to be. And we really want third parties to reach that conclusion themselves. What we don't want is, yes, we are doing our own analysis. We have our benchmark. But if I were to show up over here and say, this is how much better it is, but it's all done by Qualcomm, no, I want AMD to do it. Want someone else to do it. I want third-party artificial analysis unit do it and reach the same conclusion. That is what is happening as we speak.

It's interesting you say AMD, I mean, does it compete with Sara Rock.

That's something that they are kind of -- that's a question for them, by the way, but it was nice to actually have them say, this is something we want to try out. Where that takes them, that's a different story, and we'll take it from there. And the final part that I wanted to mention, which is kind of -- it was kind of a humorous anecdote because the modular -- I was talking to Chris, they had brought up modular software on everyone else's hardware, except Qualcom, which is kind of so odd. And so last week, for the first time, we also unveiled running on our platforms. including 1 of the older generation AI 100. That's on the data center side, and we even showed it on 1 of the laptops, the X2 Elite platform.

So it's the beginning of -- and we said any hardware, it, of course, must include Qualcomm as well. So we're beginning to do that.

Got it. Can you talk about the partnership you have with Hugging Face?

Yes. So that was the second part of that presentation at Investor Day, Hugging Gace 15 million developers flocked to the website on a daily basis, largest repo of AI models, open rate models out there, 3-plus million models and on the other hand, we have some data that indicates that the most popular ones are probably like far more concentrated, not all 3 million are the same and some of them are derivatives. But clearly, it's the place to go if you're a developer, you're in the business of building apps or writing agents or building agents, first, you go to shop. You actually -- that's like the place that you land up with and say, what do I have to work with, like all the things that are available out there. So in our partnership with Hugging Face, we want, first of all, agentic onboarding of any of those models. Any of those models, agent onboarding of those models onto our platforms. What does it mean? Today, I go to Hugging Face, I'm going to click, I have to manually click on something and then it gets downloaded. They don't have to write something with that and then have to write some specific comments, and then I can start building the applications. I don't want to do that.

I have 1 small thing. They actually have something called Hugging chat. I would like to pick some model that does object detection and classification or a text to video generation model, pick 1 of the good ones and make sure that it runs on this platform from Qualcomm. That's your prompt. Everything else happens behind the scenes. You can actually see the code being written down out there and it gets done and it comes on to your platform. That makes it extremely easy for developers to start adopting our platforms over there.

Second piece of the puzzle, the Tuen modular as a part of that as well. And that means if I'm a developer, I would say like instant PyTorch, maybe I want to actually use modular. A developer might say, do I have to now learn another language. Mojo just happens to be very similar to Python. But these days, increasingly, developers themselves are not necessarily writing all the code. They're using coding agents. So I'd rather say using module, I would like to see an application being done. -- that's actually written because it does all the tool calling, looks at all the documentation getting done. This is where Hugging Face has emerged as a really good partner for us to do a direct engagement with developers while at the same time, those developers get exposed a lot more with the modular stack.

Got it. It sounds like a great partnership.

It is.

Yes, for sure. And then you kind of touched on this when you're talking about running modular on some of your older technologies, Qualcomm technology that you just announced. But can -- can you talk about how it might then support your cloud-to-edge strategy? And can it be like a universal layer for -- to write an enterprise agent AI agent all once and kind of trust that it can run on these different devices, whether it be a laptop or data center CPU.

So that's the, I think, the final piece of the puzzle. We talked about it even in our hugging face engagement. If you take these emerging agentic AI devices that we are talking of, a good fraction of them actually the user interface is directly to an agent. You ask something you want a specific task to be done. The agent then decides what needs to be done. Do I run some inference on the device. First, I need to shop around inside the device and see what models do I have to work with? Is this good enough for what I need? Or do I need something better? If it's good enough, I use that, but you might still need something else that's running in the cloud, so you go there and you do a second inference instance that's running there. That's 2 parallel inference instances running concurrently.

It's not 1 or the other, but it's both. And meanwhile, there might be some sort of a tool calling with a web call in which you're extracting information from somewhere else, that's the third part. When you put it all together, we have reached the conclusion that it's never going to be one or the other. It's going to be all of the above, and the agent is running several inference instances. So hypothetically, if you were to picture a world in which you have Mojon MAX, the modular stack actually running on the device, and it happens to be running on the cloud instances where the infants is running, does it make it more efficient? And the answer is yes. So that actually might be -- and by the way, it could be any rack. It could be an AMD rate, it could be a Qualcomm back, it could be someone else's that. It doesn't matter.

But does that make it more efficient? We do believe the answer is yes, but that's what we are working on.

Maybe it's a basic question, but like how did modular -- like what was the secret sauce? Because you said so many people have tried this. What -- why were they able to do this when no 1 else could? Because I think a lot of people would like to break the kudos mode.

It's a combination of things, in my opinion, but 1 of them happens to be that if you kind of think about the existing stacks that are there, including CUDA, for example, CUDA itself is about 15 years old I mean what have we initially written for traffic and specifically focused. It evolved over time to this. So there's a lot of legacy as well. So 1 thing that's certainly been done is when it comes to the development of Mojo and MAX, it's kind of a fresh look at it, and it's turning the fundamentally different way, which makes it cleaner, simpler and for lack of a better phrase, more modular, like literally, it is modular, the objective.

The second part of it is that as you go through the compilers and this goes 1 level below into the technology in itself, compilers themselves have been gradually evolving there is notion of using AI within compilers itself, and that's emerging as well. So a lot of these things have been built into it. We believe that's the real differentiation. There's certainly far more to it, but I would actually reserve that for a tech deep dive.

Got it. And then maybe just to talk a little bit about prefill and decode in terms of modular tool stack. How does their software -- how does that software handle that? Can it do it dynamically?

So just in terms of the modular stack, the way that it's currently envisioned and constructed, it doesn't -- it kind of sits on top of like I would say, prefill and decode are like system-level concepts sitting upon certain hardware, but the software layer itself is kind of independent of each other. So everything that I said earlier about modular equally applies to both stages.

Got it. Okay. And then also, when you think about the memory wall that we've been and also the power wall, is there ways to does modular also kind of attack some of this in terms of like harnessing idle devices, et cetera, that might be able to help improve the efficiency or performance?

I think it's a little too early to come to that stage. We are still doing a lot of our analysis in terms of what else we can do in this space. But at this stage, I think it's a little too early.

Got it. And when investors are thinking about the modular, -- like what kind of milestones and proof points should we look for? Are you talking about like number of developers using or...

We have a lot of things actually planned out over the next quarter or 2. As we start bringing in speeds and feeds and data points in terms of, okay, here's an existing rack, Here's with the native stack. This is the performance based upon all these workloads and you compare in contrast with what happens when you have MojonMAX running on top of it. That's like the first set of tons and tons of data coming up on that one. The second proof point of that is, how well does it actually scale into all the other platforms beyond data center. Keep in mind that as Qualcomm, from a device perspective as well, we are in the business of any framework, any run time, any operating system. And we provide all the debuggers, the compilers, the tools and whatnot that actually go along with it. And there is a very rich set of run times that already exist. I mean you have executors from Meta. We have light RD from Google. We have Winmill coming in from Microsoft. They're all great partners of us. We will continue to support them while at the same time, also bringing in our stock as well. So the next step would be to how does it work on a given platform, when you already have support for 1 and where does modular come in? That's the other part that you will see over the next couple of quarters.

Got it. I mean we've kind of touched on a lot of different things. HBC, CPU and modular, a lot of exciting new opportunities for Qualcomm. I guess, like -- do you think there's something we didn't cover right now that you'd want to point out to investors? I know you just had your Analyst Day, but I thought you might wrap with that in the last.

I think yes. I think the 1 part that I would like to say is that there was a fourth piece of this. We talked about GPU NXP and HPC, but custom silicon, I think, is going to be important for us. So post our acquisition of Alfa way, we've retained the custom silicon business. Increasingly, all of our discussions with hyperscalers involved some level of customization. It's a mix and match of everything that we bring to the table, along with the -- and keep in mind that we are in the networking business now. We have CDs. This is what we acquired from Asaf, that will continue to evolve. And that's 1 place where I would like everyone to be reminded of the fact that prior to acquisition, Alphawave already had these hyperscaler customers. So we'll continue to go in that direction.

Great. Well, thank you so much. I appreciate you.

All right. Thank you.

Ladies and gentlemen, thank you for standing by. Welcome to the QUALCOMM Third Quarter Fiscal 2026 Earnings Conference Call. [Operator Instructions]

As a reminder, this conference is being recorded, July 29, 2026. The playback number for today's call is (877) 660-6853. International callers, please dial (201) 612-7415. Playback reservation number is 13761080.

I would now like to turn the call over to Brett Simpson, Senior Vice President of Investor Relations. Mr. Simpson, please go ahead.

Thank you, and good afternoon, everyone. Today's call will include prepared remarks by Cristiano Amon and Akash Palkhiwala. In addition, Alex Rogers will join the question-and-answer session. You can access our earnings release and a slide presentation that accompany this call on our Investor Relations website.

In addition, this call is being webcast on qualcomm.com, and a replay will be available on our website later today. During the call today, we will use non-GAAP financial measures as defined in Regulation G, and you can find the related reconciliations to GAAP on our website. We will also make forward-looking statements, including projections and estimates of future events, business or industry trends or business or financial results. Actual events or results could differ materially from those projected in our forward-looking statements.

Please refer to our SEC filings, including our most recent 10-Q, which contain important factors that could cause actual results to differ materially from the forward-looking statements. And now to comments from Qualcomm's President and Chief Executive Officer, Cristiano Amon.

Thank you, Brett, and good afternoon, everyone. Thanks for joining us today. In fiscal Q3, we delivered revenues of $9.9 billion, coming in at the high end of our guidance and non-GAAP earnings per share of $2.21. QCT revenues were $8.5 billion, with another quarter of record automotive revenues as well as growth in IoT. Licensing business revenues were $1.3 billion. .

At our recent Investor Day, we lay out the next chapter of Qualcomm built across 3 dimensions: one, expanding into the data center with 4 unique product lines; two, driving agentic and physical AI compute everywhere; and three, expanding beyond silicon to full stack software and platform solutions. We also updated our fiscal 2029 financial targets, which now include more than $24 billion in revenue across automotive and IoT plus more than $15 billion in data center, bringing our total non-handset revenue outlook to $40 billion by fiscal 2029, up from our previous target of $22 billion. This reflects our conviction in the opportunities throughout the end of the decade and the scale of our business diversification.

In the short term, the entire industry continues to be impacted by unprecedented memory prices, higher manufacturing and input costs as well as supply chain shortages driven by overall data center demand. In addition to the resulting revenue decline in mobile and consumer electronics, this is creating short-term pressure on QCT gross margins which will be slightly below our historical range. We're implementing price increases and as they take effect, we expect to see gross margins realign to our operating model.

Despite these headwinds, we expect top line growth for QUALCOMM in fiscal '27, driven by an inflection in non-handset revenues throughout the fiscal year. We're incredibly excited about the next chapter of Qualcomm, our relevance in the next phase of AI and distributed intelligence from edge to cloud, and we remain firmly focused on the execution phase of our strategy.

I will now share some key highlights on the business. Let me start with data center. This is the ideal and logical time for Qualcomm to enter the market as agentic workloads are reshaping the economics of AI. Efficient token generation and total cost of ownership are fundamental to scaling AI. And as a result, inference is becoming disaggregated in the data center and will be increasingly distributed. This means hybrid inference will evolve across the entire compute continuum from data center to on-premise network edge and edge devices. Given Qualcomm's assets, it's a natural evolution of our growth story.

We are developing a differentiated set of product lines, including connectivity in fiscal '26, custom silicon and AI accelerators in fiscal '27 and server-class CPUs in fiscal '28. Our portfolio is rolling out in phases over the next 2 years, leveraging decades of leadership in power-efficient compute and strong ecosystem presence and relationships.

Our 2 near-term custom silicon wins will be revenue generating in the December quarter, and we have begun wafer production. Both projects are in the first phase of strategic multiyear customer relationships that we expect to expand over time. Our innovative high-bandwidth compute solution is designed to address one of the industry's most difficult bottlenecks by integrating compute directly with high-density memory, improving performance per watt, memory efficiency and total cost of ownership.

I'm pleased to report that we have completed the tape-out of HBC Gen 1, an engineering milestone that move us into the next phase of customer engagements. We expect to demonstrate HBC performance on silicon in the coming quarters ahead of the launch of our first HBC solution in mid-2027. Across our merchant platforms, including HBC based AI accelerators, SerDes connectivity and CPUs, we are in active conversations with nearly every leading data center player about building long-term partnerships.

We're pleased with the activity and interest across these opportunities and expect to share more as they advance. As announced earlier today, we have closed our acquisition of Modular Inc., and integration is now underway. Modular strengthens our ability to deliver an end-to-end software stack for data center and edge AI deployments. It will also be hardware-agnostic helping simplify AI software complexity across multiple platforms and giving developers a modern and open environment for heterogenous compute.

This is an important step in how we see AI infrastructure evolving with software and hardware coming together to deliver better performance, flexibility and efficiency. Our vision and objective with Modular goes far beyond augmenting our AI software capabilities. We have the ambition to change the current industry approach to AI software from closed to open systems to promote enhanced competition, innovation and resilience.

Modular will host ModCon in August with some incredible announcements from industry partners, and we look forward to further engaging with developers and ecosystem partners at this event. In automotive, customer momentum continues to drive exceptional revenue growth. This quarter, we signed a landmark expanded agreement with BMW, winning a highly competitive selection process to become the lead compute silicon provider for their next-generation ADAS as well as digital cockpit.

This agreement represents a material expansion of our automotive pipeline and establishes Qualcomm as the lead compute silicon partner for BMW extending across model programs well into the next decade. We look forward to building on our existing cooperation with BMW in the years ahead. Additionally, our recently announced collaboration with Stellantis supports our automotive pipeline well into the 2030s. These agreements reflect the broad interest we're seeing for digital cockpit and ADAS.

Customers are shifting from socket-by-socket design awards to multi-generation strategic engagements as they increasingly recognize the value of our broad technology portfolio, platform approach and long-term commitment to partnerships, the automotive industry and open ecosystems. Further, with our fifth-generation Snapdragon digital chassis ramping in September, we're delivering a significant increase in content per vehicle, and we are on track to become the #1 automotive semiconductor player by revenue.

Last quarter, we said we were targeting an annualized revenue run rate of $6 billion as we exit fiscal 2026. Today, we're raising that outlook and now expect annualized sales of approximately $7 billion exiting fiscal '26. Within Industrial, we're strengthening our position across many verticals as they embrace AI at the edge and open weight models. At our Investor Day, we introduced a fiscal '29 projection of $8 billion in revenue for industrial networking and robotics.

We're happy to report that our industrial design win pipeline exceeds $7 billion with over $3.5 billion in design wins secured this fiscal year. This reflects strong customer demand and a meaningful increase in new businesses. We have a very broad portfolio of purpose-built silicon and full stack software solutions for this category. Arduino presence exceeds 38,000 customers supported by a deep partner ecosystem that is already yielding results. And with Arduino and edge [indiscernible], our reach now extends to more than 30 million users.

Moving on to handsets. Despite overall industry contraction caused by the current memory environment, we're seeing early signs of an agentic smartphone cycle that will grow over time. In China, major OEMs are preparing to bring new on-device agents and orchestrators to market, and we believe agentic experiences will play a larger role in premium tier demand as adoption grows.

Our share position at Samsung remains strong, with Snapdragon powering approximately 70% of their flagship devices as announced at Samsung Unpacked. Our collaboration is now expanding across the wider Galaxy ecosystem from the latest foldable phones and Galaxy watches to intelligent eyewear developed with Google, bringing new agentic experiences to more devices. This reflects a broader potential to reimagine mobile for the age of agentic AI.

Beyond smartphones, PCs, smart glasses and other new personal AI form factors are all becoming endpoint for agents that creates a significant multiyear upgrade opportunity for Qualcomm as today's installed base needs to evolve to enable more personal, contextual and autonomous AI experience. In PCs, we're growing our leading share of design wins in Google Books, bringing Snapdragon together with Gemini Intelligence for a new generation of AI first laptops.

With Microsoft, we're collaborating on Project Solara, a chip to cloud platform design for agent-first enterprise devices. And two, our Snapdragon Start program for smart glasses, we're delivering a complete reference platform that enables eyewear brands to develop their own devices. With roughly 600 million global eyewear units shipped every year, this program will help expand the ecosystem and accelerate the transition of this category to smart glasses.

You will hear more about this as Snapdragon Summit in September. At Investor Day, we laid out our vision for Qualcomm's next chapter and our path toward our fiscal 2029 targets. We're already seeing an inflection in our non-handset businesses, which underscores the success of our diversification strategy, and there's a lot more to come. Our data center business is just at the beginning of its journey, and we recognize that investors want to see more proof points that we can successfully execute on our plans as a new entrant.

We welcome the challenge ahead and are confident we will prove as we have many times before, that Qualcomm can execute and win in new growth areas, including data center. With that, I will turn the call over to Akash.

Thank you, Cristiano, and good afternoon, everyone. Let me begin with our results for the third fiscal quarter. We delivered revenues of $9.9 billion and non-GAAP EPS of $2.21, with revenue at the high end of our guidance. QTL revenues of $1.3 billion and EBT margin of 69% were in line with our expectations.

QCT revenues of $8.5 billion were at the high end of our guidance, and EBT margin of 26% was in line with guidance. QCT handset revenues of $5.1 billion reflect the impact of industry-wide memory dynamics on the global smartphone market. QCT IoT revenues of $1.8 billion were up 9% versus the prior year driven by growth within the industrial networking and robotics category of products.

In QCT Automotive, we delivered another record quarter with revenues of $1.6 billion, with 61% year-over-year growth driven by accelerating demand and increasing compute content per vehicle. Total non-handset revenues in QCT, including automotive and IoT, grew 28% year-over-year, underscoring the continued execution of our diversification strategy. Lastly, we returned $2.3 billion to stockholders, including $1.4 billion in share repurchases and $937 million in dividends.

Before turning to guidance, I'd like to provide an update on a couple of factors reflected in our financial performance. First, consistent with our expectations, we estimate that QCT handset revenues from Chinese OEMs reached a bottom in the third fiscal quarter and will return to double-digit sequential growth in the fourth quarter. Second, the semiconductor industry is experiencing broad-based increase in input costs across wafer fabrication, assembly, test, advanced packaging, memory and other materials.

We are taking concrete actions to reflect the higher input costs in our product pricing. These actions will benefit our gross margins over time as the pricing changes gradually come into effect. Finally, as a result of our supply constraint, we now expect an acceleration in the step-down of Apple product revenues starting in the fourth fiscal quarter as our share for upcoming iPhone launch is expected to be materially lower than our prior estimate of 20%. All these factors are contemplated both in our third quarter performance and fourth quarter outlook.

Against this backdrop, I'll now provide our guidance for the fourth fiscal quarter. We are forecasting revenues of $9.7 billion to $10.5 billion and non-GAAP EPS of $2.05 to $2.25. In QTL, we estimate revenues of $1.2 billion to $1.4 billion, and EBT margin of 68% to 72%, reflecting normal seasonal trends. In QCT, we expect revenues of $8.4 billion to $9 billion and EBT margins of 23% to 25%. We forecast QCT handset revenues to be approximately $5.2 billion, driven by sequential growth in Android, offset by lower Apple product revenues.

We expect QCT IoT revenues to remain approximately flat versus the year ago period, with double-digit growth across our industrial networking and robotics category of products, offset primarily by the impact of memory constraints on tablets and other consumer products. In QCT Automotive, we expect another record quarter with approximately 60% year-over-year revenue growth.

Lastly, we anticipate non-GAAP operating expenses to be approximately $2.7 billion in the quarter, reflecting the acquisition of Modular and continued investment in our data center product road map ahead of revenue ramp.

Before I conclude my prepared remarks, let me summarize the key drivers of QCT's growth trajectory going forward. We are well positioned to execute on the vision we outlined at our recent Investor Day with QCT non-handset revenues expected to grow to $40 billion by fiscal '29, nearly double the target we had previously provided. This forecast includes data center revenue growth to $5 billion in fiscal '27 and $15 billion in fiscal '29. As a result of our diversification execution, we now estimate non-handsets at more than 50% of QCT revenues in fiscal '27 and grew to approximately 2/3 in fiscal '29.

In the short term, we anticipate growth in non-handset revenues relative to prior year to accelerate from 24% in fiscal '26 to greater than 60% in fiscal '27, a significant inflection point in the execution of our growth strategy. We expect this growth from non-handset revenues in fiscal '27 to replace total Apple product revenues in '26. In handsets, when memory industry dynamics stabilize, our Snapdragon product leadership and emergence of agentic AI experiences will position us well to reinstate QCT Android revenue scale and growth rates.

Lastly, I'd like to welcome the Modular team to Qualcomm. We're excited to have completed this transaction, adding a world-class team whose AI software expertise will enhance our ability to execute on the significant opportunities ahead. This concludes our prepared remarks. Back to you, Brett.

Thank you, Akash. Operator, we are now ready for questions. .

[Operator Instructions] The first question is from the line of Joshua Buchalter with TD Cowen.

Congrats on solid results in a tough backdrop. I wanted to start on the gross margins. It's pretty clear. You explained what was going on with the rising input costs and now you're raising prices. Can you walk us through how we should think about QCT gross margins returning to their prior levels, how long between the ASP increases kind of match the input costs rising?

Josh, it's Akash. So as I said in my prepared remarks, there's 2 key drivers on the impact on gross margins. I think the first is a little bit of a weaker mix within premium tier. As you know, we have multiple chips within premium and you're seeing operator -- OEMs making a choice on which chip to use and also using prior generation as a response to kind of the memory cost increase environment.

And then the second factor is the higher input cost across the supply chain. And so as you would expect, kind of we're taking action to increase the prices and reflect it in our customer product pricing. And we expect this benefit to show up in our gross margins over the next couple of quarters. These changes, as you would expect, come in gradually as we have some contracts in place, and so we need to get past those contracts. There are also product cycles that happen.

So this will come up over time. But I think when we get through it, we expect to be consistent with the historical gross margin range we have.

Maybe to follow up, you mentioned that data center revenue from your ASIC engagements will start to layer in, in the December quarter. Any help you can give us on the sort of the shape of that contribution through fiscal 2027 as it's obviously quite a material step up in revenue growth there?

Yes. So I think as we said, the revenue starts in the December quarter, and you should expect a ramp as we go through the year. As we've said this before, we have 2 custom chip engagements, and both of these are global scale hyperscalers and we're going to expect to start seeing revenue from both of them starting in the December quarter. As you know, the December quarter is right there. And so we do have POs from these engagements, and so we've already started wafers. So we're very confident about the engagement with both the customers.

Our next question is from the line of Joseph Cardoso with JPMorgan.

Maybe just as a first one and a follow-up kind of on the pricing dynamics. Can you just flesh out the pricing actions, like any commentary in terms of the magnitude of price increase that you're looking to take? And then whether these actions are broad-based or you're going to look to be more concentrated across the portfolio? And just as we think about maybe pricing actions in some of these consumer markets like handsets, how do you navigate rising prices in a market that has already seen demand affected by cost inflation and other components? And then I have a follow-up.

Sure, Joe. So the way you should think about it is this is a pricing action that we are taking broadly across different end markets. As I said earlier, there are certain places where we have a contract or we are waiting for a product cycle to come through, so it will layer in over time. But it's no different than what a lot of the peers in our industry have done, and you should expect something that the scale of the increase that we're looking at is double digit and consistent with some of the actions from other players.

Maybe, Joe, this is Cristiano. I'm just going to add 1 comment. I understand your comment, but I think the market is actually down because of the magnitude of increases in the bill materials with memory. So even a double-digit price increase, which is just a pass-through of the input cost increase and wafer price increases, exactly is small when you compare it to the order of magnitude of the memory bill of materials. So we actually don't expect that fundamental changes in the premium tier and the higher tier volume. And we maintain the position that China handset this Q3 is the bottom.

Fair. Understood. And then maybe just as a second one, can you provide any early thoughts on how you're thinking about handset seasonality into the December quarter, just given kind of the moving pieces here around, obviously, Apple and then the recovery on the China OEM side of things?

Sure. So as you would expect, we're not necessarily guiding the December quarter at this point, but let me give you some kind of qualitative comments about sequential trend on revenues in QCT between fourth quarter and first quarter. So a couple of key factors there. First is Apple. As I mentioned in my prepared remarks, we expect materially lower share in new launches versus our previous estimate of 20%. And as a result of that, we are forecasting approximately 50% decline from September to December quarter. This obviously accelerates kind of the exit of Apple revenue out of our model. .

The second is Android. We expect Android to grow and significantly offset the reduction in Apple and then data center revenue that starts to ramp in the first quarter. So the revenue profile in the December quarter will be a combination of these things, and we expect it to be slightly up on a sequential basis. When you look at the full year, you should not think of previous seasonality where first quarter was the high quarter for us. The profile obviously changes with Apple having a very strong December quarter, not in our model anymore. So we actually expect revenues to grow when we go from the December to the March quarter.

Our next question is from the line of Stacy Rasgon with Bernstein.

I wanted to ask about the accelerated Apple. The wordings on the paragraph in the slide, it sounded like you guys are making the conscious choice not to sell to them nearly as much as you were before. I mean, you sort of blame the supply -- is that true? I mean I don't want to be too dramatic, but are you basically starving Apple getting them out quicker than you could have and using that silicon to send it elsewhere. I mean is that what's going on?

Yes. I wouldn't characterize it as such, Stacy. You should think of it as supply -- our supply constraints were a part of it and then where discussions ended up is that we'd have a share less than -- materially less than 20%, and our revenue as a result in '27 from Apple product would be less than the previous guidance we had given, which was a little over [ $2 billion ]. The way you should think about '27 is really the growth that we are targeting in the non-handset areas, which we said is going to be greater than 60% on a year-over-year basis will replace the entire Apple product revenue within the year.

Got it. Which must be about $7.5 billion. If I just take the 60% growth year-over-year, that would be something like $7.5 billion in non-handsets. So that's about what the Apple revenue is in '26. And most of that is going away in '27. That's how I just think about it.

I think that's a fair range of estimate. I think we have given some additional disclosures in our website that you can look at to get some precise data points.

Got it. And for my follow-up, I know the gross margins, hopefully, get better over the next several quarters, but you also have data center ramping as well. And you sort of talked to it at the Analyst Day that data center stuff should be dilutive to gross margin. So I mean does it offset some of that gross margin recovery? Or how do we think about that?

Yes. I think great question, Stacy. So you should think of it as our baseline business has a certain gross margin range, which has been in the 48% to 50% range. And so with the price increases coming online over time, we expect to be at that range. And really, the data center revenue coming in since the first revenue is mostly from custom chip engagements. We do expect that to be significantly lower than our ongoing kind of baseline gross margin percentage. And so that will be a drag of 1.5% to 2% on the weighted average gross margin for QCT.

Our next question is from the line of Joe Moore with Morgan Stanley.

You talked about the stronger automotive ramp exiting the year. Can you just give us a sense of what that mix looks like at this point? How much ADAS is coming into that revenue stream? And then how do you think about future drivers? Do you see ADAS as a bigger driver next year? And do you see any autonomous kind of creeping in?

Joe, this is Akash. So if you look at the kind of the breakdown we shared at Investor Day, you'll be able to see how ADAS is a significant portion of our design win pipeline, and that's a representation of how revenues will flow through over the next several years. I think one of the key things to take away from kind of our ongoing traction on the revenue side and design win side, in automotive is that we are transitioning to a place where a lot of our engagement across major OEMs is across platforms. So rather than kind of competing for individual sockets, or individual capabilities within the automotive within a car, we are winning across the board across the entire platform because that just brings a lot of technology synergies to our customers.

And maybe, Joe, it's Cristiano. I'm just going to add one thing. We did talk about that before as a driver and that continues to be the case. As you move to our latest generation silicon, the amount of processing capability, it's an order of magnitude increase. We have seen a step function in silicon content. So this OEM's choices to apply next-generation silicon as part of the mix is actually increasing the automotive revenue, even ahead of our expectations.

That's helpful. And then separately, just wonder if you could talk about your supply chain your price increases reflect some input costs coming up. Just how are you feeling about your supply chain? And particularly, are you seeing constraints on the wafer side going forward?

Yes. That's a great question. Look, I think the industry now is probably operating very similar to what was in the pandemic. Everything is at 100% utilization, everything. So we're seeing a shortage in price increases across wafers, across assembly across the testing tester every thing.

The second answer is the same answer I had -- I was providing, I remember at the time of the pandemic. It's good to have scale. It's good to have scale. It's good to have significant volume across different nodes. This is actually helping us. And you're probably seeing what we did, not your question, but what are we doing on inventory, inventories also in times of shortage is a strategic advantage as well. So we're actually very comfortable. Nobody is comfortable. Everybody wants more. I don't -- I have not met anybody that is using leading nodes today that there's not one more, but we're comfortable that we have the supply to basically execute on our plans.

Next question is from the line of Ben Reitzes with Melius Research.

I wanted to talk about your data center initiative a little bit more. In June, you had enough visibility to double your target. And just wondering if how is the reception outside of the 2 customers? And when you talk about the pipeline and some of the commentary around that, are you implying that there could be more than 2 customers? And how do you see that expanding beyond the current reach?

Yes. Very good question. I'm going to break that conversation in probably 2 parts. One is the existing customer engagements. As we said before, we have line of sight to how those things expand across the custom silicon, the beginning of the accelerator with high-bandwidth compute as well as the ramp of the CPU to go from $5 billion to $15 billion.

And I think we continue to restate those numbers. I think as we get to $15 billion, you're going to see a combination of multiproduct, multi-generation on the custom ASIC plus the ramp of the accelerator and the CPU. Now the second part of your question is actually the most interesting one, and that's why I provided some commentary in the script.

I know everybody will live in an era of everybody wants instant [ gratification ]. But the reality is there's a lot of customers that want to see silicon. And that's both true on the accelerator on our HBC plus the accelerator as well on the CPU. So we're actually very happy. Everything is going according to plan on HBC. That's a disruptive technology. I said in my prepared remarks. In the coming quarters, we're going to have silicon, and we are going to be able to do silicon demonstration evaluation. I think that has the potential to unlock new opportunities. We have additional conversation with customers. But the next milestone is Qualcomm, we needed to see silicon and see how it performs before we make a decision that we'll be marching forward.

Can I just follow up on the comment around China bottoming. Are you hearing from these customers that they're getting access to domestic Chinese memory and they're feeling really good that they're going to have supply? Or I just think I'm getting some questions after hours is why you have that confidence? And is it sustainable? .

Yes, Ben. So a couple of parts to that question. I think, what we've seen in China is, first of all, our revenue was the lowest in the June quarter, and we are forecasting double-digit increase in revenue in the September quarter, which is the quarter we are in, obviously, very high confidence. And then also going into next quarter.

And one of the key drivers there is earlier in the year, the OEMs were buying based on the size of the market, but they were also drawing down on channel inventory. And where we are at now is the channel inventory is -- has thinned down that they can't draw down anymore. And so we are reconciling -- the revenue is reconciling to the size of the total market. And so that kind of change in profile of purchases is what gives us confidence.

To the first part of your question is around Chinese memory. That's certainly been a big part of what the Chinese OEMs use. This is not a new trend. It is something that has existed for the last several years. And certainly, it's something that will strengthen as we go forward.

The next question is from the line of Vivek Arya with Bank of America.

This is Liam for on for Vivek. I guess just to start, how are you thinking about potential for U.S.-China restrictions impacting your data center business? And how is that factored into your decision-making leading up to kind of this ramp in fiscal '27?

For the engagements we have today we're not restricted. And it's within the category of other companies, they're also providing solutions for China. So that is not a concern at this moment.

And then how are you thinking about handset market TAM for Qualcomm in 2027 if memory prices continue to be a headwind. Is it going to continue to kind of remain depressed? Or can it rise up to the kind of 5% CAGR that you highlighted at the Investor Day?

Yes. So I think at this point, where we are looking at for the handset market is we think the market will be down low teens relative to [ '25 to '26 ]. And a lot of the impact is in the lower tiers. And so this is largely consistent with how the industry analysts are viewing it as well. The net impact on QCT Android revenue is -- the revenue is down 20% year-over-year. and the EPS impact of it is greater than $1.50.

So the way we think about it is longer term as conditions stabilize. And with our Snapdragon leadership and the agentic transformation of phones, we have the opportunity to reinstate the scale of the handset business and the growth that comes with it. So you should think of this greater than $1.50 of EPS as a potential tailwind for us as we go forward and the markets normalize.

Our last question is from the line of Chris Caso with Wolfe Research.

My question is on QTL. And it does look like QTL revenues have held up pretty well despite the decline in the market. Do you expect that to continue? And are the QTL revenues helped by any way by some of the price increases that your customers have had to implement?

Yes, Chris, you should just think of QTL consistent with the model it has, right, which is the -- there is a cap on the total royalties based on the device ASP. And so to the extent that prices go up below the cap, there is some benefit that accrues to QTL. And then above the cap, it really doesn't make a difference in the revenue for QTL. And so this is execution of licenses that we have in place. .

Got it. As a follow-up question, just a question about OpEx and your spending. And you've obviously been ramping spending. You've done some acquisitions, which I assume are now fully reflected in the September quarter guidance. What do you expect to be the trajectory of spending from here now that you've had the step up, what kind of color can you give as per spending into next year?

Yes, Chris, no change to the guidance we provided at Investor Day. You should think of it as we have a current scale. Now going forward, it will reflect both obviously Alphawave and Modular in that scale. We are investing in data center as we execute on the CPU and accelerator road map. So you'll see some growth there, but that's the primary driver of how OpEx plays out going forward.

That concludes today's question-and-answer session. Mr. Amon, do you have anything further to add for adjourning the call?

Yes. Thank you. Just before we wrap up, I just want to take a moment to thank our employees, our dedicated employees for their outstanding execution, commitment, especially for a company that is changing, is developing new capabilities. We could not do this without great employees. They are definitely the best part of the whole company.

I also want to express gratitude to our customer partners and suppliers, the current environment for the trust they place on us every day. I think the confidence of our customers ability that -- give us the ability to deliver on the strategy, which has been instrumental to our success today and will be instrumental to our success in the future. And I will talk to you guys in the quarter. Thank you.

Thank you. Ladies and gentlemen, this concludes today's conference call. You may now disconnect.

Good afternoon, everyone, and welcome to Qualcomm's 2026 Investor Day. It's great to be here in New York, and it's great to see so many familiar faces.

Now a lot of you have been asking me recently why I joined Qualcomm. And well, I think it's pretty clear. I think we have a really compelling investment case. And today is an opportunity to really share with you why we're so excited about what lies ahead for Qualcomm. We've got a lot to share with you today.

Before we jump into things, I just want to say a big thanks to everyone involved from Qualcomm and making this day possible. It's a huge amount of work. I really had no idea how much man hours goes into put an event like this on. And just wanted to say thanks to everyone. It's really amazing. And I also wanted to say a big thanks to all the executives from Qualcomm who are here today. You've traveled a long way, and we really appreciate your support. And I also want to thank the modular executives here today. We have Chris and Tim, and we'll hear from them a little bit later.

Now an investor day wouldn't be the same without a disclaimer. And so I direct your attention to this slide, which contains important information regarding our use of non-GAAP financial measures and forward-looking statements as well as the use of Qualcomm throughout the presentation.

So we have a packed agenda today. Cristiano will start the day with a strategic overview. Lots of things happening at Qualcomm, which he will go through. And then we will pivot to data center. And I know you've all been waiting to hear from us on data center. Tony Pialis will -- who's new to Qualcomm and runs our data center division -- will talk you through why we're so excited about the differentiation we can bring to the data center and why we see big architecture changes happening with agents.

Nakul will then address how Qualcomm is moving up the value chain with full stack solutions across automotive, but more recently, industrial AI and robotics, where we see a really big opportunity for Qualcomm over the next 3 to 5 years, particularly in physical AI. And then Cristiano will come back to talk about how agents drive new edge opportunities for Qualcomm, including 6G.

We're also going to share some important updates on the software side. I'm sure you saw the announcement today with the acquisition of Modular. Akash will be our last speaker of the day, he will be bringing everything together and giving some important updates to our financial outlook. After Akash, we will have a short Q&A, and then we encourage you all to join us for drinks behind the curtain in the demo room. And you'll also see some of the new products and experiences we want to share with you today.

And just one more thing. I was actually going to put my [ kilt ] on today. It's a big time in the U.S. with the World Cup. And tonight, Scotland will be taking on the mighty Brazil. And I think Scotland only need 1 point to qualify here. So after 6, you may not see me, I'll be going to join the Tartan army.

So with that, please join me in welcoming Qualcomm's President and Chief Executive Officer, Cristiano Amon.

Thank you so much, Brett. Appreciate it. Thank you. Thank you, everyone, for joining us. I really appreciate it. This exciting day for Qualcomm. Actually, as a matter of fact, June 24, this -- I became CEO of Qualcomm June 30. So it's been -- actually, I picked this week on purpose. This is exactly 5 years since I became CEO of this company. Well, thank you. This incredible company that I kind of joined as an engineer about 30 years ago.

And what I'm going to tell you today and to tell the reason this is special, it's going a little bit about the transformation. 2021, we put a strategy together. We acted an Investor Day in 2021 in that year, and we said that we're going to do. And I think we are at a point now that we put everything in place to execute on the strategy, and it's time to start the new chapter of Qualcomm. And that's what my intent is to tell you about that today.

This is really exciting days. We have a lot of things back when we did the best that we could to pack a lot of big pieces of information in a short period of time. Brett walked you through the schedule, we have Q&A at the end. Many of you will be tempted to run to some memory guy, earnings call starting at 4:30, but I would encourage you to reconsider and stay for the Q&A. It's going to be an awesome Q&A.

So with that, let me just get to the -- with the presentation and before we start to each one of these chapters. So this is where we started. And I think it's been a company that has been probably the most focused semiconductor company in the mobile market. And I know -- I think the mobile market, to some extent, I think, out of favor when everyone in Earth had a smartphone, but it's going to become very interesting.

But the reality is we were the most focused semiconductor company on this 1 market. And we put together a strategy and start executing that strategy, that's what I told you. And the next -- the following 5 years, this is what we built. We built a diversified edge leader across multiple end markets. We built an automotive business. We built we call an IoT, and we're going to start -- I know there's a lot of things within this IoT segment for us. So we started to break down for you, what we call personal AI and compute, which -- how wearables and virtual reality of materiality evolve into personal AI devices, as well industrial, networking and robotics.

And that is the company that we built, really focus on the edge to create a diversified edge leader across multiple industries. And we are now in this transition to what we're going to do in the next 5 years, and this is the next chapter of Qualcomm. And that, there's basically 3 pillars to it. There's actually 3 dimensions to the Qualcomm future.

The first one, and I believe many of you came here today to see what we're going to do in the data center. We're building a data center platform. It's a comprehensive portfolio of solutions. I think you should think about it -- what we have done in the past few years like a submarine strategy. We've just been executing, executing, collecting assets. And when we get to this point, we feel that we have a comprehensive portfolio to enter the next phase of the data center. And you hear about it as inference scale as we see the segregation.

A lot of people ask me this question. And before you ask in the Q&A, I'm just going to even answer. A lot of people ask, "Oh, this crowded market, is it too late?" Never too late for Qualcomm, I think, because this is a market that moves very, very fast. So if you have technology leadership, there's always room for you. And I think that's how we think it, people cannot like the company about many different things, but I never had anybody that said Qualcomm does not have a solid, I think, technology capability. And I think that's what we put into work.

The second part is now that we have been building this platform of devices at the edge, we're moving to a full stack player. So as agentic start to transform the devices on the edge, we're going to build on those assets. And it is one of the broadest portfolio of semiconductors, and we'll be kind of a full stack player in physical AI, compute, everywhere. And that's what you're going to see in the presentation today.

And then it gets to the part number -- actually, I wanted to -- before I go to Part #3, I wanted to break that down. There's actually one more thing before the full stack execution in automotive, industrial network and robotics.

It's what's going to happen with our mobile business. And I actually have a portion of the presentation today to tell you how you should be thinking about the future mobile edge devices that consumers are going to utilize, and that's part of the second dimension of Qualcomm.

And that gets to the number 3. And then number 3, is how we're going to go from silicon to platform solutions, fully integrated platforms across hardware, software developer ecosystem also making Qualcomm a developer-first company. So that is the 3 dimensions of this new chapter of Qualcomm and what we've been going to be doing in the next 5 years, and we're going to show you some of the early wins that we have on this.

And I want to summarize to you as I get off the stage, what is the Qualcomm advantage. We're always very proud of our technology and IP, but we're also a company that can build very broad relationships. That's one of the assets of the company. We have built strategic relationships across industries. We were not a player in many of the new industries that we enter. And if you look at today, Qualcomm built very strong relationships with all of the leaders of the industries, and we built an ecosystem around it.

The other part of it, and that's why when people ask about if it's late to enter the data center, you think about scale and execution. Our engineering capabilities, our operations and supply chain is one of the leading scale and execution machines in the semiconductor industry.

When I say about the technology advantage of Qualcomm it's really very broad. It's really a broad portfolio.

And we take pride in providing leadership solutions. We have been, in many areas, a creator of standards. And that with everything connectivity was just wireless, now it's wireless and wireline, including connectivity in the data center. And that is the portfolio that we have been building. It's every form of compute.

When data centers talk about disaggregated computing, this is the reality of mobile from a very long time ago. So we have every disaggregated compute -- and actually, we do this for different industries, including what we do on safety and industrial-grade processors. We're building a comprehensive software and developer ecosystem, and you're going to hear more about that today.

Sensing becomes even more important when you think about physical AI multimedia, advanced packaging memory. And we have a very strong patent portfolio, which is a result of about $100 billion cumulative in R&D investment. Our customer reach -- and you should be thinking about those concentric circles. We started building an ecosystem partnerships in mobile, and you see as we go into every other industry has now expanded. And that is a unique asset of our company. Our customers believe in Qualcomm, and that is true, especially when we enter new markets, and that's what we're going to be doing right now in the data center.

And the last part is the execution. I will spend a little bit on this. We're actually a proud member of the TSMC fabless ecosystem. We work with a large number of foundries and manufacturing partners, and we have incredible scale.

I'm just going to give you a few highlights. We consume over 1 million leading node wafers, this is just annual scale. We do over 75 chips, tape-outs per year. Of that 75, over 30 is an advanced transistor. We ship 40 billion components, total of 2.5 million wafers. That's a lot of scale.

And we also have some things that are very unique. Some other semiconductor companies actually tell us, Cristiano, you're crazy. Sometimes we tape out a chip. Before we get the chip back, as soon as TSMC finishes the mask, we go to production, and we go to production at scale. That's the maturity of our manufacturing capabilities, and we ramp completely new nodes to 100,000 wafers. That's the nature of the phone business within about 2 quarters. There's a very big ecosystem. And I think that is an asset that we've also been bringing to the table.

And the unique thing about Qualcomm -- and I said about this in Computex is we are now present across the entire compute continuum, from sub-2 milliwatts to about 200 kilowatts. And if this gets transformed with AI and we apply our technology and we build on a developer-friendly unique software platform. I'm sure you're going to hear more about that -- that creates an incredible opportunity for the company not only for the business that we have built, but what we're going to build in the next 5 years.

So that's our strategy. This is what we're doing. And now what I'd like to do is start unpacking that to you. I think I want to bring to stage Tony, who joined us from Alphawave and now runs our data center business. And I'm sure you're going to be amazed about what we're doing and what Tony is going to share with you. Tony, please come on stage.

Hey, everyone. I'm Tony Pialis, General Manager of Data Center for Qualcomm. First of all, where is Brett Simpson? Brett, I thought I told you. I never want to follow Cristiano. I mean, in terms of being a visionary in the industry and a legend, I'm so lucky to be part of his team.

So folks, one of the most common questions I've been receiving this morning as I've been meeting with many in the press is why would I want to leave my role as Founder and CEO of a semiconductor company to join Qualcomm to run their data center business? For me, look, I've always been good at math. The equation for that has been really simple. It's all about accelerating value creation. That's what we're here to do. That's Cristiano's vision. That's why I came, and that's what I'm here to explain to you today, how are we going to accelerate value for all of our shareholders and customers. So let's get to that.

Agentic AI changes the economics of compute. What does that mean? It means token counts are skyrocketing as we introduce agents. It means CPU attach rates are soaring through the roof. You can't find CPUs anywhere. They've already been bought up. So traditional infrastructure will not scale to the needs of agentic AI. So the industry needs a paradigm shift in order to deliver this. I will explain this paradigm shift and how we will lead it.

But first, let me introduce to you, Dragonfly. This is our data center infrastructure to lead the industry into the next stage of AI. Let's hit it.

[Presentation]

Well, Cristiano, the submarine is surfaced and here we are today in New York. All right. So team, agentic AI requires a new compute infrastructure. This chart illustrates why. The industry has blown through Gen AI and reasoning, and here we are in the cusp of deploying agents en masse.

But what does that mean? It means a single agent queries generating 50x to 100x inference requests. We have over 1 million tokens being generated by a single query. Traditional compute infrastructure cannot support the scale. A paradigm shift is needed.

So first, what you see up on the screen, which by the way, I love the screen, I need to get one in my basement. What you see up on screen is a traditional GPU-based compute that's been deployed across data centers worldwide. It's been built to support both training and inference. It runs hundreds of kilowatts today. Other competitor solutions will be running north of 500 kilowatts.

How can we deploy this and support 100x more inference calls than we do today? We can't. A new solution is needed to lead the industry. And so what we and our submarine have been doing is building a world-leading disaggregated compute infrastructure. Cristiano used the analogy, well, we already did this in mobile. That is what's needed to lead in data center, bespoke solutions that deliver hardware acceleration for each and every function needed to deploy agentic compute. Various forms of CPUs performing specific functions, unique XPUs, some targeted to attention for prefill, some targeted for KV caching during decode. All of this blaze together using both copper and [ optical ] interconnect that deliver world-leading connectivity, creating 1 compute network that will transform the industry.

Now as I've also been asked earlier today, how is Qualcomm positioned to win? Aren't you late? Let me give you some background. I've been formally in the company for the last 6 months, but I've worked with the company for years as a supplier and partner to them. What I will tell you is this is an engineering-first company. They blazed the trail and led the world in wireless communications.

Fast forward, they lead in mobile compute. Now we are leading in both automotive and PC. When the company turns its attention to solve a new problem, we revolutionize the solution and push our way to the forefront. And folks, I am here to tell you today, that is what we will do and are doing in data center. We are pushing our way to the forefront, and our customers are pulling us the rest of the way in.

So how are we going to deploy our agentic infrastructure? 4 simple steps. The first one starts right now. We have our connectivity portfolio that comes from Alphawave. That is in high finally. I've been talking about getting this into production for the last 2 years. I am proud to say this year, we are qualified in our first leading major hyperscaler, generating meaningful revenue for the company over the next 4 quarters. Follow that by custom silicon. Lots of speculation, and I'll have a lot more to say, so you can hold your questions. We are winning in this space, and we'll be delivering meaningful revenue at the end of this year, calendar year, starting fiscal quarter 1, 2027.

Then in 2027, we are launching our third generation of AI accelerator. What makes the third generation unique is it's the industry's first near compute AI accelerator that will transform inference. A lot more to come on that. And we're not done yet, folks. You layer on top of that, in the middle of 2028, we will be launching the industry's first Oryon server-class compute solution. We will launch a fleet of a agentic, general purpose and head node compute that will complete the Qualcomm infrastructure.

Now let's get to it. So this is the best worst kept secret or the worst best kept secret, however you want to view it, that's limiting the industry. I've been in the industry for 30 years. Compute has increased by more than 60,000x in that span. Kudos to the amazing Qualcomm engineers for accomplishing this.

But guess what, where transformer sizes are growing 240x over a span of 2 years these days, look, compute memory is only doubling in that same time span. So what does this mean? It means there's no point packing more compute unless we solve the memory bottleneck.

So before I show you how we do that, what you see here is how a modern day XPU works. You see that GPU off to the side. It is pulling and pushing data to a HBM stack. You have thousands and tens of thousands of wires constantly carrying data, generating tremendous amount of heat, burning significant amounts of power, moving data back and forth in order to support the scaling transformer sizes. This solution cannot keep up with the growth in AI models. So an innovation is needed, and this is what the submarine has been doing.

This is what I was most pleasantly surprised to find when I joined Qualcomm. We have broken through the memory bottleneck. How have we done it? We have rearchitected compute for XPUs. We have separated the AI accelerator from the XPU. And what you see, we now put our XPU right under a DRAM stack.

What does this mean? Very, very important. We offer all of the performance advantage of SRAM, but with the density and the memory capacity that HBM stacks offer. And so what does this mean to the industry? It means that congestion that you saw with HBM is gone. What happens now? A great analogy I use with reporters is imagine working in the same building that you live in. And so you only travel up and down. And what does that mean for the highways and the roads that connect the suburbs to the city? Guess what? The roads are clear. So the value this brings to the industry is lower power consumption, less heat. And that expensive road of silicon interposer that HBM solutions use are no longer needed. We can deploy multiple HPC stack within a single compute device using standard packaging. That is a tremendous value that we deliver to the industry in terms of performance per cost advantage.

Now how do we quantify these benefits? Look off to the side. These are ultra low latency workloads, things like coding workloads. HPC offers 200x capacity per watt, better solution than SRAM. There are many that are announced seeing SRAM-based solutions today, look what we can deliver with HPC. Now you look at the other end of the spectrum for high-throughput workload. These are the workloads that GPUs and HBM have dominated for years. With HPC, we deliver 6x the bandwidth per watt versus competitor HBM-based solution.

And so while the rest of the industry is now trying to deploy 2 unique solutions to solve these bookmarked endpoints -- and those in the middle, well, guess what, they're kind of stuck in the lurches. With HPC, we offer a single solution that can seamlessly span this entire sphere of workload and deliver multiple fold performance per watt and performance per dollar benefit. That is a direct TCO advantage that we offer to the industry.

So I go back to those asking questions. Are you late to the game? With this kind of performance advantage, the industry is demanding and pulling us in. So who better to introduce HPC to the industry than one of our lead partners, Microsoft? So it is my greatest pleasure and my first but not only surprise to you today to introduce Satya Nadella, who will talk to you about how Microsoft will be deploying HPC within Azure data centers. Let's hit it.

Hello, everyone. is great to be back at Qualcomm's Investor Day. At Microsoft, we have had the opportunity to partner closely with Qualcomm across multiple waves of computing, from the PC to mobile, and now, AI. And across all of them, we have shared a deep commitment to innovation at the systems level, bringing together the silicon and software to deliver meaningful advances for our customers. This includes our continued collaboration to reinvent the PC for the AI era, which will only become more critical as we deliver unmetered intelligence at the edge with Windows.

And we are not stopping there. In fact, with Project Solara, we are collaborating on a new platform purpose-built for agent-first devices. And it's been fantastic to see the reception since we announced it together earlier this month. And now we are excited about your innovation in the data center, especially around high-bandwidth compute, and we look forward to building on that together. HBC implements an innovative architecture with high memory bandwidth and integrated compute that unlocks significant improvements in cost and performance for the next generation of AI infrastructure, and there's so much more to come. We look forward to our continued partnership as we build next-generation of computing together. Thank you so very much.

Folks, if I had a microphone rather than a lapel device, I would drop it right there. So that's our first surprise. Stay tuned, many more on its way. All right.

So how are we deploying HPC? Well, we're deploying it to target a $680 billion addressable market for us. HPC earns us the right to win a significant portion of that market over the next few years. Look on the left side of the chart. To quantify Satya's comment in terms of performance benefit deploying HBC, we deliver, depending on the workload, anywhere from a 4x to 8x advantage. That directly translates to TCO advantage that we deploy to our customers.

I am super excited for our launch of our first HBC product in the middle of 2027. That's it, folks. Take those pictures. You're going to be seeing this to last for years to come. Our AI250 product will introduce the first near-memory compute that employs HBC and will be a complete game-changer to the industry. We're following that in 2028 with AI300, which launches our second generation of HBC. It will integrate [ UAL and East ] on the latest scale-up network fabrics. And for scale-out, we will be deploying both copper and optical networks to connect AI clusters.

Now look, as amazing as this hardware stack is, it really is just a foundation for running software. Software is where the magic is, and it takes a lot for me, a hardware designer, to acknowledge that. We will be deploying a full software solution stack that includes the most sophisticated orchestrators that will manage and route the traffic across a disaggregated compute cluster, all the way down through frameworks. And most importantly, open frameworks that will allow model developers to both develop and deploy at scale their models. We offer all the kernels and compilers that are necessary in order to optimize the latest models onto our specific hardware.

While others in the industry build moats trying to protect their hardware deployments, myself and we at Qualcomm, we believe in building bridges to unite the industry. And so my second surprise, which I know came out through an announcement, but it is still my honor to announce that today, we published an announcement for Qualcomm to acquire Modular. Modular is a world leader in developing AI software solutions. And who better to introduce how we will jointly transform the AI industry with our open solution to disaggregated hybrid compute than Tim Davis, Co-Founder and President of Modular. Tim, will you come up on stage?

Thanks, Tony. And you can obviously tell between Cristiano and Tony. That's why we're so excited to be joining Qualcomm.

But hey, everyone. I'm Tim, one of the co-founders of Modular, and I've been building AI infrastructure for almost a decade. First, at Google Brain for 6 years, building core AI data center and edge infrastructure for mobile devices and TPUs. And then at Modular as Co-Founder and President.

Modular has assembled one of the best teams in the industry that have helped found, build and contribute to most of the core AI infrastructure in use today. Now you may have read this morning that we are beyond excited that Modular is joining Qualcomm to supercharge our AI infrastructure and distribute it to the world.

But you might be wondering, what is Modular? Well, I'm here to tell you, Modular is building AI's unified compute layer, a software layer that enables AI models to run on any hardware and is heterogeneous by design. For developers and enterprises, that means building once, deploying anywhere, lowering the cost of running AI at scale and accelerating innovation into their production data center workloads.

Modular is the portable alternative to NVIDIA's software stack, designed from day 1 for every AI accelerator.

Let's walk through the stack. Mojo gives developers the high-performance, low-level programming model they need without locking them into 1 platform. MAX gives them the model in serving layer that they need without relying on [ Triton ] or TRT LLM. And Modular Cloud gives enterprises the distributed serving infrastructure they need without being tied to a single silicon vendor.

Together, this is a full AI compute platform for the heterogeneous data center. And we are up to 50% faster when executing AI inference workloads on third-party hardware. And we have the numbers to prove it. After 4 years of R&D, Modular is rapidly coming to market with industry-leading performance across many of the world's most foundational AI models.

Now importantly, our platform turns heterogeneous data center systems into multi-silicon AI token factories. In this world, enterprises, partners and developers can use the best silicon for each workload without being locked into a single hardware stack. Because Modular is heterogeneous by design, the industry can achieve lower TCO, higher performance and greater portability across the world's compute infrastructure.

We are incredibly excited that Qualcomm will help us scale our technology to data center customers everywhere, enabling broad hardware independence for the world. Chris Lattner, my co-founder at Modular, will share more about our incredible future with Qualcomm later today. Back to you, Tony.

Thank you, Tim. Look, rather than preparing for this event, I've been fielding calls all morning from hyperscalers and customers asking, "Hey, how can we begin incorporating Modular's technology." So I'm super excited about what we will be doing together.

So now let's talk about CPU technology. Qualcomm has a long lineage in leading in CPUs. They pioneered mobile compute with Snapdragon. We are winning now in both PC and automotive. The company's focus is now transitioning over to data center.

And so today, there's been a lot of speculation about this, but I am excited to introduce Qualcomm's C1000. It is a data center fleet of processors. These processors will run the industry fastest cores, running greater than 5 gigahertz. This is more than 30% faster than any of the competition. Coupled to that, we offer more than 250 cores to run the highest throughput workload. Combine that with Alphawave's leading [ PCI-Express ] technology delivering greater than 2 terabytes of IO bandwidth.

Then add on Qualcomm's memory leadership delivering the highest performance, lowest cost memory solutions, employing LPDDR. We have server-class [ RAS ] security embedded directly in the hardware. And finally, our CPU is also AI native. That amazing HBC technology that I walked you through for our AI inference engines couples directly as an HBC attach to accelerate AI workloads natively onto the C1000. Now how are we deploying it? Through 3 various product lines for the C1000. The first is our agentic CPU. Leveraging our HBC attach, we deliver industry-best performance. We then deploy it through general-purpose CPUs running virtualized container workloads. And finally, our AI head node CPUs running and orchestrating all the traffic across disaggregated, heterogeneous compute data centers. All of this targeting a $200 billion market, and that number is growing every day with each and every analyst report that gets published.

So my next surprise for you, I'd love to introduce to you Mark Zuckerberg, Founder and CEO of Meta, as he introduces how Meta plans to deploy the C1000 into its next generation of data centers. Hit it.

Hey, everyone. Great to be here at Investor Day with you. Meta and Qualcomm have been partners for a long time, and we're doing some great work together. We first started on the Quest headsets, and then we brought Llama to Snapdragon so people could run AI right on their phones. And today, Snapdragon is powering our AI glasses, too.

Now we're bringing that partnership into our data centers. With our latest model, Muse Spark, we're delivering AI to billions of people every day across our apps. The data centers, the energy, the compute to run billions of model inferences, that's what makes it all possible. Our goal is to deliver personal super intelligence to everyone in the world. And as our teams work hard to build state-of-the-art models, we need to innovate with how we get the power we need, scale it and make it accessible to everyone.

So that's why our work with Qualcomm is so critical. They've spent decades figuring out how to get the most performance out of every watt. They're really good at it, and now is the right time to expand this partnership. So today, I'm excited to share that we've entered a multigenerational collaboration for Qualcomm to supply CPUs for our data centers and help power our next-generation server fleet. This will help put personal super intelligence into billions of people's hands.

There's a lot more to come, and I'm looking forward to building together for a long time. So thank you to Cristiano and all of the teams at Qualcomm for all the work you do here.

Satya and Mark already in my presentation, and folks, I'm not done yet. So let's move on to the third product line, custom silicon. There's been rampant speculation in terms of what we're doing here.

First off, I want to establish, we are in custom silicon to target the highest tier of customer where we can deliver the most value add by bringing our incredible IP portfolio to play. So our wins to date are based on both Alphawave legacy wins that are scaling into production. And most interestingly, in the first 6 months here, I am extremely excited to announce we have won 2 major hyperscaler deals that will contribute meaningful revenue to Qualcomm starting at the end of this year.

And so folks, how do we win in custom silicon? We work with our customers. We take their specs, and we help them build their chips, whether it's in the front-end RTL design or whether we help them convert their designs into chiplet-based solutions, delivering the most advanced computer networking solutions in the world. And then using our manufacturing scale and know-how, we optimize their yield and enable them to deploy their bespoke solutions en mass to their data centers.

I've been in this space for 30 years. The way you differentiate and win in custom silicon is through your IP portfolio. We have the world's best custom silicon IP portfolio. We have our own compute that we can optimize for our customers. We have HBC, a complete game changer in the AI industry. Add on to that, Alphawave's leading electrical and optical SerDes that kicks the butt of its competitors. We've been working in the silicon photonics and optics space for more than 5 years. We bring that to play in order to bring connectivity directly into compute. And you couple that with Qualcomm's leading manufacturing and supply chain. This is how we've been winning in the first 6 months.

I have not had to push my way into hyperscale customers. They've been pulling us in. And when they pull us in, it gives me a chance to expand and bring the rest of my solutions to play.

The final product line I will walk you through is our connectivity. This is the third bottleneck in the industry. The first was memory. We solved that with HBC. The second was the software stack. Modular, Tim and Chris will help us solve that.

The third is connectivity, and this is near and dear to me. This is where hyperscalers, as they deploy clusters of compute, as AI compute doubles, so does connectivity every 2 years. And then you have this transition from copper cables to optical solutions with new low latency scale-up and scale-out fabrics. This is a race to the forefront, and we have the technology pieces needed to win.

We have everything you need to scale from the millimeter of connectivity all the way through to tens of kilometers, from our leading die-to-die technology to our co-packaged optics interfaces that bring the world's fastest lower power connectivity right next to our compute. Add on to that, leading PAM4 electrical and optical SerDes to drive scale up and scale out networks. Today, in production at 224 gig, soon, we'll be in production with 448 gig.

And finally, Coherent-lite today, it drives optical connectivity across campuses. But when PAM4 runs out of steam, Coherent-lite will be connecting compute clusters within the data center. So with all these pieces, I'm excited to announce our connectivity portfolio. We are already in production with our first generation of 800 gig electrical and optical DSPs, including our first generation of Coherent-lite By the end of this year, we will be in production with our second generation of electrical and optical solutions, deploying 224 gig solutions.

These products are anchored with the lead hyperscaler win already. And then looking forward to 2028, we will be bringing in our third generation of connectivity solutions. This will be based on 448 connectivity. And we'll also deploy our next generation of Coherent-lite solutions.

So folks, stepping back now. We have developed a transformational infrastructure that is already winning in the industry. Four product lines, each of them already anchored with multiple customer wins and a pipeline that will blow your heads in terms of accumulated value. Incredible metrics. Look at that, up to 8x better tokens per watt per second than traditional GPUs, greater than 200x memory capacity compared to SRAM solutions, 6x memory bandwidth per watt. And for our CPUs, greater performance than 2x than our competition. All of this is direct TCO advantage.

So why are we entering now? Because we have the performance that the industry needs. So folks, final takeaways I want you to remember. Tokens per watt replaces FLOPS. The race has changed. Embedded solutions, embedded providers, they're playing the old game. There's a new game in town, and it's all about delivering agentic first rack scale platforms that delivers the world's best TCO. That is what we've been building in our submarine.

Add on to that, 4 product lines that we've already anchored with hyperscaler wins. And now with today's announcement, we have the world's best software stack that will build bridges across all the hardware of the industry. And finally, and I will leave the numbers to our man, Akash. I am very, very proud to announce within the first 6 months on the job, we will deliver multiple billions of revenue starting fiscal '27. And for those of you that aren't aware, that means starting this calendar year.

And so I am excited now to introduce my colleague, Nakul. But before he comes on stage, I have one last person I want you to hear from. It's Tareq Amin, the CEO of HUMAIN. He is a visionary in the industry, and he has been our first data center customer. Let's hear what Tareq has to say. Let's hit it, and thank you.

Congratulations to Cristiano and the entire Qualcomm team on this bold milestone. AI is no longer a technology trend. It is becoming the operating system for every industry, every economy, and every society. I believe the next decade belongs to inference, billions of agents, trillions of interaction, continuous intelligence operating across devices, enterprises and government.

Through our collaboration, HUMAIN and Qualcomm are deploying the next-generation AI infrastructure by combining Qualcomm breakthrough in semiconductor innovation with HUMAIN full-stack AI capability from infrastructure, cloud platform, foundation model and [indiscernible] AI system. Success will not be measured by peak performance alone. It will be measured by performance per watt, performance per dollar, performance per outcome.

This is where Qualcomm brings something extraordinary, a really fundamentally different approach to AI compute that challenge really conventional assumptions about power consumption. Years from now, we'll look back on this moment as the beginning of a new era for AI.

Good afternoon, ladies and gentlemen, and a very big round of applause for Tony, first of all. Nakul Duggal here. I run Qualcomm's automotive industrial and robotics businesses, and someone who's been with the company for over 30 years. There has not been a more exciting time to be at Qualcomm.

Each of these 3 businesses are quite different. They're quite unique, and they need different strategies. But over the years, to diversify Qualcomm, we've had to build new muscles that strengthens over time. We see the next several years belonging to physical AI and massive transformative change that physical AI is going to drive in industrial and enterprises and especially robotics becoming a key catalyst.

You've seen what we've done with auto. I'm going to give you a sense over the next 30 minutes, how we are preparing for physical AI. Physical AI is the next great computing wave. It doesn't run in the cloud. It runs on the edge. It's going to run in factories and warehouses, in retail, in hospitality, in hospitals. And robots are going to be a very important part of physical AI.

Automotive is the first example, and you've seen what we have done with the automotive business. Industrial and embedded, I gave you an update on this space in November of '24. And we've been preparing and I'll give you a sense as to how this business is progressing. And robotics is a space that we entered only in the last 9 months or so, and we have expanded very quickly into multiple environments. We are finding strategies to move up the stack.

One thing that is very unique about Qualcomm. While these are 3 unique businesses, there is one single common theme, one IP road map, one product foundation, one physical AI platform. If you look at the way physical AI is now moving into our lives, you've seen human facing AI. You've started to now see machine facing AI and ultimately, embodied AI. These 3 layers are highly interwoven, and they compound over time. And you will start to see how this plays out. As I walk through automotive, industrial and robotics, hopefully, this will all become clear.

But they're all underpinned by the same technology layers that create vast automation capability. Humans facing AI has changed our interaction layer, first with chatbots and digital assistants, but now with body cameras, with XR glasses. And you are changing the physical space interaction layer between humans and the devices, the products that they own.

Instrumented AI, machine AI is about putting AI that is embedded into sensors, embedded into cameras. And it mostly comes down to that sensor, that endpoint being situationally aware. The real economic unlock is, however, physical AI and embodied AI. And what we see here is the ability for devices to perceive, to reason and to actuate with the goal to be able to complete a physical task, and that evolution is just starting.

We find ourselves at this inflection point. And as this matures, we see a massive edge content update cycle hit us. This content shift across the edge is expected to be pretty significant across automotive, across industrial and across robotics. In automotive, over the next 7 years, you will see 500 million vehicles produced that will have AI cockpits that will have anything from L2 to L4 autonomy. This was not the case if you look at what was getting deployed, what was getting shipped over the last 5 years. 50 billion IoT endpoints by 2035. And over 1 million robots will get deployed globally.

What is today a $300 billion addressable market is going to become over $1 trillion within the next decade. Right now, while we see automotive and industrial as very large TAMs, we expect this to invert, where robotics will actually become very large. This is the market that we are going to lead.

Our ambition is simple. We need to own the solution, the silicon, the software and the stack for physical AI. We will build full-stack platforms where there's white space. Our operating mantra and you've seen this, we will win in automotive. We will disrupt industrial, and we will define robotics.

Let's get started. We introduced our first generation of automotive products, especially compute products 10 years ago. Today, we are one of the largest automotive compute and advanced connectivity players globally. Five generations of compute silicon delivered in 10 years. Today, from first silicon to start of production of the vehicle, we have brought that time line down to 15 months. This is as fast as consumer product life cycles.

We now have over 500 million Snapdragon cars on the road, 90 million cockpits powered, and we only entered the cockpit business in 2016. We have launched 450 new car models since 2021, which is 2 new models every week for the last 5 years. The Snapdragon Digital Chassis is the underpinning of vehicle compute and connectivity globally. There is not a modern vehicle that is built without the Snapdragon Digital Chassis.

We will exit FY '26 at $6 billion in annualized revenue. And this is after delivering 23 consecutive quarters of double-digit year-over-year growth. We have built a $65 billion design win pipeline. Our content value from Gen 3 to Gen 5 has uplifted 8x. We are engaged with over 70 automakers and over 100 Tier 1s and Tier 2s globally.

This is what true diversification looks like. We are a systems company, and we have shaped the automotive industry, its platform architecture across hardware, across software, across compute, across AI, across opportunities like ADAS. And we will do this consistently over multiple generations. We are, ladies and gentlemen, on track to become the largest automotive semiconductor supplier across all pure-play automotive.

If you look to the right, we have built leadership in cockpit because all of the value, all of the IP, all of the differentiation that Qualcomm brings, all of the breadth of access that we have to so many ecosystems. As we start to see AI coming to our lives, AI is transcending the traditional domain architecture of what is a cockpit and what is ADAS.

So we designed for this. Gen 5 was designed keeping in mind that we can't really be traditional in the way of thinking. We're designed for a mixed criticality fabric. That means the customer can run cockpit applications, ADAS applications as they feel. They can run them separately, and they can run them together.

As we build more powerful chips, that necessitates customers to be able to figure out how do these architectures change. And this is creating tremendous optionality for customers to be able to figure out how do they design the next generation vehicles. With AI now, you can process any sensor input across any specific vehicle domain. You don't have to tie the physical hardware to a specific domain. We are running 30 billion parameter models on the cockpit today already, commercial. Concurrently, we can run L2-L4 stacks.

And then as you see on the left, what we used to call the software-defined vehicle, SDV, last year or the year before, this has now become an AI defined vehicle because we can now run agents directly on top of SDV that get access to vehicle context. We now have use cases where a car drives into a parking lot, sees a QR code, scans it, pays for it, and that's an agent. This is how quickly the AI-defined vehicle is moving..

ADAS was a new space to us about 3 years ago. We didn't really have any customers to talk about either in the SoC space or in the stack space. We are now at 25 OEMs. The open platform strategy that we have built with ADAS has allowed us to be able to provide tremendous optionality to customers because this is a complicated business. It is about safety, it's about cost sensitivity, about which part of the world you're deploying in. We have a dozen different stack partners that we have engaged with, and we are building our own stack as well.

Now Snapdragon provides you the best performance per watt per dollar capability across the industry. And that is why customers are moving to our platforms. In parallel, we have built the Snapdragon Ride Pilot stack ourselves, we debuted this last year with BMW, and we are now validated in 60 countries. Stellantis is the latest OEM who has picked not only our Ride Pilot stack, but the entire Snapdragon Digital Chassis, which we will deploy starting SOP '28.

As automotive is evolving, we see a tremendous amount of new growth opportunities. Robotaxis are things that you might have questioned how -- when will the robotaxis be real. They're starting to happen. We do expect that by the end of this decade, we will start to see these scale. And our strategy with robotaxi is actually very straightforward.

Tony talked about HBC. We will actually build accelerators that will connect our SoCs and HBC Gen 2 to provide that same tech to our automotive customers, and we are starting to plan to go to them in the '28 time frame. We're also seeing this very interesting transition around token generators inside the car. As automakers are putting in so much of compute and so much of memory to be able to run models locally, we are receiving requests to see an offline mode. Could those be part of a federated use case for token acceleration. We will use the same exact HBC capability for token acceleration in the car.

The other area that we are seeing a lot of interest is in AI/ML use cases for the car, for the powertrain for the drivetrain for battery management. But there are so many domains in the vehicle that need local machine learning processing. We acquired a company called Edge Impulse a year ago, and we are actually running their MLOps locally in the vehicle. That allows us to use the Snapdragon NPU for local AI compute. Same exact chassis, no big difference, just add more local processing capability.

And finally, we are starting to see a buildup in satellite connectivity need. We've had a tremendous telematics and connectivity portfolio over the years. Customers are now looking to add satellite to that as well.

So here is why we keep winning. We operate at the full system architecture of a car. We support a global footprint. We are multigenerational in our silicon road map. We go across every tier. We go across every domain. We are compatible across generations. And we allow customers to plan a decade against our road map. We are building an ADAS, L2+ stack ourselves because we know that this technology is going to be standardized across every vehicle.

We have built the deepest and widest software and AI stack in the industry. We partnered with Google. We've integrated Gemini. We partner with every global digital ecosystem, and we bring in OEM preferred ecosystems. We've built years of safety expertise. We're a smartphone company, but we have made this transition, we have diversified. We now build safety as part of every single chip, every single piece of software. Even our stack, our tools are all safety grade.

And we've built tremendous supply chain complexity and resilience because we have become one of one in the automotive space. We understand supply complexity, capacity complexity, reasonable geopolitical complexity, and we've been able to scale this business up very well. So a big thank you to the Qualcomm team who's been driving this for years.

Automotive is a playbook for diversification for the company. And we've gained leadership, we have gained tremendous scale, and we've built a multigeneration strategy. And as AI is upon us, we find ourselves very well prepared for the transition.

Now we've learned a lot about how to diversify through our automotive experience. And I'll share with you what we have done in the last 18 months in the industrial and embedded space. Before we get started, if you look to the slide, the OP or the operational plane in any industry, in any enterprise that has traditionally never had to do any processing at the edge. It was always about capturing information and sending that information in the cloud. It was mostly deterministic. It was mostly static. And the concept was send data to the cloud, and processing happens in the IT layer.

Now as you start to see AI come into the picture, it's really the same concept. You have data at the edge, you have enough information at the edge to be able to process to get to a specific outcome, whether it's to detect an anomaly, whether it's to extract specific analytics. The operational technology plane is where these endpoints are getting deployed, and that OP plane is getting rearchitected. Even brownfield settings are now being rearchitected because you can add an intelligent AI aggregator. This architecture creates a once-in-a-generation opportunity for the entire OP plane to become more intelligent, more data, smarter models, better insights and more intelligent endpoints.

So we started to build our road map for this specific space. If you look at what we have built with Dragonwing over the last 18 months, a variety of different solutions that are very vertical focused. So we have dived very deep into what our vertical customers need, and we have built solutions based upon that. We have a silicon road map that addresses connectivity, camera, commercial processors, industrial processors. These power everything from AI boxes, connected industrial gateways, edge appliances, industrial PCs, payment terminals, smart home appliances and drones, body cameras.

And we picked 3 vertical categories that we focus on: industrial, commercial and mobility. And these are further segmented across 12 verticals. So we are building for every device class, every connectivity standard and every stack layer. Let me give you an example of what we have done in the vision space. So we believe that vision is the major unlock in industrial. We've had tremendous expertise in the company from our smartphone heritage, from our automotive heritage, and we are building industrial machine vision. We are building robotics, and we are building surveillance as 3 additional layers of capability.

Now vision AI is a major unlock because the physical world is best processed through the lens of a camera frame. We understand lighting conditions and how to improve them, we can reconstruct scenes in 3D. We can semantically annotate these scenes. We can feed them to a VLA so that it can tell us what the artifacts of interest are. We can even predict what the next scene needs to look like in a situational awareness scenario.

And so we've built an entire video AI stack, from camera chips, edge AI boxes, on-prem appliances to a full video AI service. And we are deploying this across every vertical, retail, small and medium businesses, smart cities, venues, any use case, all verticals. We believe video intelligence is a major edge play.

The other area that we had to spend a lot of time on was to figure out how do we become much more developer-centric, much more developer-first, as Cristiano mentioned in his remarks. And these were two problems. One was, how do you simplify access to the product? And how do you accelerate the journey from prototyping to commercialization?

As Tim mentioned in his remarks, Modular is going to be a significant game changer for Qualcomm because we will now be able to write and serve models faster. We will also be able to build application-specific acceleration in the libraries that Modular is going to bring to be able to make our entire AI stack that much stronger.

Over the last 18 months, we made 3 additional acquisitions to be able to be very developer-centric. One was Arduino. Arduino brought us 33 million developers, massive global footprint, completely open source, allowed us to be able to get access to pretty much every vertical out there. We acquired Edge Impulse. That allowed us to be able to get model training and tuning and containerized development of models at the edge. And we acquired Foundries, which allowed us to be able to manage industrial-grade Linux. So developers now have the ability to prototype rapidly on Arduino and Dragonwing, and we scale these projects across standardized system on modules or chip-on-board capabilities across all of these various use cases that I mentioned.

Last October, we launched Arduino UNO Q, which was the first Dragonwing product on the Arduino ecosystem with tremendous success really across every vertical, every type of developer and consumer market. We are able to launch VENTUNO Q in August this summer, 40 TOPS of AI, octacore, 12 cameras, safety island built-in. Built-in AI models, part of this overall ecosystem that we talked about. It will focus on industrial, consumer and embedded. And, of course, robotics.

This runs full upstream Linux. We have not been a company that has been on that path. We are now a full upstream Linux company, and these will run out of the box. So all of the goodness of the Qualcomm platform available in full upstream Linux.

We are also working on agentic development. So you can essentially take these development boards and code directly with Claude, with Codex, with Cursor. and start by coding. And you can buy these on Amazon.

Qualcomm is changing, and we are becoming massively developer-centric. So while we were pulling together this product portfolio, this developer centricity, this video AI, we were also, in parallel, building out focus on verticals. We have 12 different verticals across 3 major industry types, and I'll give you a sense as to what we are doing in 3 of them: retail, energy and utilities and oil and gas.

But we are building blueprints so that they are repeatable. So we understand what the OT blueprint is that we need to be able to replicate. And we are leveraging the entire Dragonwing portfolio for this. We've even built a solutions engineering team for Dragonwing.

Let me step through a few of these quickly. In retail, the store is evolving. It is continuously sensing. It is making decisions by itself. It has to act because stores are now hybrid. Some stores in the evening are almost not manned. They need to have more autonomy. There are AI cameras that are getting added for loss prevention, for shelf intelligence, electronic shelf labels, RFID for dynamic pricing. AMRs and robots for automatic restocking, for removing products that have expired, for cleanups. And we are partnering with key partners in retail like VusionGroup to be able to drive this expansion.

Similarly, in energy and utilities, Qualcomm addresses every step of the value chain, generation, transmission, distribution, the entire grid. We go after sensors, industrial gateways, fixed cameras, meters. In our hometown in San Diego, we worked with San Diego Gas & Electric to help mitigate wildfires with autonomous drone inspection to execute automatic power shutoff. Schneider has been a great partner of ours. We've built with them, capability to be able to deploy industrial gateways as part of their substation concentrators.

In oil and gas, we have built solutions for upstream, midstream and downstream capabilities. We are proud to have worked very closely with Aramco, with Schlumberger to bring connectivity and edge compute to very high-risk environments in a vital industry, from autonomous drones, to well drilling operations, to monitoring the safety of workers. A tremendous amount of complexity in these industries.

And the results are showing. Our indirect revenue is up 77% from '24 to '26. We have tens of thousands of unique customers, and we are working to grow many thousands more. Over 200 hardware and tech solutions, more than 35 leading distributors, 45 global GSIs. Our partners span all verticals, and we address every step of the value chain. The channel has become the multiplier of what we are building.

To summarize, AI is and will continue to rearchitect the operational plane. And that disruption is going to create this upgrade cycle across billions really of endpoints, which is a massive market opportunity for us. And in the last 18 months, we have rebuilt our entire product portfolio, our developer platform and a vertical go-to-market. We have purpose-built silicon for software and AI. We have a clear prototype to commercialization path across 38,000 customers, and this is repeatable full-stack blueprint across industries.

This business has really helped us in understanding very deeply as we take on new challenges like robotics as to how to figure out how to diversify the company. And really, all of these build up on each other, all these learnings compound. The automotive and industrial business has created a lot of focus to on-ramp the teams to be able to go after robotics. We understand how to do safety. We understand how to get into vertical-specific markets.

And robotics requires 4 key building blocks: computing at the edge, connecting the edge, enabling high-performance AI hardware and orchestrating intelligence systems. And all of this builds on our collective experience, knowledge and investments from various automotive and industrial initiatives.

Now robotics is where embodied AI gets physical. So the objective is to perform human tasks, which would include mobility or motion, perception and reasoning and actuating or manipulating in the physical space that is around you. So these are systems that they have a sense, they have to think, they have to act. This is, in our mind, at least a $1 trillion opportunity over the next decade. And it requires a very broad set of technologies, products and experiences that no real general purpose chipmaker has today.

Before we get into what we are building, I want to maybe describe to you a little bit as to what will robotics do. And then really, you can think of this as a time continuum. So embodied AI implementations will encounter many tasks that will have varying degrees of complexity across the mobility, the actuation and the intelligence domains. Starting with inspection, where mobility is the underlying skill. Tasks are going to include reporting status, visually documenting, measuring and surveying the real world.

Next is transportation and the movement of goods, of tools of packages and even people. After that comes the interaction with the physical world, core skills to start and then finer, more precise, more dexterous skills, pick-and-place, sorting, assembly, insertions. Some can be single shot, some can be a longer horizon. This builds up to multi-agent fleets, teams of robots that are working in partnership, in coordination with each other across multiple use cases.

And then finally, robotics comes to the home with consumer interaction, which will require tremendously high levels of safety, of testing, of trials. And each of these steps adds more and more unique value. We have silicon shipping in every single tier today, and let me share with you what we are doing in this space.

So as we have done in pretty much every market that we get into, we always look at our full stack approach. We want to be able to make sure that we capture value across the stack. We want to be able to make sure that we can drive the pace of the inflection point if a market is ready to mature.

We are focused on 6 layers: the compute, the next-generation operating system, the simulation, the data pyramid, the data flywheel, the models and the hardware reference design. And as we have learned from our automotive experience, we capture disproportionate content value. And we've built tremendous systems expertise whenever we take a full stack approach.

A robot is not 1 computer, it's 3 computers. They work in concert, and they have an in tandem hierarchy. System 2 is the reasoning brain. It's the cerebrum. It's the heavy mixed critical AI workloads that require deliberative thinking. System 1 is the action layer. It's the layer that plans the motion. System 0 is executing the motion. It is your reflex system. It's the millisecond control. It's the highly decentralized part, what you call the nervous system.

And the main takeaway is this is a heterogeneous compute problem statement, something that we understand a few things about. And this requires you to have this optimal balance of distributed thinking, planning and real-time reflexes that require the appropriate compute engine. We have solved similar challenges in the history of the company, most recently with flex in automotive, which is, in our mind, another initiation of physical AI. And to my knowledge, we are the only company that is architecting across all 3 domains.

I'm going to get a little bit technical. The hierarchical compute architecture comprises of the thinking brain with central compute, of several limbs and joints with their own local compute and split second motor control and intelligent sensing for end effectors. We are building this entire system from brain to fingertip. The Dragonwing IQ10 is our center compute, SOC. It's purpose-built robotic silicon, and it is already commercial.

The perception IP allows us to visualize the world around us across multiple context modalities. The motion control IP for trajectory and balance, the actuation and control IP at the servo motor control level, wireless and wired IP for time-sensitive networking and obviously, all this on sensing. We've taken this chassis mindset that we adopt from our learnings in automotive, where we start off with a specific area that we are good at. And over time, we expand to get to a system level focus. This platform-level thinking allows us to think about the embodiment, which is a unique differentiator that Qualcomm has.

One maybe last complicated slide, second last. Let me orient you to this slide. If you look to the robot to the left, and I use this as an example that Cristiano and I were discussing this. Think about the concept of a robot picking a jug of water and pouring it into a cup. As it is doing that, the weight and the shape -- it is a paper cup, the weight in the shape of the cup is going to change. And that requires a robot's hand to sense and to adjust its grip pressure in real time. This is decentralized. This is exactly what you will feel as a human. That is the complexity of a robot.

When we talk about bringing a robot into the home, just think about what level of complexity you're talking about introducing technology to. And that is why this is a longer gestation cycle time period. 3 systems are active at the same time. The brain controller system 2 is highly perceptive. It is aware of the kinematic shape of the embodiment. It knows its degrees of freedom. It knows the range of motion. It's responsible for balance control. It can identify the cup. It can identify the water jug.

The body controller, system 1, it controls the limbs, it controls the movement of the hand to go to the jar, knows where the cup is and actually takes that motion on. It coordinates that movement. System 0 is actually able to sense the grip pressure, the tactile feedback, the temperature, the moisture, the weight, which allows it to act flexibly. Across these different systems, we have multiple real-time local loops that run within a system, and we have slower looks at running across systems.

And we are building embodiments across all these 3 different systems. We are also building a full software and application stack complete with ROS support and SDKs for manipulation to write to various types of central end-effectors. We will ship sample applications, pick-and-place, robotic arms, office scout applications, AMR for navigation, follow me applications. And this is open to every developer ecosystem, including the Arduino ecosystem that we just enabled.

The other aspect of robotics development is the simulation data and training model flywheel. We are building this environment in-house. We are building, as you can see on the left, a simulation platform where before a robot ever touches a real world, you have to be able to train it in the virtual world. It has to be aware of the physics, the sensors and the rendering as to how that will take place in the real world. This saves you tremendously in terms of the physical involvement of trial and error.

Then we have the data pyramid. That's the fuel for these systems. We combine real-world data that Qualcomm has access to, synthetic data that we generate ourselves and a lot of open source data. And then on the right, we train the foundation model. This is a single model that can take multimodal input like vision, like depth, like touch, natural language and it generalizes across use cases. We train these models with simulators, with behavioral cloning and teleoperations, and the reinforcement learning. So the workflow is end-to-end. We build the hardware, we generate the data, we develop the models, we deploy them into the customer environment.

We announced our IQ10 reference design at Computex in June, and this is purpose-built silicon, which is shipping today. We also have IQ9 and IQ8 for simple embodiments. And we're already designed into the NEURA robots, which you can see outside in the demo area, with whom we offer a complete reference design that powers the NEURA MAiRA cognitive robot arm as well as the 4$NE1 humanoid. These robots are trained in the new edge gen with a robotics foundation model, and they run the Neuraverse application platform.

This is a full stack, silicon, solutions with a key customer, a key partner in less than 6 months. Today, we are powering every type of embodiment, and several are shipping already. We have over 100 engagements spanning the entire robotic stack with companies like NEURA and Figure, KUKA. We are working with several drone OEMs, many AI sensor and embodiment partners.

With physical AI upon us, IQ10, the robotic reference design, the end-to-end solution stack ensures that customers and partners is with us. We are taking the same approach that has allowed us to scale very quickly in other businesses. Robotics is already a reality at Qualcomm, and we are very excited to be powering this next generation of physical AI, where we believe we are very well positioned.

To conclude, a few takeaways. Automotive, I hope you are all believers, is now a track record. We've had 23 consecutive year-over-year double-digit quarters of growth. We don't expect to let you down any time soon. $65 billion in design win pipeline, we have delivered, and we are still accelerating. And we are on track to becoming the largest automotive semi player globally. We are now a category leader in every domain we enter, and that's not easy to do.

Industrial and embedded IoT is now scaling. 18 months in, we have built a product portfolio with Dragonwing. We have built developer muscle with 3 acquisitions, 4 with Modular. And a full vertical stack that goes from silicon to solutions. And robotics is happening now. It's already shipping Dragonwing IQ10, IQ9 and IQ8 are all in production. Partners are integrating them into every embodiment, from humanoids to quadruped, from cognitive arms to AMRs and drones. 3 industries, 1 IP foundation, 1 physical AI platform.

Thank you very much. And before I turn it over to Cristiano, I would like to play a video from one of our partners, David from NEURA. Thank you.

Hello, everyone. My name is David Reger, and I'm Founder and CEO of NEURA Robotics. What we do is we are building robots and enable them to have cognitive abilities to see, hear, feel and think and react fully autonomously, all kind of physical tasks, like humans do.

The benefit of working with Qualcomm together, is giving a robot more than just a brain. What I mean by that is, today, we're seeing physical AI is mainly seen as the vision language action model but it's actually much more. It's a little bit similar to swimming. You can't learn swimming by just your brain and by vision. So it means you cannot just watch a video and think you are a swimmer. How to learn to swim is simply going into the water, trying it out for yourself and then actually training your memory effect of your muscles, then training also your reflexes and nervous system, how to breathe, how to move your body to actually stay above the water.

The task always require more than actually just vision. They need a feel of touch. They need to hear. And combine that all to build the foundational model, which can actually do all the physical tasks on this planet. And we did build the physical AI platform we call Neuraverse. This is a deployment platform where everyone in the world can actually train and contribute to make this 1 brain actually smarter.

What you can expect with all the partnership with NEURA and Qualcomm is basically setting a new standard in physical AI, enabling every robot on the planet and every human on the planet to actually train the robots and enable them for all kind of physical tasks on our platform, Neuraverse.

All right. So I got to the last part of the presentation, I think, before Akash will come in to walk you through the financials. And I think -- before I start what I'm going to tell you next, I think this is what's unique about Qualcomm. I know we have a limited amount of time, but there's a lot of new vectors of technology that -- and hopefully, you'll be able to see that.

It's not only about one solution in the data center. But also, when we think about we're doing automotive, we're thinking about in industrial, which is a whole different industry in this field of robotics. You need to have the breadth of semiconductors and technology that we have. And I think that's an opportunity with Qualcomm.

With that, I'm going to talk to you about the future of mobile edge devices. I'm going to try to unpack a couple of different trends that are going to happen in -- as we think about the role of agents. I said it in the keynote of Computex. The advent of agents and orchestrator was a very significant milestone that actually provide clarity how those devices are going to evolve.

And the industry tends to think in binary terms. Is there all of a sudden everything they can stop and you're going to go do the other thing. No, but they're going to coexist. But devices are going to have different type of use cases, and that's what I'm going to try to unpack in this presentation.

The first one I want to talk to you is, we all have been used with the mobile, the smartphone at the center of your digital life. And everything is around that smartphone. The OS, the app store, it becomes the control point of the OS. And the apps understand the human intentions and everything is around the smartphones, even other devices is just an extension of the smartphone.

That's not the case anymore. Actually, every AI company, every foundational model company now talk to us, the devices are the end points for agents. That's where the humans are. And the agent is at the center. It's not about the phone at the center anymore. The agent for the agentic experience, once you understand human intentions, the agent is at the center, devices are just endpoints of the agent.

And the purpose of my presentation right now is to tell you how the device is going to change because everything that has happened. Let's start about -- start talking about the user experience. So those devices have been built for the human as the user. So the workflow is based on the human going to an app and doing things at the human speed. But now the device with the orchestrator and the agents are also going to do other things on behalf of the human. So it's going to operate the device, and we're starting to see that right now.

If you ask me where is the epicenter of the start of new agentic use cases exactly happening in China right now? And you started to see the agents go to your device and operate the device for you, and it goes to the web for the agentic experience. That tells you that the device now has two use cases. It was interesting. I said this before, I'm going to repeat it. When people want a computer to run OpenClaw and they have the computer running OpenClaw. Once you start having that experience with you, not about just the amount of software developers they use -- they exist in the world, but the 6 billion people that have smartphones. When they started to use agentic experience as part of your experience and interaction with the device, you're not going to carry the computer. It's all going to happen in the same device. And the device is going to have 2 users because there's 2 different workflows. There's you and there's agents. So that's one big change.

The other big change is perception in sensing. And that's what is actually changing the device in itself and enabling different endpoints like personalized devices. So if we have now the computer that interact with us the way we interact with other than the context that we are inserted, especially as we think about we communicate with audio, we speak, we listen, we see. And then we are integrated into our surrounding, the context become important. So you now have a lot of sensing data that needs to be a part of the processing of those devices, and that is enabling also a different class of device.

That's why I told you about what happened to wearables and what happened to augmented reality, mixed reality and virtual reality transition is very important because the reason we've been very, very focused on glasses is because glass is as close to our sensors, close to our eyes, to our mouth to our ear. And those devices, wearables as extension of the smartphone when the smartphone in the center for the agentic experience is actually an end point of an agent. And the stuff that you actually wear becomes very interesting.

And especially when you think about glasses, just as a side comment on Microsoft Build, Satya announced Project Solara, a badge with a camera. It's a new class of devices. We have 40 different designs today with some of the largest AI and model companies in the world thinking about those new form factors. There's one form factor we know is going to get scale is glasses.

And the use case is see what I see and hear what I hear. And I'm going to come back to the use case because there's also going to talk about another change that's going to happen on devices at the Edge. But now what I would like you to do is new companies are looking at this big change. The devices now are endpoints, the barriers to entry of OSs and app store for new experience is no longer the same. You're going to see a lot of excitement. I want to start by showing a video from Amazon.

We're in the middle of a fundamental shift in computing right now. And what's most exciting about this next wave of AI is what it unlocks for people. For their creativity, their curiosity, the things only humans can do. Now as technology continues to fade into the background, the customer must stay at the center. That shift creates new requirements not only for the devices already in people's homes, but for entirely new devices built for AI-first experiences.

At Amazon, we're building for this future with Alexa, creating experience that works seamlessly for a customer when they're in the home as well as when they're on the go. Qualcomm is one of our critical partners in expanding both the capabilities and reach of these experiences, so AI can seamlessly move with people throughout their daily lives. This relationship is about building what comes next together.

So there's a lot of exciting things coming. And by the way, I'm actually so grateful. I have a relationship with Panos Panay, more than decades. He's an incredible individual visionary, and I'm actually very grateful for the partnership.

So the other thing is we talk about the orchestrator. And you heard about today, those agents, they generate demand for a lot of tokens. The reason a lot of the hyperscalers just see a wall in front of them of compute demand is because the economics of AI are fundamentally changing. Agents and orchestrators, that's why I said the OpenClaw was an incredible milestone. They're redefining the architecture and economics of AI, not only creating an entry point for Qualcomm into the data center, but actually, creating a fundamental change in the architecture of compute that touches all the devices on the edge.

And this order of magnitude increase. If you look at how we started with conversational to now agents, you see the order of magnitude increase. The projection is 40x the increase in annual token demand between 2026 and 2030.

Now I'm going to show an example for you next. And I'll tell you what you're going to see. We've been doing a number of those things. And you can do, you can try yourself. You can try different prompts. And you're going to see it changes. Sometimes you get 10%, 20%, 30%, 40%, 50%. I just pick a very simple example for you to see today. We've been showing this.

So what you're going to see right now is we've got 2 computers. Those computers have orchestrators. Those orchestrators, you give them a prompt, a complex prompt. You ask them to do some research, design a web page, put the results. One, we're going to be using is just Claude. We use Opus 4.6 100% of the cloud. You see the thing working. The other one use smart routing. You use some models are locally installed into the machine, and some others in the cloud.

And you get those things to work. And what you see at the very end that you see what hybrid AI really means because you will get to exactly the same outcome that you want. But now when you think about things like mix of experts, when you think about different kind of models, you can actually see how the architecture of AI is evolving. The reason you saw some foundational model companies saying, "I am going to give up to do video creation." It's very obvious because you can use your compute capacity to monetize tokens of higher value.

And that's what we're starting to see. Actually, when Microsoft also set a build that they now create an unmetered intelligence and people really understand what AI PC is, this is happening everywhere. And the beauty of the devices like phone has actually happened on the same device with the different users. And you can see lot of useless debates I have seen over the years about is this edge or this is cloud. It is actually the wrong conversation to say, I have something that need to do on the cloud. Can I do it on the edge and vice versa. That's the wrong approach.

Things are going to be done in the cloud is going to be done in the cloud. The growth of the cloud is incredible, but the edge now also become a computing that is going to generate tokens and I think how the industry is naturally going to evolve. And the message here to you is like what's happened in the data center, inference is actually becoming disaggregated and distributed everywhere. This is a new form of compute.

For example, when you look at the architecture, you're going to have the cloud data center, which has hundreds of megawatts. This diagram, I'll bring it back to you. You'll understand I'm going to overlay 2 diagrams that are going to make it very interesting. But you look at the data center with hundreds of megawatts, 2 gigawatts, you're going to have regional data center with tens of megawatts and on-prem. There's a lot. If you look at results of some of the server companies like Dell, there's a lot of movement for on-prem right now.

It's not against the cloud. They're both growing. And you also see, as you have more of a hybrid AI, you're going to see more computing happening on PCs and devices at the edge. So inference become distributed. Just don't take my word for it. I was -- we have an incredible partnership with Google, and I want for you to hear from Google about that.

Hello, everyone. I'm Rick Osterloh, Senior Vice President of Platforms and Devices at Google. And I'm incredibly excited to talk about our long-standing partnership with Qualcomm. As the industry shifts towards the agentic workflows. We're moving beyond simple responses to offer true digital agents. These are proactive multistep systems that seamlessly anticipate user needs across your entire ecosystem of devices.

To bring this to life, our teams are focused on bringing a shared full stack vision. We're combining Google's advanced Gemini models and Android system-level intelligence with Snapdragon silicon. Together, we're innovating side by side to scale on device AI and Gemini intelligence to the most advanced mobile devices. With AI, we're redefining next-generation automotive experiences and pushing into new frontiers with wearables like XR glasses and intelligent eyewear.

And this innovation is also at the heart of our brand-new Google Books effort to revolutionize the laptop experience. True agentic workloads require what we call distributed intelligence. By efficiently balancing processing between the cloud and on the device edge, we can deliver seamless experiences that are private, instantaneous and personalized.

Our deep engineering collaboration with Qualcomm ensures that the broader ecosystem and our OEM partners can scale these innovations rapidly. Gemini Intelligence will elevate the Android ecosystem, making it smarter, more proactive and more helpful than ever before.

Cristiano, congratulations on Investor Day. Thank you for our outstanding partnership, and I'm really excited about our road ahead.

So big thank you, Rick, for the partnership and the confidence what we're going to do together. So I have to pick up the pace now, and I get to my last part of this presentation, which is about 6G. So we have been designing 6G for this AI era. And I'm going to now show the role of 6G in this conversation we just have, and I'm going to go fast. So I just thought about new classes of devices like glasses. So the goal of the connectivity of 6G is to transform all of us in walking cameras in this world. I need a very fast high-definition video, the opposite of what we did with 5G, which is enable streaming a high-definition video. We're going to do that for uplink and across the cell site, so everyone has the ability to stream high-definition video, foresee what I see. That's very important context for Agentic experience. That's why we're going to the connectivity. I'll go to the details. There's a lot of improvement on the connectivity side. But that's only 1/3 of the story.

The other part of the story is the computing part. And the computing is going to be required because the 6G infrastructure is no longer a dedicated equipment for communications. The network of 6G, actually, not -- you need to be thinking about that was a network that was designed for voice. It now transported in bits and is going to also generate tokens. And the reason that's important because the premise of how 6G is going to deal with radio frequency is going to look at radio frequency is physical AI, and you're going to need compute. Look at the architecture, it's kind of the same of the distributed. You're going to have a big data center, you're going to have a regional data center that's the core of the network. You have to have an Edge data center and you're going to have the cell site and the devices at the Edge. And that's why we're also building our data center solution to be scalable because when you think about 6G becomes one sovereign AI workloads. And some operators will actually be selling token generation machine capacity like a CSP for AI companies for this distributed compute. And as you have all of this compute, it brings the next part of 6G, which is sensing. Because every single RF is going to be treated as a radar and you use model trained on that RF performance and the RF radio characteristics, you're going to be able to sense everything. Drone detection is on the top of mind. It becomes critical infrastructure, everything that moves and fly and becomes important context for models, and that's part of the perception.

So with that, I'm going to summarize it what's happening with devices at the Edge right now. They're evolving for agentic experiences. They're going to have more than one user. And we're building the next architecture of those devices. You need now a CPU for the orchestrators. The orchestrator is going to operate your device for you. I think everybody now understands the CPU matters and it's important. You need to have a different architecture on inference because you have to have very high performance, low-power inference even when you are not using the device as a human. And that -- even that high-bandwidth compute that we're doing on the data center, we're building a version of that as a coprocessor for phones. There's a completely change on the perception and sensor as well as a new modem, and it's just not about just the phone. It's different classes of devices. And that is also bringing other people to the space. So I'm going to finish this part of the presentation asking Hark, a new company, also founded by Brett, which is going to show you what they're doing.

I founded Hark because today's AI is just not good enough, and nobody is taking real advantage of it. They constantly forget things about you, and it runs on devices built a long time ago. So Hark, we're an AI lab building the world's most advanced personal intelligence. I think the next AI platform is something that truly knows you, that can see, hear and act in the world with you. And you don't get there by bolting AI onto existing devices. You do it all together. The models, the hardware and the interface is one product. That's what we're building at Hark, an intelligence that thinks like you and sometimes ahead of you. We're grateful to have excellent partners like Qualcomm, who are helping us realize some incredible ideas. We're excited to release the Hark platform this summer and then the next generation of consumer devices after.

The obvious question that everybody is asking, what is the timing? It's hard to predict the timing right now. And I think actually, the mobile market is dealing with the uncertainty of the memory situation. But it's very interesting. We are incredibly encouraged with the design activity, the number of new entrants. Actually, when I talk about China, I am in a situation right now, I don't know who the mobile customers are anymore because there are OEMs, but every single AI foundational model company building agents, also our customers. The surface area is tremendous.

There's 6 billion phones, 2 billion personal AI, 2 billion PCs and 500 million cars. And when you think about this new form factor, I would just give you an idea of how you should think about the glasses. We're just at the beginning of this, just at the beginning. 600 million glasses are shipped annually. There's less than 1% market penetration for smart glasses. But all you need to do, you can actually -- we're building a very small reference design that you can build the glasses. And you can see you can have ability to access an agent for audio, multimodal or premium display, and you can see the smart E-BOM. So that is a great opportunity, and that's what's also going to happen on the Mobile Devices business of Qualcomm. So I hope you saw there's a lot of interesting trends in technology.

And as I get to the end of my presentation, and I think you're all eager to see what Akash has in store for you. But I am just going to go talk about this #3 pillar that we've been talking in this presentation about the next chapter of Qualcomm from silicon to platform solutions. Into really building a fully integrated platform for hardware, software. But changing also the company to a developer-first company mindset. And you saw we're doing this with all the new business we built on the Edge, and we're going to be doing this from everything we have on the compute continuum. That's why this acquisition that we made of Modular is so significant for Qualcomm because it also builds on the pillars of the Qualcomm advantage, not only the technology, but the focus on deep customer partnerships and creation of ecosystem and the scale.

We have proved that we can partner across the industry. It's never the role of one company to innovate. Though that creates an incredible opportunity. Before I bring Chris up here, and he's going to talk to you for a few minutes. I am going to say -- and I may be a little bit aggressive saying this, but I'm going to say, I'm willing to bet that not everyone here will understand what we're trying to do. I think some of you will understand. And this is not a negative comment in any shape or form. It just takes a while to understand how we've been thinking about this. I've been -- we've been starting this journey with Modular for more than a year, and it took a lot, I think, for me to convince Chris and team, and he will share his story. But I think we may have an Android moment here. And maybe I'm going to be as bold to say, maybe there's even a Linux moment. I don't know.

But we -- I think we have something good. And I think we have momentum as AI goes everywhere, as compute becomes distributed, as you have every single endpoint become an endpoint for agents doing inference and you have an industry that wants an open ecosystem, maybe that's what Qualcomm can do, supporting everyone. With that, I'd like to bring to this stage a legend, Chris Lattner, I think the Founder and CEO of Modular, who will tell you about what we're going to be doing. Chris, please come on stage.

all right. Thank you, Cristiano. By way of introduction, I spent my early career building today's software platforms. This includes the compiler technology that runs every phone, no matter from whose vendor is in your pocket, the data centers that span all the hyperscalers. Built the Swift programming language at Apple, also built the software stack that powers Google's amazing TPU AI scale platform. Now I joined -- I decided to join Qualcomm because I found the team recognizes something. They recognize the opportunity of AI today. They have the ambition to do something big. But also they've made all of the investments already that put them in a perfect position to do something about it. Today feels familiar to me. It feels a lot like back at Apple when it was about to take off. Back then, there were a lot of doubters. People did not really understand what was going on, but we had all done the formative work already. And so all we had to do is get people to see it through the products in their lives.

So let me walk you through what I see today. So it turns out that compute has fundamentally changed. You've heard about that a lot today. It's no longer about a single chip. Compute today is a large-scale data center distributed systems problem. We all need to program diverse AI accelerators from multiple different vendors. We need to get the best performance, the best TCO, and we need usability because doing all this is harder than it's ever been before. Now the world is still struggling to get individual systems to compete with the industry leader, but that's where Modular comes in. And Modular, we spent the last 4.5 years building a novel platform that actually scales, all with the goal from the beginning of unifying the industry and opening a new chapter for accelerated compute. Now we built this platform to scale across the full spectrum, starting from the data center, but then going all the way down to the Edge.

And so this is why I'm so excited that Modular is joining Qualcomm. We're bringing together the perfect combination of scalable hardware and scalable software. This is joined with a shared ambition from both teams to make a world that is better for everyone with AI. And I got to tell you a little bit about where we're going. So now let us remember that nobody wants -- everybody wants tokens, but they also want amazing economics. Now most people don't really actually want to know how it works. The systems, the software components, all the different pieces go into this are amazing. And for nerds like me, I love it. It's great. I think I'm a fellow company here. But a lot of people just want a solution. And so together, we're lifting the Qualcomm AI silicon business. Silicon no longer. Now it's about solutions, full -- it just works solutions and the AI solution business is far more valuable. Now this end-to-end solution approach starts in the data center, of course. That's our core focus, but it won't end there. This is the first software platform that was designed from the beginning to unify Edge and data center, utilizing diverse accelerated compute in all the crazy form factors that are pervasive in our lives.

Now we've been using a lot of operating systems over the course of the last decades, but this platform will grow into a full operating system, built natively distributed, natively accelerated and agentically native by design. And we are not just building this for us. We're building an open developer platform, enabling AI developers, AI researchers and app developers to innovate like never before. Together as an industry, we understand that we need to scale incredible amounts of compute. We know that this will require many different organizations to come together to make it possible. The result of doing this all is an incredible opportunity for everyone, all of our partners, millions of developers and of course, all the consumers that will benefit from AI products in our lives.

Now throughout my career, I've had the privilege to drive open standards. I built several very large-scale open source communities. I've been part of multiple waves of compute. And I feel that today, Qualcomm is really quite well positioned to be the best in the industry to lift the entire world. And as such, we're committed to this being an open platform. Qualcomm is obviously an incredible hardware company, but today, we're going further. We're now a full-stack vertical AI solution company. I couldn't be more excited to build together. So thank you.

I hope you share our enthusiasm about this change in the company, this opportunity to change in the industry, what we're going to do together. And we're super happy to have Chris, Tim and the rest of Modular as part of the Qualcomm family. So I was about to end here, but I thought maybe there's one more thing. Hopefully, you are entertained, but we have one more thing, and that's just building on this. We're also very happy to announce a very strategic partnership that we're making with Hugging Face. And I'll tell you about this partnership.

Qualcomm and Hugging Face started a very unique collaboration because they share on exactly the same vision we presented to you today. For the data center Dragonfly, Hugging Face will be very focused on demand creation from Qualcomm Dragonfly silicon. Both their inference and storage services will map to all the Qualcomm Dragonfly products. There will be agentic model onboarding, combining the Hugging Face, their 16 million developers and what we're doing with Modular across the whole family of Qualcomm chipsets from Snapdragon to Dragon Wing to Dragonfly. All models are going to be onboarded on Qualcomm technology platforms using an agent that is going to handle the setup, the optimization deployment with 0 manual integration work. And then just what you heard from Chris, build on an end-to-end agent AI with distributed intelligence and that distributed AI framework when agents can operate seamless across the entire compute continuum, leveraging Qualcomm technology model and also cooperating with everyone. And I want to hear from Clément.

Hi, everyone. I'm Clém, Co-Founder and CEO for Hugging Face. If you haven't noticed recently, something big is happening in AI. More and more, the world is running on open source and local models and for good reasons. They're way more affordable than the big LLM APIs, more customizable by companies. And because they run on your own device, they're private by design. Your data stays yours. Today, over 16 million AI builders create this future in the open on Hugging Face. And that's why I couldn't be more excited to announce a new collaboration with Qualcomm.

Together, we're going to make open models easy to run everywhere from a device in your hand to a full rack in the data center. Snapdragon, Dragon Wing and Qualcomm's Dragonfly Cloud, all powered by the open source community. You'll be able to take any model, big or small, deploy it, optimize on any Qualcomm platform with agents running on device and orchestrating across the cloud. We'll also offer Hugging Face Pro subscriptions to many developers using Qualcomm platforms. Local, private, affordable for everyone. That's the future of AI we want, and we can't wait to build it together. Thank you very much.

So that's it. I think I got to the end of the presentation. Hopefully, we gave you an opportunity to understand, I think, the -- what Qualcomm is going to do within the next 5 years. I'm going to summarize it to you. Data center will add a meaningful new vector of growth. And I think what you saw is when we originally talked about this, we talked about we're building a data center portfolio. We expect revenue to be in fiscal '28, then we get more traction, we move it to fiscal '27. We get more traction, we get to fiscal '26. And you heard from Tony, we're just starting. Automotive, Industrial and Robotics will extend Qualcomm in the next frontier of physical AI. We build the platform. We have the market scale, and we're executing on all the technology trends.

Agentic AI at scale will drive an upgrade cycle on across Edge devices. Those are going to be machines going to generate tokens, and they are going to be interacting not only with the users, they're going to be interacting with agents, and that's going to happen across the entire industry. The token economics will make distributed inference inevitable. We're excited about the growth in the cloud. That's what a great opportunity for us to enter in to disaggregate, and that will continue. We're just at the beginning of that. But everything will become an AI computer. And I think that is going to fundamentally change and create a massive opportunity for us across the compute continuum.

6G will be foundational infrastructure for the age of AI. And if you haven't forgot, we have that asset, too. And we're going to be expanding beyond silicon to full stack software platform with the most industry-friendly, I think, platform. And at the end of the day, I think we have seen in our industry, open horizontal systems will win. And I think that's kind of our bet. So with that, thank you so much for listening to our presentation. And now I think the main attraction of today, our CFO, Akash.

All right. Good afternoon, New York. It's incredible to be here. Lots of familiar faces. Great to see all of our investor friends here in the room as well. It looks like all of you decided to stay here rather than go to an earnings call. That's a great decision. That's a great decision. We're going to make it worth your while. This is the climax of the show. So we have closed the doors now. You're stuck here. You'll have to listen to the rest of what I have to say. Just kidding aside, we -- you heard Cristiano, Tony, Nakul talk through all the great stuff we are doing across our businesses. And so my job is now to try to wrap it up in a financial framework, and so let's just get to it.

Through my presentation today, I'll try to address these key areas. Revenue and EPS are going to grow much faster than what we had told you before. We're going to see with diversification and growing into data center, the mix of businesses will change radically versus our previous estimate. Our operating scale is going to be a key differentiator for us going forward, and you heard a little bit about that in the various presentations. And then capital return, capital allocation remains consistent with what we've told you before.

But before I go through all of this, let me just quickly address how Qualcomm has changed over the years. We obviously started off as -- by inventing 3G. We led 4G. We led 5G. But today, we have changed. We are more of a computing company than we are a connectivity company. We're still the best in world in connectivity, mind you. But we are a computing company today. So for a lot of investors who've known Qualcomm for a long time ago, it's important to make the switch that we are a computing leader that also happens to be best-in-class in connectivity. The second change in Qualcomm -- The second transformation started when Cristiano became CEO 5 years ago. We went from being a smartphone company to all Edge devices. Auto, Personal AI, Networking, Industrial, PC, all of these devices, we are a leader in now.

The third transformation starts now. We're going to go from being a devices company to being a cloud and device company. And as we change, as we transform, it completely redefines what's in front of the company and how you should think about the company going forward. Let's start with some financial look back. This is the last 5 years, what has happened to Qualcomm overall. We've doubled revenue during this period, $44 billion. During this period as well, we've tripled EPS. And this performance and track record validates our growth strategy. It also sets the platform for what we're going to do going forward.

Within this period, what did QCT do? QCT far exceeded the performance of overall Qualcomm, much more than 2x growth in revenue. We also had double-digit CAGRs in each of our revenue streams. I'll highlight Auto, 44% CAGR over this period. Within Handsets -- within Android Handsets, we grew at a CAGR of 12%. This is a market that's perceived as mature. But during this period, we grew because content increased and the mix across tiers got stronger. We expect this trend to continue. If you think about what happened to earnings during the same period, we grew 4x, 2x faster than revenue. And this is because of operating leverage and really investing and growing and diversifying across the businesses.

This slide outlines the businesses we are in today. We're in Licensing, Android Handsets, Automotive, IoT and now in Data Center. Across these businesses, we address $1.7 trillion of TAM. And as we diversify more, as we launch new products, a very large portion of this becomes addressable to us. Over the next few slides, I will focus on three key areas in terms of financials. I'll talk about data center, I'll talk about Automotive, and I'll talk about IoT.

But let's first start with the numbers. I'm sure you're waiting for the revenue forecast. Last time when we were here 18 months ago, we set a target of $22 billion in non-Handset revenue. This is Auto and IoT, $8 billion in Auto, $14 billion in IoT. And I remember a lot of investors said, "Wow, that's a very aggressive target. Are you going to be able to meet the target?" 18 months later, we are here again, and we are very happy to say that we are revising the target. Our fiscal '29 revenue target is now $40 billion. Just to repeat. Let me just repeat this. Same year, last time, 18 months ago, we said $22 billion. Now we're saying $40 billion. This is nearly 2x increase in the target for revenues in fiscal '29. What this also means is 4-year CAGR from '25 to '29 is 40%, Very strong growth as a result of the diversification effort. We have an incredible opportunity in front of us.

Okay. So now I'm going to start with Data Center and talk through the data center financial forecast. As Tony outlined, incredibly exciting product portfolio. It's built on the basis of technology leadership, and this is by far our largest growth opportunity. In terms of revenue ramp timing, we have revenue today in fiscal '26 from the connectivity products we acquired through Alphawave. As we get to fiscal '27, that's only 3 months away, we will start ramping custom silicon revenue in first quarter of fiscal '27. In second half of fiscal '27, we'll ramp AI accelerator revenue. And then second half of fiscal '28, we will ramp CPU revenue. All these revenue streams will layer on top of each other. It's similar to what we did in Auto. We started with connectivity cockpit, ADAS. Same thing in data center, and that's how our revenue will scale. Across all of these markets, there's a $1 trillion TAM opportunity for us. So we're incredibly excited. As we launch these new products, we'll be able to access a very large portion of that TAM.

So this is our financial forecast for the Data Center business. Let's start with fiscal '27. We're targeting $5 billion of revenue in fiscal '27. And let me highlight a few key things. We will have 2 hyperscaler customers that are at global scale that will drive at least $1 billion of revenue within the year. This is not a concentrated revenue stream. We have diversification within the customer base. We also expect custom silicon gross margin to be slightly below our overall Qualcomm gross margin, but it will be accretive at the operating margin level. So this is a very attractive financial business for us. In terms of AI accelerator and CPU, we'll be investing ahead of revenue ramp in fiscal '27.

Let's talk about fiscal '29 now. We are targeting $15 billion of revenue in Data Center in fiscal '29. And this revenue -- what gives us the confidence in being able to achieve this revenue is the diversity of products, diversity of customers and the fact that we are talking about multi-generations across our customer base. So this is the forecast that we're putting out for fiscal '29. But let's talk about the opportunity beyond that. As we discussed earlier, it's a $1 trillion market cap -- $1 trillion TAM opportunity for us, and we have an incredible product portfolio. Across all this, we are targeting greater than 5% share in 5 to 7 years. So this is not a question of what will we do in fiscal '27, which is very strong. It's not just a question of what we'll do fiscal '29, but the long-term opportunity for us is incredible.

So I'll go to Automotive now. Snapdragon has become the platform of choice for the Automotive industry. We are very proud of what the Auto team has built. And as Nakul discussed, that becomes the platform with which we jump into Industrial and then into Robotics. Content increase has been a key part of the story. And I often get the question on, "hey, Auto market is mature. When will you stop growing? You're growing much faster than everyone else in the industry." The reality is that the part of the industry that we are in, the amount of content growth is tremendous. Between our fourth, third-generation product, fourth-generation product and fifth-generation product, there's an 8x increase in content. This translates into financial growth and really a very long runway for us. We just launched our fifth generation product.

As you think about the overall TAM for cars in that flat market, our SAM is going to grow -- easily grow double-digits, and we are going to grow much faster than that. The drivers of those growth includes digital cockpit capabilities, increased sensors, ADAS going to L2++ and then finally, generative AI capabilities, and we are leading in each one of these areas.

Let's go to the Automotive design-win pipeline. As Nakul mentioned, $65 billion. 2 years ago, we were here, we're talking about $45 billion design win pipeline. Now we are at $65 billion. As a reminder, what design win pipeline conveys is really the cumulative revenue expected over the designs that we have won. But what's incredibly impressive about the design-win pipeline is the diversification. Diversification of products, you have cockpit, connectivity and ADAS. We have diversification in customers, and we have diversification within the regions. And this is a very important attribute of our pipeline. We are really winning globally across all OEMs. We're looking forward to continue to grow this pipeline going forward. Okay.

So now I'll talk about the revenue forecast. This is the last forecast we gave 18 months ago. And we said we'll be at $8 billion in '29, and we'll be greater than $9 billion in '31. And at that point, we had pulled in our revenue ramp by 2 years. So what we're going to do now is we're going to pull it in again by 2 more years. We will hit $10 billion of revenue in fiscal '29. And as Nakul mentioned, we'll be the largest Automotive silicon supplier shortly. And we are not done yet. If you think about the growth vectors that remain, we're incredibly excited about them. Robotaxis, autonomy going to L4, token accelerators, a separate accelerator that we'll use HBM for and then AI HBC for and AI workflows. So tremendous amount of vectors still remaining. We'll continue to grow this business for a very long period of time.

I'll go to IoT next. Cristiano outlined this. We're thinking about IoT in 2 buckets going forward. You have Personal AI and Compute and Industrial, Networking and Robotics. For personal AI and Compute, agentic AI is driving an inflection point in these devices. We used to talk about glasses. We don't talk about glasses anymore. It's not just glasses. It's a bunch of different devices that come with it. For PCs, it's -- of course, we are in Windows PCs, but now we're extending to Chromebooks in addition to tablets. For Industrial, Networking and Robotics, digital transformation is creating an incredible opportunity for us. So I'll talk through both of these areas in some more detail, but here's the financial forecast. We're going to be over $14 billion across these markets, which is a CAGR of 20% from where we were in 2025. We have strong growth drivers and pillars in each one of these areas.

Let's start with Personal AI. Personal AI is a set of devices that can see what you can see, hear what you can hear and you can have an Agentic AI conversation with it. We have the broadest technology portfolio that allows us to win here. If you think about things that are required, small form factor, low power, connectivity, great camera, sensor fusion, object tracking. We have all these technologies in-house. We integrate it into our products, and it has been adopted by all kinds of customers. You have the traditional OEMs, you have the hyperscalers and you have new entrants and all of them are using our chips. We have included a modest forecast in our financials for this market. If the vision that Cristiano outlined plays out and this market turns out to be much bigger, we have tremendous financial upside opportunity that is not captured into our forecast.

Let's talk about PC next. Actually, this week, 2 years ago was when we got into the PC market. It's been a very short period of time. And here's our report card since then. We've launched products across all tiers, flagship to entry. We are the performance leader in every single tier. We have multiple generations of products now, so we can launch current generation products and previous generation products at the same time. But when we got into the market, one of the questions was, will you be able to build the channel? Here are the metrics on the channel. We now have not just design-wins at OEMs, but very broad acceptance of applications that have been ported over to our platform, printers, peripherals, consumer channels with retailers and enterprise channel. We've built a very, very strong platform over the last 2 years, and we are ready to scale the volume as a result of it.

Agentic AI is also driving an inflection point, and it changes the way the device is going to be used, and we are very well positioned to take advantage of that. Rick mentioned about Google Book. We are the lead partner for Google Book. We'll be launching devices with various OEMs over the next several months. And these are devices that will bring Snapdragon along with Gemini together to deliver the best Agentic AI experiences. So while there is more work to do for us in PC, we are tremendously proud of the progress that we have made, and we are set for takeoff.

Finally, Industrial, Networking and Robotics. This is another market, as Nakul outlined, AI is accelerating the transformation. 18 months ago, when we were here, we were talking about how microcontrollers will move to microprocessors and AI. Today, I know a lot of our peers come in and talk about that same transition. This positions us extremely well for what's going to happen going forward. Like PC, one of the criticisms that we heard is, "are you going to have the channel that is required to drive the volume here?" So we've built that channel. Now we have hardware and technology partners, we have distributors, we have system integrators and 38,000 unique customers, and we are ready to ramp the volume based on the products we have. We are targeting $8 billion in '29 for this business as well.

Okay. So let me summarize what we just said on the financials and add a few more metrics to it. Not $22 billion anymore, $40 billion in revenue outside Handsets with $15 billion for data center. We are forecasting Android Handset revenues to grow modestly at 5% going forward. This assumes that the current memory environment doesn't materially change, and it also does not assume any uplift from the Agentic AI conversation we just had. Both of those things would be upside to our Handset forecast. Licensing, we continue to be very happy with the way that business is stable, and it will continue to scale with 4G and 5G units globally. And then finally, operating margins. No change to the targets we've said before. We expect QCT as we include data center into it, to be at 30% in the long term, and we expect QTL to continue to be at 70%.

So what does this do to the mix of businesses across QCT? When we get to '27, Handsets will be less than half of our revenue. This is a conversation we have with investors all the time, but you're a Handset company. Here's how the mix is going to change. We'll be less than 50% in '27. In '29, with the forecast we just showed you, Handset will be 1/3 of our revenues. We will be truly diversified across Handsets, Data Center and industrial, IoT and Automotive. Our capital allocation strategy really is unchanged. We're continuing to proceed down the path that we have discussed with you in the past. Our priority is investing in the business, maintaining the leadership on technology and continuing to accelerate diversification.

For M&A, we've done 35 acquisitions over the last 5 years, and each one of them was contributing to our growth strategy. We've picked the strategy and each acquisition comes in and says, "how do you help me accelerate something that I'm planning to do anyways?" And there's Alphawave, Modular are great examples of what we've done. In terms of capital returns, over the last 5 years, we've returned $40 billion to the shareholders. Over the last 10 years, we've retired 30% of our shares. As we go forward, the strategy remains unchanged. We're going to keep increasing dividends in mid- to low single-digits, and we'll return most of our free cash flow to shareholders. Also through this, we'll maintain a strong balance sheet. We'll retain financial and strategic flexibility, which is so important in our industry.

Let's talk about OpEx for a second. What has happened over the last 5 years is as we have diversified, we have grown revenue at 15%, but we have grown OpEx only at 6%. This is in spite of funding all the diversification efforts. And what that did is OpEx as a percent of revenue came down from 31% to 23%. As we go forward, we expect that to decline further to 19% to 20%. So very happy with the way we are managing the need to invest and focus on diversification while realizing operating leverage financially.

So we've talked about '27, we've talked about '29. So the question is, are we well positioned to continue to grow beyond that? So before I wrap up, I want to talk about all these growth drivers that we have beyond '29. Data center, clearly a massive growth vector for us as we look forward. Robotics, as that becomes one of the largest markets in the long term, we are positioned to win there as well. Industrial, we'll see an upgrade cycle that will happen over a very long period of time, positioned to win there as well. ADAS, both from ADAS and autonomy perspective, we have an opportunity in silicon and stack, and we're executing on all those opportunities as well. Personal AI, as Cristiano outlined, tremendous opportunity. If this becomes a scenario where everyone in the world has a couple of devices that are personal AI devices in addition to phone and PC, we're going to have a very significant growth vector beyond '29. And then finally, 6G, as we outlined before. So very excited. Our growth curve is not done in '29. We have a long-term growth opportunity, and we are focused on executing to it.

Finally, I want to wrap this up with some key takeaways. We have a clear line of sight to diversified revenue base where Handset will be 1/3 of our revenues in '29. We expect non-Handsets to be $40 billion within QCT by '29. Our EPS target is of greater than $18 in that same time frame. And then finally, when you think about long term, we have an opportunity with the things we outlined to scale our revenue to $100 billion.

Thank you very much for coming. Thank you for listening to us. We're very excited about what's in front of us. Thank you very much. And I'd like to invite Cristiano, Nakul and Tony back on stage for Q&A.

Thank you so much for staying with us. I'm glad that most of you stay.

Before we start, Akash, given that financial performance, can I buy some more shares?

Go for it. Ask [ Ann ] before you do. All right.

What's going to go first?

Chris Rollins, Susquehanna. I think data center is probably the most interesting here, the $15 billion and then it sounded like more than $50 billion over time. If you could talk about the linearity of this given your product releases and then also customer deployments, I think that would be great. And your overall ability, you think, to address this with your partners as well, given supply constraints, et cetera?

I think the best way to answer the linearity question is you have a number for fiscal '27 of $5 billion, fiscal '29 of $15 billion, right? And as we said, our product launches, the way it pans out is '29 will have the benefit of CPU accelerator and custom silicon and connectivity. So all 4 product launches. And then '28 will CPU comes in at the end of the year. So you're going to see this ramp that happens between $5 billion to $10 billion, but it aligns with the product launches time line across the period.

Maybe just add a few things to try to answer your question on the supply side. So as we outlined, we have visibility right now of $5 billion in fiscal '27. For that revenue, we have secure capacity as well as secure memory. So even our customer commitments right now on the high-bandwidth compute, HBC technology, we have secure memory as well. We're not a small company. I think we have capacity allocation, we consume a lot of leading node wafers. I also think our suppliers are betting on Qualcomm and want Qualcomm to succeed. And I think that is reflected in the capacity commitments we have for the projected revenue we made of fiscal '27 of $5 billion.

The one thing I'll just add is '29, as Akash said, is when all 4 product lines truly launch, and that's just the beginning of the launch. So '30, '31 is when we start delivering multiple generations of these products, and that trajectory is going to change from what you'll see over the next 3 years.

Who's driving the microphone? I think there's a couple of...

[indiscernible] Arete Research. I was wondering if you can zoom into the CPU commentary that you laid out for us. I mean some of the statistics, performance statistics versus your peers were quite compelling, the 5 gigahertz. And I was just wondering if you can maybe just hold my hand a bit and just tell us how you get to this, I mean, versus your competition, given how significant the TAM expansion has been over the last couple of months.

Look, that's a great question. As I mentioned during my talk, this is a company founded by engineers. Everything we do is about technical innovation. The Orion CPU core is transformational, as you've mentioned, 5 gigahertz is remarkable. and it's not even a custom hand-built design. It's built using automated tools, and it's updated and refreshed each and every year.

So look, it's foundational in architecture, okay? You cannot just stitch this type of performance in. It has to be built from the ground up. So even though it's based on mobile compute, the server class compute has been built from the ground up to lead in terms of performance. So I've been asked, "why are you launching in '28?" Because even in '28, we will have the industry's best performance in compute and I/O capability. And then when you think of bolting on the HBC attach to integrate native AI inference workloads on top of this industry-leading compute, folks, this is game over across the industry.

Maybe I was just going to add a few things. So I'll kind of remind you a little bit of our journey, right? So we have been building, I think, our own CPUs. I think the first thing we did, we build a CPU to have an Apple compete, I think as we enter the PC space to create an Apple compete. This is following the Apple M series. Then we built a CPU for mobile devices. We build a CPU safety grade for Automotive. So this is the next-generation CPU that we design.

One commentary, I think, to what you said. Right now, the demand for CPU is massive. So everybody has CPU chips. I've seen this in the pandemic. But if you look of what happened in the other markets with design CPU, our metrics have been very, very good from a performance, from a power. The feedback that we got on our CPU that we -- Tony outlined, I want to tell a little bit, we receive very specific requirements from all the hyperscalers about the CPU that they want to see in '28, want to start shipping it. And the feedback is this is good to be true, when can I get the silicon.

So you should be thinking about us having the capability, understand where the puck is going for Agentic and building a CPU for that. Right now, everybody is shipping. It almost like it doesn't matter. The demand is high, but we expect by that time frame, if you have like true competition, Qualcomm is going to fare very, very well in this area.

Josh Buchalter from TD Cowen. I was hoping you can maybe speak to your software maturity as we think about you merging into the data center ecosystem. I think we appreciate Qualcomm's rich heritage in silicon design and manufacturing, but it's a new venture for you guys and one where others have been inhibited by their software platforms. So could you speak to that and sort of what Modular brings specifically as we think about merging that into your road map?

Very good. Thanks, Josh, for the question. So let me break this conversation into pieces. I think what we're doing right now and what we're going to do soon, right? So first, we're being very focused on inference. And that is the focus right now, the disaggregated inference. It doesn't mean that's the only thing we're going to be doing, but we've been very focused on inference for the disaggregated, I think, accelerator into the Data Center.

We also bring another interesting capability, especially what you hear when you think about open-source models because when models have to run on the Edge, they have to run on Qualcomm. We have been embracing. I think a lot of the industry standards we've been supporting [ executor ] been supporting Triton, for example. And we have been working -- over the years, there was the purpose. I told you that we have been preparing for this, building assets. There was a purpose on AI 100. The purpose of AI 100 is basically to start understanding how we need to mature a software stack to the point now, we have new models in 24 hours actually is running on our accelerator.

So that's what we have been building. We've been building the capabilities supporting all of the different open ecosystems. But now there's something else we're going to be doing. Because the reality is some of those platforms right now are old. I think the incumbent platform has been designed about 20 years ago. And I think a company like Modular has a very modern platform, which has been probably designed for the disaggregated heterogeneous compute and is open. And I think that's how we're going to change the conversation a bit, not only creating something that delivers higher performance and is easier for developers to use, not only for us, but for the rest of the industry and actually make sure that happens on the data center as well on the Edge. So those are the 2 vectors. We'll continue to check the box and do what everybody is doing, I think, for inference and -- but we want to do something much better.

[ Luke Masoshi ] here with Daiwa Capital. So to continue on the Modular situation, it definitely seems very interesting. Maybe you could just talk about the obstacles. Where do you think it's going to be deployed, hyperscalers, Neo clouds, enterprise? Because obviously, once software does get defined, obviously, NVIDIA has made a lot of progress in this area. It's hard to overcome as you've seen with Windows and other things, but there has been success stories like VMware throughout the year. So it seems like you could have an opportunity. So more details would be great.

Well, so I'll definitely -- I'll tell what we see right now, and I'll talk a little bit about the vision. But first, I'll answer the question Is If for inference, NVIDIA was the only option, NVIDIA will be the only thing shipping right now, which is actually not the case. And I also believe that as you think about what the industry really wants and what a lot of the industry has been developing, you see the progress of Google with TPUs as an example. You saw the acquisition that NVIDIA made of Grok. So you have kind of different architectures. And I think the moat of inference is actually not as strong as it has been for training. But it also created an opportunity because you now have clusters of compute with a bunch of different hardware and you wanted a solution that exactly could actually abstract that problem for developers.

Now I'm going to tell you what I see. What I see is the Modular team, I think Chris and team, which are here, and we can't wait to get them as part of Qualcomm. They develop something that is modern and actually abstract this for developers and get a lot of performance of the hardware. And just don't take my word for it, if you actually look what they have done, they have achieved a lot of performance working with NVIDIA hardware, AMD hardware, with CPUs. And it's truly an open platform that can actually run across the entire different types of environments and compute and also scale for the Edge. That's how we start working with them.

So you're always going to have this conversation, which is somebody is going to say, I'm just going to stay with CUDA, which is tied to NVIDIA hardware. I'm just going to stay there. And I'm just going to get I am just going to get one ecosystem or I'm actually going to see the benefits of having heterogeneous compute, a disaggregated compute and see what's going to happen on the Edge, which will happen regardless. Like I'm telling you, I'm starting to see -- you see -- I am getting some of the major AI companies meeting with us. You saw some of those videos today saying, I need to move all of those workloads to the Edge. I have a better use for some of my tokens in the cloud. That's going to happen, and that's going to bring different type of hardware.

With that, I'm going to tell you what the vision is. At the end of the day, we have a lot of customers that make things and those customers that make things, they are going to start adopting a lot of AI in inference compute. You saw that across Physical AI. So I think the customer reaction is they're actually looking forward for a platform that scales, scales across the Edge, it's open, it's easy to use. And we expect that there's a lot of different customers in the cloud. They are dealing with the fact now they have 3 or 4 or 5 different software stacks they have to play with it. That's the bet. And as I said, we're going to be actively driving it. We're going to continue -- what you heard from Chris. It's going to be open. It's going to be available to everyone, and we're going to see what happens.

It's Chris Caso from Wolfe Research. For the fiscal '27 data center guidance, I think it would be helpful if you could clarify exactly what's in that guidance? And I guess from the product launches you've discussed, it sounds like it's the accelerator plus maybe some of the connectivity you got from Alphawave. And then with regard to the customers, you've already announced Humane as an accelerator customer. Microsoft was today, and you talked about 2 customers. So are those the 2 that are included in that?

So from a product perspective, I'd say the largest part of the revenue base will be custom silicon. As I mentioned, there are 2 customers who are global hyperscalers who will each be greater than $1 billion. So by far, that will be the largest part of the revenue. There will be a portion of AI accelerator coming in towards the end of the year. And then connectivity products that came from Alphawave acquisition will be also a portion of it. So it's really those 3 product lines with AI accelerator really coming at the end of the year. From a customer perspective, as I said, for custom silicon, the 2 large customers will drive the custom silicon revenue. And then we have a very large customer base in connectivity that came through the acquisition, and we'll have Humane as a significant portion of it as well. So that's the base.

Jim Schneider from Goldman Sachs. Maybe just to follow up on the prior question. Can you maybe talk a little bit about how you expect the customer diversity change from fiscal '27 onward? You expect -- I think you talked about 3 customers. You sort of indicated maintaining like the largest kind of lions share of the custom silicon. Do you have orders for all of that $5 billion already to be covered with that revenue? And then maybe talk about how much more diverse do you expect that revenue base to get in '29? And can you actually hit the '29 targets based on the customers you have now?

So let me address it in 2 parts. I think first is we have high confidence in our forecast, and so I'll leave it at that. The way you should think about the '29 forecast is we're engaged across a variety of customers today. We are engaged across a variety of products with them, and then we are talking about multi-generation. So it's a combination of those factors that gives us confidence in the fiscal '29 number.

And the one thing I will add is, remember, in data center, the discussions are moving from megawatts to gigawatts. And when you deploy full infrastructure, as I outlined today, a few gigawatts can get you to the '29 numbers.

[indiscernible] Constellation Research. I would be amiss not asking the Brazilian against who Brazil will be in the world cup...

I don't know. This is -- it's a tough one. I think I'm kind of encouraged that Brazil starting playing bad because when they start playing bad, usually, it brings down -- I think it brings more humility in the team. They start playing together and they start to improve in the second half. So I'm going to take that as probably a consolation for what the pro forma has seen in the beginning of the season.

Perfect. To the serious question, AI tells me you're building between 250 or 500 chipsets. I know you don't make that number complex, but I estimate with all your plans for 2029, that number might easily double -- how do you plan to handle the complexity because every chipset is an adventure, not all adventures and happy from a human skill, capacity, risking, supply chain perspective?

Okay. Look, I'm going to give the -- I think the answer that we always give to ourselves, right? So we're actually called quality communications. Maybe I need to -- it was interesting. I think during their 40-year anniversary of Qualcomm, we had an event and our founder, Dr. Irwin Jacobs, he came and to speak and he said to me, Christiano, I make a mistake -- I think I made a mistake. I should have done Qualcomm with 'm' because then it could be communications or compute interchangeably.

But I think the story here is we actually have a quality reputation. You saw -- when you talk about our customers, we get award all the time from mobile customers about the lowest defect density. We ramp brand-new chip and IP in a very fast period of time. We get over and over messages from Apple has been probably one of the best quality suppliers. And you saw what happened in Automotive. So I think we look at that skill that we have developed as something that is very important. What we saw -- and this is a little bit complicated. This is a little bit complicated.

I'm going to try to simplify the complexity. There's often discussion about leading node design and process technology. There's often the discussion about who is best in yield and all of this. And one of the things we learned, for example, when we do Snapdragon for mobile phones that has to ramp very, very fast. You have to design the product to a very narrow spec. I could never afford to do, for example, what Intel us and say, I have this distribution, and I'm going to sell this as a high clock speed as the i9, this is going to be the i7, and I'm just going to bin it. I have to develop the same exact part because you never hear Samsung saying, this is a fast Galaxy or not. And we saw what happened with this incredible demand on the Data Center. We hear anecdotal data from a lot of the customers about parts are going to have rework and failure rate. We actually look as a vector of differentiation for Qualcomm, which is our ability to reliability.

And we're not small. We ship 40 billion components, I think, every year. As I said, we do a large number of tape-outs of leading node chipsets, do all of them in parallel. You saw record dates on very complex Automotive industry, which is very strict from all the gates we have to go to. We're breaking new records from tape-out to cars. So we're going to bring all of that to the Data Center. And this is already happening. I wanted to use this opportunity -- sorry for the long answer, but I think sometimes the word is actually a lot more simpler than it looked like. Alphawave had customers. They have been licensing IP and they have engaged them with customers.

What we saw as soon as we closed Alphawave, that conversation accelerated because just Qualcomm add more muscle. We just had more capacity. We have a bigger supply chain, and we're willing to actually make commitments that people will bet a large volume on it. And I think that's kind of what happened. That's why they accelerated a lot of the custom ASIC engagements they had for Alphawave IP on it. And I think that's what I expect to happen. I think size really matters. As I said, all of the $5 billion we outlined and we forecasted right now, we have wafers and we have memory committed to those.

We have time for one final question.

It's Joe Cardoso from JPMorgan. Maybe more of a question for Tony on the connectivity side. Nice to see the road map here across copper and optical solutions. However, curious in one of your earlier slides, you also mentioned CPO. How are you thinking about that opportunity on the connectivity side? How is Qualcomm looking to participate and maybe time line around the road map there?

Thanks for that question. So at Alphawave, we had started working on silicon photonics and co-packaged optics about 5 years ago. And so what the plan is right now is to initially deploy first generation of silicon photonics in our AI 300 series, all right? So that will immediately empower optical scale out, drive down power consumption dramatically because you're no longer going from copper to optics with you're going straight to photons. And look, beyond that, AI fabrics will be moving optical. So we start with scale-out in and around 2028 and then scale-up beyond that.

That's it. All right.

I think that...

Thank you so much. Thank you. Thank you for being here with us. Really appreciate it.

Thank you.

Good afternoon. Thank you all for coming. I'm Stacy Rasgon. I'm Bernstein's senior research analyst for U.S. semiconductors and semiconductor capital equipment. And it is my honor to welcome our guest here today, Cristiano Amon, the President and CEO of Qualcomm. Now I always say Cristiano has been here for many years, and I always say here, Qualcomm is a company whose stock has been through a lot over the last 5 or 10 years. But I think now sitting where we are today, it really looks to be coming to its own, especially recently. Like handsets, as we know, the biggest part of the business today, we'll talk a little bit about that. But I mean the fruits of the company's diversification strategy really do look like they may now be ripening with auto and IoT and now potentially as AI goes mainstream, data center excitement. Even the handset piece, we'll see what it does in the near term, but it may have its own moment if and when AI broadly moves to the edge.

So we'll talk about that and probably many other things today. It gives me great pleasure to welcome Cristiano. So thank you so much for joining us.

Thank you so much.

You don't have like a safe harbor statement or anything...

No, we don't.

And again, I've been looking forward to this. Thank you so much for coming. We will get to the data center and diversification stuff. I think I want to spend most of my time there. But I think I do need to talk a little bit about the other businesses today. And I mean, like my own view, handsets are probably not in such a great place given memory prices, although I think you guys have tried to incorporate that into the forward outlook. In the meantime, though, I mean, IoT is recovering pretty well and automotive seems to be cruising along. And I guess maybe just to give you just an opportunity just to discuss sort of the recent results in near term. I don't want to spend a ton of time on it, but -- Just where are we sitting as we are like standing today, both for the existing business as well as maybe where the future may be taking us, like just given where we're seeing today, what are you seeing?

Yes. Look, I think, what we said on handsets, we -- handsets, it's kind of a stable business for Qualcomm. It is -- we have been very, very focused. I remember when I outlined our strategy for handset, that was kind of back in 2021 that we have an Investor Day, and we said we're just going to be focused on share of wallet. It's a market that it's going to -- it's stable. It grows single digit. I think we haven't still recovered in terms of total units used to be before the pandemic. But if you look at our performance, it's been great. We've been growing content. I think we've been growing -- the mix has improved. I think we've been growing share. And I think that's kind of we think about the business. We significantly improved the operating margin. Nobody was interested, not even ourselves in this whole price war that happened in the...

Yes, we've seen those in the past...

That's -- we had a completely different strategy in '21, focused on the premium and high tier and focus on silicon content. Now what the handset is right now it is artificially constrained by the memory situation. It's artificially. I think we have seen data points showing about the market was down like a 15% year-over-year. It is not a statement on demand. And a Qualcomm specific, I think, commentary that we provided in the last earnings call, which is we can now see the bottom in Q3. And the reason we can see the bottom is because we're significantly undershipping...

Not necessarily a bottom in the market. It's a bottom in your...

In Qualcomm. We are significantly undershipping to consumer demand. As you know, we have a sophisticated licensing business, and we have a lot of visibility on the channel. We have visibility in activations and what happens. So we can see even with the new prices, the prices became higher for devices. We're undershipping to market demand that give us confidence we're going to see kind of sequential improvement after Q3. And the magnitude of the market, I think, on handsets is going to go back to a larger -- once the memory situation gets resolved, and you go through the shift that we're starting to see right now that is happening on the market because of AI. Happy to elaborate on that. But that's kind of the 2 things going to drive the handset business.

We're very happy with our position. I think we had -- historically, we had on the Samsung about 50% share. Our baseline now is north of 70% share. And I think Snapdragon has become even more relevant when you have a smaller market. I think what we have seen from our customer base is for the high tier, the premium tier, then if I have to design for a smaller market, I want to make sure I design Snapdragon. So we've seen that dynamic play out. IoT, happy with the trajectory. I think a very diversified business for us. There's a lot of categories. Happy to elaborate what's inside.

And then automotive is this machine that we built. We talked about a $45 billion pipeline. We've been executing to that. And I think it's even exceeding our own expectations, how well is it performing. I think we just talked about now ending the year with like a $6 billion run rate, very comfortable when you look at our trajectory.

I can't remember what the target was, was it...

[ 29. ] We're very, very comfortable with that number.

Got it. So maybe that's just a good segue into this broader diversification story. So I mean, it was -- I can't remember the numbers said today, handset is probably still 70% to 75% of the chipset business...

It's smaller now. Memory has helped accelerate...

But I mean maybe talk a little bit about that diversification journey. And I guess it's not like the handset position that you have has not helped feed into that, especially in areas like IoT and auto, where I do think there was a lot of like crosstalk between the businesses. But maybe just talk a little bit about the diversification strategy, sort of like what set it off and where are you now? And then we can start to maybe talk about some of the individual businesses within there.

So I think before I answer the question, Stacy, I'm going to go maybe give you 2 comments, like higher-level comments. I think one is we are incredibly blessed that we have the handset business. The handset business has proven to be an incredible business. First, from an engineering standpoint, the handset is incredibly unforgiven. You have to pack, especially if you're focused on premium tier, you have to pack the computing performance you find today in PCs into -- and a lot of AI into a small device that has to manage thermals, cannot get liquid cool, cannot get hot. The battery has less out.

So it forces you to create a very comprehensive IP that has proven to be incredibly valuable as we diversify the company. And I gave us -- and I think you have seen that and you have seen even when you look at Apple, for example, their -- how handset can influence their computing strategy from the Neo to the M-Series, et cetera. So I think that is an incredible asset that we have in the company from a technology standpoint. The other thing is the handset become so concentrated as an industry that I decided when I became CEO, I'm going to do the opposite. I'm going to diversify Qualcomm to the max. And that's what we're doing right now. We're doing a lot of things in parallel because we're never going to see ourselves in a position again that it's highly concentrated like that.

And I think the strategy is working very well. I think the way the way we have done that, we have done with a mission with discipline. I think people can say a lot of things about Qualcomm, but our technology and engineering and execution, I don't think there's anything that people can say about our ability to execute. So first, we build a platform for the automotive. We saw that was an industry that we could be very disruptive. Then we look at the convergence between PC and mobile. So this is more scale for our IP, become accretive to the business. We've been executing that. We saw that there was an opportunity to drive a change in industrial. Industrial was about microcontroller and connectivity, and we changed it to high-performance compute and AI at the edge. We have built what we think is going to be a very important asset for the future of mobile platforms with AI.

We'll talk about that later. That's what we call in the personal AI category. And then we have been executing on our broadband business, continue to add assets robotics and then the next step will be the data center. So I think our -- by design, we said we're going to create a very diversified company. We're probably going to be very unique in having IP that goes from sub 2 milliwatts on an earbud to 2,000 watts on a data center. And I think the strategy is going well. We talk about $22 billion on non-handset by fiscal '29, not including data center. We stand behind that number. And we're very happy to see how some of those businesses are performing right now.

And to give you a plug, you have an Analyst Day on June 24, where...

Yes.

Presumably we'll get some updates to some of these numbers, I would say...

I think you'll get a lot of details, especially a lot of details on what we're doing in data center. I think people will be very positively surprised and please reserve a seat.

Okay. Let's talk about data center. So you've talked about data center before, different aspects of it. And there's a few pieces that I'm aware of, so you can correct me. We have the data center CPU, and you've announced something already with Humain, although I don't know that we've seen much in terms of actual revenue yet, but that's out there. You've sold AI accelerators for a while, several years, this AI 100, which was an initial product, but now we're talking racks, so AI 200 and AI 250. I recall a 200-megawatt deal with Humain that has been announced, although, again, I think we're still waiting for it to show up.

And then you announced a hyperscaler ASIC customer on the earnings call, which have been speculated and you confirm that, and there's been some chatter in the press on the identity of -- I don't know if you want to comment there or not. But I guess, first of all, maybe to take those one at a time. So maybe just to start with the CPU piece. Clearly, CPUs have been catching a lot of attention recently on the back of Agentic AI. And again, we have the Humain deal. I don't know if there's any other deals that are out there. And maybe you can talk to this if you would like to. But I'm curious -- and this almost goes across all the different categories. What is Qualcomm's value proposition, maybe to start with CPUs?

What is it about it that would make a customer want to buy a Qualcomm CPU for these tasks or for whatever tasks were versus something else that is available? Because to be fair, there are lots of other vendors that are already there and already established and scaled in that space. So what is it about Qualcomm that enables you to win in this market?

Okay. So a lot to unpack there. So let me...

I have similar questions about the other pieces, by the way.

No, Okay. So let me -- I'm going to give you as much as I can without front-running our Investor Day, but...

In a front run, by the way, if you...

But there's a lot to unpack. There's a lot to unpack. So first, I'm going to recap what we had said publicly, just to make sure because there's a lot of different information. We said publicly that we saw that as a data center now the next step and natural expansion is a massive TAM for Qualcomm and especially that our IP very relevant for that, especially the direction of traffic that data center was going. We said our data center strategy is going to be based on a number of factors, CPUs, we'll talk about that in a second. XPU, which is inference accelerator. We're not focused on training. We're focused on inference accelerator when we had a unique architecture, including how we think about memory.

We don't need HBM. So we have an XPU. And we have an incredibly large semiconductor machine. Most people don't know that we ship about 40 billion components a year that we're going to be able to do custom ASIC. The other thing we said is a lot of those markets are becoming bespoke markets. So we're very flexible. We're a new entrant. So those are the assets. What have we announced? The very first contract was a contract with Humain, actually not on the CPU. It's actually on the XPU is on accelerators. And we...

You came first, but maybe I'm wrong. Yes, you would know better than me.

I will talk about that in a second. And we're very happy that has been a process. I think we're now -- we started with the AI 100, but that's not the real scale is the AI 200, 250. I think we're on track on the execution. We're very happy with that. That's the very first one.

We're not having a shipment date for that yet.

So when that happens, we did say that we pull in when they announced you what data center would start having revenue in '28 -- fiscal '28, we pull into '27. The second thing that we said, we said we were very happy with the progress of our CPU with a U.S. hyperscaler. And we're very happy with that. And I think that is how we get started. And then given the traction that we have now with custom ASIC, we pull in now revenue that was supposed to be in fiscal '27. We said we're going to even see shipments within this calendar year. So that's what we said publicly.

For the ASIC?

For the ASIC. So that's what we said publicly. Now I'm going to answer your question. The first one was on the CPU. Look, I think if you want to support, I think, the valuation that you probably see right now companies I can tell, I think people realize that CPUs are very important, especially when you get mix of experts, you have orchestrated CPU, I think, is going to be a big asset when you think about the TAM. We have a pretty good CPU. I'll even say I probably -- when you think about the ARM compatible instruction set, we probably have one of the best CPUs in the industry, right...

Is that performance? Is it power? Is it TCO?

All of it.

All of it.

And a little bit of a track record, right? So we're the first one to put a 5 gigahertz CPU clocked on a smartphone. We did that. I think you saw what we did in PCs CPU. We also have our CPU with safety grade in automotive and the Snapdragon Ride Elite cockpit Elite. I think it's been second to none. So all of those markets, our CPU has been the test and benchmarks, and we have designed a CPU specifically for the data center. Very flexible architecture. It scales in the number of cores to a very big data center all the way to on-prem, and it's going to be -- this conversation is going to be relevant for future 6G networks. And it also can be a chiplet that can be part of a custom design. So that's -- I think the reason you should look at Qualcomm for the CPU is CPU performance, power consumption and TCU will matter a lot. And I think we have an asset that's been proven in at least 3 other markets.

Got it. Got it. What about the racks, the 200 and the 250? You mentioned some unique memory architecture. I mean, is that the value proposition of that rack?

So the value proposition of our NPU, I think the industry has been talking about XPUs and they have a lot of those, and they have different customizations. I think you have to look of where is happening with the data center right now. The data center is going to a process of disaggregated compute. And let me take a minute to explain this because -- this is actually -- I go back to what I said. This is actually very familiar to Qualcomm. The smartphone is the most disaggregated compute ever because there's -- I like to tell this story because it's a very simple example. When iTunes started, everybody buying their iPods and people will go have an iTunes running on your PC, the whole MP3 decode happened on the CPU and Intel CPU. Intel architecture has always been CPU does everything. You can never do that on the phone.

You have to have a separate hardware just to accelerate MP3. So the data center is going into that disaggregation. You started to see what it used to be one GPU from NVIDIA for training and inference. Now you start to have a different solution for inference. Now inference further disaggregate and prefill and decode. So we're building the XPU that is going to be very good and efficient in terms of compute density and power. for inference, some of those disaggregated workloads. We also have something that is very unique. We don't need HBM. We have a different memory architecture. I think we're going to talk...

Can you get the memory?

We're very, very confident about the ability to get memory for our engagements right now. And then the other thing I'm going to tell you is we're flexible. So we're building the chip. We're building the cards. We're building the servers. We're building the racks because different type of customers want different type of deployments. There is a value, I think, of at the system level, we can tap into the ecosystem that already exists for that. So those are all options of our solution on the accelerator. And happy to talk about the custom ASIC.

Yes. Okay. So the customer -- so part of this is you recently bought Alphawave, and that brings a number of components that I think are sort of table stakes for that. You have SerDes and other things. Now Alphawave used to talk about their own ASIC engagements. This particular -- I just want to level -- this is not something that was acquired with Alphawave. This was something that was...

It was the case, we will not get away paying $2.5 billion.

I get the question. So -- is there anything you can tell us about this ASIC? Or do we have to wait a few weeks...

So here's -- so there is a science -- I think there's science behind Amendus in Qualcomm, especially when you look at M&A. I'm going to give you a couple of examples, right? I think we bought Arriver. [indiscernible] had to do with [indiscernible]. We bought a Tier 1 Veoneer, got Arriver spin out. We bought Arriver because it was an important asset to complement everything else we're doing for automotive and help us complete the platform. I think we bought NUVIA to help us accelerate our efforts to design our CPU, combined with everything else we had and we enter some of those markets.

We bought Alphawave because it provided a lot of IP that we needed, a lot of the connectivity IP, a lot of the expertise on customizing combined to the -- what we had in engagement and then allow us to go. That's why we think we're very focused and consistently said, now we're serious about data center. We're missing some assets. We're going to go buy Alphawave, and we're going to complement the platform. And I think what that happens, there are 2 things that happened with the acquisition of Alphawave. One, it complements our portfolio. Second, it builds a lot of confidence, I think. Then what do we bring to the table custom ASIC?

We're not a small semiconductor company. There's a lot of companies that they decided to enter the semiconductor business, they probably have not a lot of experience doing a chip. We do have a lot of experience doing a chip. If you look at the number of chips that we do in leading node, we also have such a sophisticated -- and I'm going to say this, I don't want to brag. I'm just going to give you the fact. We have such a sophisticated machine to build SoCs. For example, what we do in mobile, like clockwork on Snapdragon 8 every year, we tape out a chip on a new node. When TSMC finishes the mask, we go to production. We don't even have to get the chip back, how sophisticated our [indiscernible] is. All of those things became incredibly attractive, I think, for companies they're trying to do bespoke solution. So it's our IP, a lot of the IP that comes off of wave, our ability to do this and our willingness to be very flexible. This has been incredibly helpful and kind of accelerate our data center ambitions.

Got it. And so you've talked about -- I'm assuming we'll hear more about the revenue outlook and everything to grow that like in June. But you talked about like -- I can't remember the wording substantial or meaningful revenue.

Material.

Material revenue in '27. What does material mean?

Well, just look at the Qualcomm revenue scale. Material has to be in the multiple billions of dollars. That's really what it means.

Okay. And what about the margin structure of these? I mean we're seeing a lot of other ASIC. I mean, Hawk over there, Broadcom gets pretty high margins in ASICs. A lot of other players don't quite get that kind of margin. How should I think about the margin? And to be, I guess, your margins right now are more mobile related anyway, so they're probably not as high as -- but I mean, can we think about this as being margin -- gross margin and operating margin accretive?

Operating margin accretive.

Operating margin.

It is -- all of those things that we're doing are going to be operating margin accretive for the company.

Okay. What about gross margin?

Well, say, I think we did provide a framework, I believe, for the gross margin of the company. The reason I'm going to hesitate about gross margin right now is because a lot of those engagements, some are chips, some are cards, some are servers, some are rack. So they kind of fluctuate. But I think it's going to be probably very healthy, and it's going to be about making operating margin accretive at the company level.

And the material revenue, that's across all of these things. It's not like it's CPUs and XPUs and ASICs collectively material.

That's correct. I think they all start ramping, but it is such that we -- like I said, we're going to -- it will pull in. So we're going to start shipping in the calendar year '26, '27, it becomes material for us.

Are you going to split it out as a segment at least so we can look at it or -- are you going to split it out as a segment at least so we can look at it or...

I haven't thought about that right now.

Okay. Okay. Okay. That sounds good. Let's talk about some of the other areas. So I mean, you mentioned Arriver in the context of automotive. Let's talk about automotive. So you've had these targets out there for a while. Again, it was $8 billion by 2020. $9 billion by 2031 or whatever it was. And you've been sort of dead on the trajectory, if not even ahead of it since then. So what are you doing in automotive now that's actually driving that? I guess Arriver brought some of the self-driving software in, but like what is a role does that play? How do I think about the split of maybe the revenue as well as the order backlog by things like infotainment versus like maybe the more attractive autonomous aspects of this?

Okay. So let's break this down in 2 parts. Let me talk about what the revenue that you see right now and go into the trajectory. And then I'm going to tell you some of the trends that are driving, I think, an acceleration and increase of content in automotive. Let's kind of break those 2 things. We did talk about publicly about a $45 billion pipeline. I think the pipeline has increased since that time. I won't disclose the new number now, but it's possible we're going to disclose that at Investor Day. And that has been converting into revenue. I think you see share gains, new models with Qualcomm silicon. That revenue today and that $45 billion pipeline, you should think about 1/3 is about all that's happening in connectivity on the car.

Another one is the digital cockpit, all the computer to power all the different screens. About 1/3 of that is ADAS. ADAS is processor, mainly processor for autonomy and ADAS, and then it has some components of the stack. This is kind of what is driving that growth. And you see those growth rates significantly higher than the market growth. I think we guided 50% year-over-year is because we're gaining share. I think those new models launched with Qualcomm technology. The exciting part is the second part. A couple of things happening in the car right now. First of all, the Chinese is accelerating, I think, premium content in the car. Actually, what they have been done right now is breaking records from silicon to SOP, and that is driving a lot of more premium computing content.

Are you talking about the speed at which they can get the stuff adopted?

New chips adopted and put it into cars and especially because -- the car right now is also being -- computing is being pre-provisioned in the car because over the lifetime of a car, we're starting to see a lot of Agentic AI experience coming to the car. We can talk about that across the board. So that is -- and it's China, you see that happening right now. The other thing that is driving a lot of the content is an acceleration, I think, of Level 2++, especially in premium car, more processing, I think, for ADAS and autonomy.

The #3 driver is what's happened with the stack. So we just -- we develop an asset, which I think is a very good asset. It's basically a Level 2+ safety stack that we've done with BMW and has been launched in the BMW iX3. We're getting a lot of interest from other OEMs. Some already engagements of that. And I think that's in combination with GenAI end-to-end stack. You always have to have a safety stack that go along the end-to-end. That's driving an increase, I think, of content for Qualcomm and cars alongside where we're seeing a lot more processing capabilities that is happening on the digital cockpit because of Edge AI as well as the ADAS transition to FSD, which is going to be like what a premium car would expect to have.

Got it. Got it. Maybe talk about IoT. And there's a couple of different pieces of this. And I'm going to go back to your prior Analyst Day. You sort of at the core business, which at the time was sort of at the bottom of the trough, I think. And you were sort of calling for growth of that business kind of reaching the prior peak in 5 years. It was, I think, getting to $8 billion or something. And then you had another $6 billion in the target, and it was $4 billion from PCs and $2 billion from XR, VR. And my at least outside-in view, it looks like the core IoT piece is in recovery, and that's great.

I feel like the XR/VR is maybe overachieving. And I guess on the glasses, to be fair, it's actually been on my big stack of things to do is to look at that market because it does seem like it's there. The PC piece feels like maybe it's lagging a little bit versus the trajectory. We haven't seen -- and now I mean, PCs just as a market are probably facing some of the same issues that smartphones are facing on that. Maybe if you could talk about the different pieces of IoT and what your perspective is on sort of where those markets are, where they're going and the growth?

Very good. Look, I'm just going to say, look, we're very happy with the trajectory. It's going as planned. I think that segment is also moving towards all the targets we set up for fiscal '29. Just to give an idea, I think we guided the IoT segment for next quarter. It's, I think, somewhere in the $1.8 billion in change. So I think it's kind of starting to approach like that $2 billion a quarter business. So you can look at the projection and you look at the growth rates to see how we're marching towards what we said we're going to do in fiscal '29.

No, let me unpack what's in IoT. there. It's a good question when you ask me about segment reporting because eventually, I think as those businesses mature, we're probably going to be thinking about how to break it down because there's a lot of things in IoT. And I think in hindsight, it wasn't a great term to represent what is in there. But anyway, we'll get there. So the first thing that you see there is the compute business, tablets and PCs. I will come and talk about that. The second thing that you have there is you used to have VR, XR and wearables, which is now all wrap around what's happening with personal AI devices. So that's in there. You have our broadband business that's in there. You have our industrial business as well. And it's kind of you kind of -- we often talk about -- if you look at some of our end markets, consumer, industrial and networking, it's kind of put PCs in the consumer side with wearables networking is broadband and industrial, it was about 1/3 each one of those segments.

Now let me tell you what's happening right now. PC, we're actually very happy with the trajectory. I think we've been tracking -- if you look at our projections for $4 billion, it's around in the order of magnitude, about 10% share. The markets that we launched, we're tracking to that. The problem with PC, you have to deal -- you have to have the patience because the first -- to be fair, you've sold into that market for like over 10 years, right? I mean it's -- it was not until Windows 11 that finally the emulator could run Win 64. And so you had a second-class Windows for ARM. And by the way, we have to do all of the heavy lifting ourselves. I think the effort to get ARM to run natively, we did it with Microsoft. No other company did it.

And what happened -- and of course, Apple did it for their ecosystem. But what's happening now, finally, on the consumer side, you now have -- there is no difference between an x86 and an ARM. It is now very mature in the consumer. We see that in the markets that we launched consumer. We're tracking in the consumer markets that we -- United States, top 5 Europeans, we're tracking 10%. We look at the position, the design traction in the channel, very strong. We're happy. That is now -- it's just going to get its own scale. We took the past year getting games and AAA games and all those things have to be more mature on the laptop side. Now we're tackling commercial.

Okay. To be fair, the 10% is 10% where you play...

No, no, no. If I get 10% of the market, I will already be at the $4 billion target. Yes. But you have to have this discipline. Now we're in the commercial, which is very important and which is the majority, I think, of the profit pool is. So we've been working. We have about 500 enterprises. We're going to get their CIOs comfortable. And I think Microsoft has been a great partner of this, and it's going to get there. You just have -- you cannot burn any step. But when I look at the [indiscernible], which is excellent because it kind of validates this convergence between mobile and PC.

And I think about the performance of some of the products we're building and what's happening in AI, I'll talk about it in a second. I think that's very good. We just have to have patience. The problem is both tablets MPCs have the same memory dynamics. There's no memory. So the market is artificially constrained until that's resolved. The other one in the IoT segment is what is now what we call the personal AI devices. I am incredibly optimistic about this. And if anything, I kind of agree with you that when we talk about the XR $2 billion by fiscal '29, I think we're beyond comfortable there. I think because of what's happening with those new mobile category of devices.

Every AI company, every AI company and every cloud company from the United States and China and other places are looking those new kind of devices. You hear here and there in the press. We're working with all of them. We have over 40 designs of what is called this personal AI category...

Give us an example, what kind of devices are we talking about?

Okay. The one that is the most popular and is going to get the most scale is smart glasses. And smart glasses come in 2 forms: processing, connectivity, cameras, and then you may or may not have a display. Why glasses? Why glasses? And you have to really understand what's happening there because it has nothing to do with the definition of a wearable or smart glasses before. As AI now has a role in large language models, large visual model in the man-machine interface, becomes very natural for glasses which close to your sensors, close to your eyes, your ears, your mouth, you turn your head, you're watching it and then you can access a model.

And what we're starting to see right now is a little bit of the dynamic of the smartphone, not the same, but similar. Every week, somebody adds a new workload, a QR code, a payment system gets integrated, like, for example, some of the digital payment systems has been integrated, all of those glasses. So glasses is the big one. I think it's already in the multiple tens of millions of units. It could become 100 million units. Eventually, if this is successful, it could become as big as phones. The second one, I will put it in the...

Are you still buying phones in that scenario?

100% you're buying phones. Yes. Because if you give me time to talk about what's happening with the AI transition of phones. -- you'll see the role of both. Now the second form factor that is getting traction is a combination of pens, pendant, jewelry. I actually -- I'd be careful what I say, but there's some other form factor very interested. Go watch Microsoft build. go watch Microsoft build. You're going to see some interesting things there, too. I think there's going to be -- the form factors are changing because it's related to the easiest way for a human to get access to a model for a completely Agentic use cases that has no legacy.

So that is now we call personal AI category. And I said we have about 40 designs of day nature, all connected to models. The other thing in there is our broadband business, stable. It is -- we're happy with it...

This is like networking gateways and...

It's access points, retail, carrier, and for enterprise as well as broadband. We also now have wired connectivity. We've been executing our FGS-PON. -- we're doing -- and then also edge processing happening for AI at the broadband business. And that's going to grow for us with 6G. We're going to take a role into that area as well. And then the last one is industrial. Very diversified, very sticky. On industrial, I'll tell you right now, we're not bound by demand. We're bound by speed of building solutions across verticals. We have focused on retail, oil and gas, energy, manufacturing warehousing...

It really does require vertical solutions, I guess, in...

You have to have the silicon, some of the AI capabilities, it's common. Like think about, for example, computer vision, it's common. I think the -- what is the edge appliance, the smart cameras. But then a lot of the players and the software that goes on top is building an ecosystem across each vertical. So that's what is in IoT.

Got it. So let's talk more broadly about AI at the edge. Talk to me about AI and smartphones.

Okay. Why do I want an AI smartphone? You want it for sure. So it's really important. I need a few minutes to explain this because we have clarity now. Like I didn't have clarity. We knew that this stuff was going to happen. We have different pieces. But now we can see everything together. And the picture, at least for us, is super clear, and we're starting to see what the change is going to be in mobile. One thing I'm going to tell you, from the early days of cellular, from analog to 2G to 3G, 4G, 5G, one thing I can tell about cellular it changes every time. It changes, the players change, companies go from hero to zero.

We're fortunate enough that we survive all of those transitions. And we always want to look to understand what is happening, how is that transitioning. Now we have clarity. I want to explain that to you. So think about what happened with AI. First thing that happened with AI is you have chat box, people go in, they ask something to a chat box and a search and get a response. Then of sudden get reasoning, you start to generate tokens, but then something incredible happened. The OpenClaw happened. Those orchestrators happen. That's what's driving a lot of the CPU. You have now -- you have the ability to get a bunch of agents, you have a mix of experts.

You have different models for different things, and you have this orchestrator that does thing for you. And it started to operate your computer for you. It's so profound that if you're a SaaS company today, you have to design your SaaS software for a client, not only to interface a human, but to interface an agent because the agent is going to go to the client. And you start to see what people are doing OpenClaw, you send a WhatsApp to your computer and they start doing things for you. And that's how computers are going to get utilized and those things consume a lot of tokens.

So you're going to start to see now finally, somebody said, oh, this hybrid AI that Qualcomm has been talking about starting to make sense with different models, mix of experts and smart routing and all of that. So what's happening on phones? So what happened on phones, you have to understand there are 2 fundamental changes happening with the phone market. One, I'm going to give you an example of what happened when OpenClaw and all those different claws started. People started buying Mac Mini. Apple was sold out of Mac Mini. They put the Mac mini, they put the thing and run. Okay.

But you can run it on anything, not you didn't need a Mac Mini...

No, but that's not an interesting observation. The observation is in the phone, you only carry one device. You cannot buy a backpack, put a Mac mini, a car battery and walk around. So you're going to have to have it in your phone. So your phone is starting to get the double personality right now. You, as the human will pick up your phone and you do phone things. You're going to go to your apps and you're going to do the things you do every day of your phone.

Actually, the reason glasses are getting traction because it's not very natural for you to talk to the phone or pointing the phones to things that needs you're going to do your phone thing. But the phone now can also run those orchestrators. And then the phone is going to start doing things for you on your behalf. And it's going to be doing things for the other devices that you interact on your behalf because the agent take the center pay -- it's not the phone, the center, all the wearables are connecting to the phone. Now it's the agent in the center.

On the phone or is the agent in the cloud somewhere...

Orchestrator runs on the phone. And see, this -- I'm going to say this to be provocative. I have seen so much discussion about cloud and edge, cloud and edge. It's a useless discussion. It's almost like for me to ask you to go to your 300 apps that you have on your phone and let's say, let's say one by one. What runs on the cloud, what runs on the end. put an airplane mode, they're going to say useless. -- is a brick. It's probably an iPod Touch, right, with all the Qualcomm IP. So then you're going to think about it, this is going to be cloud and edge working together.

You see all of those orchestrators that run on a device, they do some things on the cloud, MCP protocol, things that go to your app and they start doing things for you. So what we're starting to see right now is the orchestrator comes to the phone and start doing things for you on your phone or for other devices. You may do something on another device and your phone is going to do something for you. And that drives the second change. you're going to see right now between now and summer, every one of the phone -- every one of my phone OEMs right now in China, they have their claw -- the claw equivalent. Coming into the phone, you're going to see every OS vendor talking about a claw as part of the OS.

And the interesting thing is the nature of the business is changing. I used to think about the OEMs as the customers. Now I think that the OEMs and the cloud companies as the customers. Just look, for example, what's happened in China. Some cloud and AI company actually took over the UI of the entire phone OEM and that's going to be the new change in phones. And I think that is driving a completely different discussion. On one hand, you have the existing OEM thinking about memory cost is too high. I need -- I don't know if I can afford all of this compute, this is a smaller market. Have an AI company said, I need more compute. I need more compute. I need to have the ability to do a lot of those things into that device. That's going to propagate to the PC. It's going to propagate because it's 6 billion people use the phones.

So yes, you're Mac Mini separate, you're going to see every OS vendor is going to talk about a claw and you're going to feel which one is secure, not secure. And you're -- and those are going to be token machines. And that's going to require a lot of competition. And leads me to the last -- that's going to happen to your access point is going to happen to your car. And the last part of this conversation is this concept of hybrid AI. I am going to speak at Computex, and we're going to show a couple of interesting demos at Computex.

Are you leaving there like -- you're going there shortly, I guess, right?

Yes.

And what you're going to see in some of those demos is how much is going to cost you when you start having those token machines running everything in the cloud and how much is going to cost when those cloud companies push some of those models to the device and they break the tasks up. I think we're starting to see a lot of conversations about how companies are thinking about how much you can afford, what is open source, what is not. All of this is goodness that is going to drive an upgrade. And then my last comment on this, none of the existing devices today...

So this is the killer app, though, that we've all been waiting for.

It's going to be those orchestrated and agents that's going to be generating tokens. And we're going to finally -- finally, we're going to start talking about what the real conversation because the real conversation in the early days of smartphone, if I have to tell you that the smartphone experience is going to be defined by what the OS vendor put into the OS or the OEM put into his apps, that is nothing compared to the apps the developers did. Those agents are going to be some of the applications. And I think the good thing for Qualcomm is none of those devices are ready for it. So you're going to have to go to an upgrade cycle. I can't predict the timing, but we now have clarity how those are going to play out.

Got it. Got it.

The phones today don't have the computing power that are necessary to orchestrate something like this, do they?

The -- right now, you're going to need much faster CPUs to be able to run those orchestrators. And then eventually, as some of the hybrid AI matures, you're going to need accelerators as well.

Got it. Got a few minutes left to go to the lightning round, see some of the audience questions.

I guess it has to be asked. There's the quintessential Apple question, which is here, which is baselines now that Apple starts leaving the model in a bigger way this year, do investors have the right trajectory in mind? And I would follow this up by also saying Apple's licensing business, which does have an expiration date, the current agreement, I guess, beginning of April of next year. So what are your thoughts on that?

Look, I hope investors have the right trajectory. We have been so clear...

We've been 100% clear, by the way. There's no excuse, frankly, if...

They don't have the right trajectory in mind, but anyway...

All right. Second question. Look, we -- as you'd expect, we're in conversations about the license. The license is independent of the chip business. And I will say, I think the licensing business is probably one of the most stable times of our licensing business. I probably feel very confident about the state of the licenses business and...

Probably we feel very confident.

And look, I think you have to ask yourself, especially a lot of those customers' renewals, this business is being battle-tested. So I think there's a lot of clarity about what the patents are, what the claim charts are, I think the size of the patent portfolio. And I think what I have seen from most of my customers right now, especially in the AI transition, the last thing anybody wants in the middle of an AI transition is a worldwide patent litigation. I think everybody knows, I think that we're always going to defend that business. And I think -- and the fact that we always have done that, that in itself, it's a positive thing that brings stability. So I am -- I will repeat. I think it's probably one of the most stable times for licensing business right now.

Okay. Is there any truth to the idea though that like the last time Apple settled is -- I mean, they had no choice. Intel was late with 5G, the only option less of an issue now they've got other options.

I think you should follow the court case.

Okay.

I guess -- to be fair, you've had a lot of disputes over many years. And I think eventually, you've come out on top on every single one. I think that is true. So one more question here. What's the next -- given memory issue hopefully is abating? I don't know if it's abating yet, by the way, but hopefully abating. Handset market share is well forecast. What's the next thing that concerns you? What are you most worried about? Is it competition from China? Is it something else?

Look, I think if you're in the semiconductor business right now and you're a global company, you have a lot to worry about. But I -- look, right now, for us, I think we can't wait, I think, for the memory things to improve because like it's just going to be an immediate acceleration, I think, of our business. I think we've been.

And there any signs, by the way, that is, are you seeing any improvement or getting worse?

It depends. I think it depends on how you look at. One thing I'm going to say for this, especially on DRAM, we have seen a lot of investments in China, I think, to build DRAM capacity.

And you're qualified with the local Chinese guys, correct?

If you have a memory right now, I'll qualify your memory. I'm qualified with everything. So I'm more optimistic than people are about maybe '27, there's more capacity coming in, in line. So that's one thing. And the other thing is we're like laser focused on executing on the data center. We see a great opportunity for the company. I think we're being positively surprised with the traction, and we need to make sure we're delivering on all the expectations.

Got it. So we've got about 40 seconds left. As I always do, I'll give you your soapbox, you've got a whole audience here. Why should these fine people buy your stock?

Look, one thing I tell about the company, and I'll give them my bias. I then I joined the company as an engineer in 1995 in the company for 30 years. So Qualcomm has an incredible, I think, engineering and technology capability. I think we also -- our patent portfolio, it's kind of reflective of our innovation engine. All those markets that we enter one way or the other, even some of the markets were new players and we have to grow, we -- if you look at our execution, we have the leading platform.

And I think Qualcomm has been on this trajectory to be a very diversified company. I will never bet against Qualcomm execution. So I think investors, if they understand the capability of the company, they understand the trends that is happening on AI. It's fascinating to see when everybody thought everything is about the GPU. Now people don't talk about that anymore. There's a lot of new engines. If people didn't talk about the edge before, now they talk about the edge. I think the company is unique because it has bets in many areas that are now -- have seen positive trends in AI. And we're just going to be on this mission. And hopefully, June '24, we'll have a very compelling story to see why this is an incredible long-term opportunity.

Got it. With that, I think we'll close it out. Thank you so much.

Thank you. Thank you very much.

Thank you. Good afternoon, everyone. We are here to host the fireside chat with Qualcomm and one of my favorite sessions from the -- at the conference every year and with one of my favorite people here, Akash, who we were discussing outside, proves me wrong every time. It's always a good time. Akash, thank you for being here. Akash, who is the Chief Financial Officer and Chief Operating Officer of Qualcomm. Thank you for the time here.

A lot going on. So let's get into them one by one. You've talked about 3 different opportunities in accelerators, CPUs and custom silicon now over different points of time. Let's maybe start on custom CPUs. That's driving the most amount of discussion with shareholders or investors at this time. How should we think about Qualcomm's opportunity, differentiation in custom CPUs, and remind me of timelines and any acceleration on the timelines as well?

Sure, sure. So first of all, it's great to be here. Thanks for hosting. So we're very excited. I think we've kind of waited for the right opportunity to have a set of assets that's extremely relevant and material to what's happening in data center. And so at our earnings recently, we talked about how we're going to enter the data center business. And there's 3 key areas that we outlined. The first one is custom silicon. So to Samik's question, what we discussed is that we are working with a hyperscaler, and we're going to have revenue starting later this year from that engagement. So very excited. This is, I think, an opportunity that's going to be very material for us in '27. And then hopefully, it's a stepping stone to an ongoing engagement and then scaling up from there.

So we're leveraging the fact that we have an incredible technology portfolio. We recently acquired Alphawave and they have -- they have been doing custom silicon for a long period of time. So we are leveraging their expertise as well. And then as these hyperscalers look to make chips and look for partners who can help them with the silicon execution, Qualcomm quickly emerges at the top of the list. I mean we've -- Obviously, lots of experience in doing chips. We have expertise in 2-nanometer, 3-nanometer, 4-nanometer other technologies. We have obviously tremendous scale at the foundries as well. And all of these things become very relevant when the hyperscalers are looking for a partner. So happy to have this engagement. I think this is an opportunity for us to grow from there.

The second area where we'll be entering in data center is the CPU. As several of you might know, we have a custom CPU that we deploy in handsets, where we are the performance leader. We also deployed that same CPU in PC. And so if you compare us to the x86 players, Intel or AMD, we think we have a very significant performance advantage. And we're going to bring all of that to bear in a data center CPU solution. So that comes in as well and excited about how the CPU use is changing in data center. It was a very large market. I think it's expanded very significantly now. And so as agentic comes in, the role of the CPU in data center expands, and now this allows us to play in that larger market.

The third thing we're doing is we're building an AI accelerator that is really optimized for certain workloads and inference. So as the data center kind of drives growth in inference, within inference, you have Prefill and Decode and what we're building something that is really optimized for certain workloads that happen in decode, and so very excited about that as well. So it's a series of products, we think of this as something that layers into our portfolio over the next couple of years and excited about being hopefully a very significant player in data center going forward.

So custom silicon this year and the other 2 CPUs and accelerators over the next couple of years from a timeline perspective?

Yes. We think those 3 layer on top of each other, right? And so happy to kind of have that come through. If you look at several of the chips at data center players, they're focused on 1 or 2 of these areas. And the fact that we have such a large portfolio of technologies and now it's coming together into products, we get a chance to play in all 3 of them. So very excited. I think it puts us in a unique position in data center and semiconductor companies. And you know, we've not had a position there in a very, very large market. So super excited about it.

Let's talk about competition and maybe starting with CPUs, how are you thinking about competing against ARM, which is now vertically integrating further?

Yes. So we compete with ARM at edge devices, right? So we have a core that's a custom core. If you look at our phones, we use that custom core into our phones. Our competition uses a custom ARM CPU in those devices as well. And if you look at the performance of the two, we're pretty comfortable that we have a significant performance advantage when you compare the two.

When you take that to PCs, we're using the same in our Snapdragon PC products. And if you compare our CPU core to the x86 ecosystem, we think we have an advantage as well. So I think we are unique on the edge where we use this custom core and we have an advantage across the other players, and that same advantage shows up as we get into data center.

Interesting. Okay. Accelerators. You talked about the opportunity with sovereign customers. How is the target customer for maybe AI200 different from AI250? And how are you thinking about customer adoption right now?

Yes. So I think this product line, the AI200, 250 product line is very much a merchant product line, right? So we're developing it and we'll make it available to really all hyperscalers globally. And we think we're bringing our low-power heritage that we've worked on for several years on the neural processing unit, NPU, on edge devices. We're bringing that to data center. We're bringing some novel techniques in terms of how we solve the memory bandwidth problem in decode solutions for inference.

So it requires us to kind of combine memory along with logic. And it's a unique way of solving the problem. We don't think there's anyone else in the industry who does that, and we think it brings tremendous performance advantage in terms of how strong the memory bandwidth performance is. So excited about that as well. We are engaged with multiple customers. And then, of course, looking forward to giving a lot more details when we do our Investor Day on June 24th.

Yes. So any more color on what -- how is the product line for AI200 different from 250, Who do you exactly target as a customer?

So we'll talk about it at our Investor Day, so I don't want to front-run it. But I think you should think of the -- a lot of the innovation that we're doing coming through with the AI250 product and AI200 becomes kind of the set up for that launch.

Okay. Maybe just taking a step back, I think one of the questions we hear often on these initiatives that you have is, okay, you have a design here like is it that easy to come into the market, break into the market and scale. As you're thinking about it, what are the primary hurdles you think about as you have to scale these platforms and address customer demand? What are the primary hurdles you have to overcome?

So as you know about Qualcomm, right, we're all about technology. And so we have a very, very strong portfolio of technologies. We're bringing all of those to bear. We're very confident that we can take those and deliver chipset solutions at scale, whether it's for custom silicon engagement or whether it's for merchant engagement. So I think given our track record, hopefully, everyone in the industry believes that. One of the things that we had to work through is the software ecosystem. It doesn't apply to the custom solutions because those are built custom for a hyperscaler and they are typically working on the software. But as we get to CPU, which is porting workloads onto the ARM architecture, which has largely happened already, but there's some additional work to do. And then when we get to our merchant solution for AI accelerator kind of fitting into the architecture that already exists and supporting the industry standards on it is what we're working on.

So that to us is kind of normal execution when we get to a new market. But when you start from a position of strength where you have competitive differentiation, if you think about the data center players, most important thing is how does performance per watt -- power is a very important metric now in data center and performance per watt leadership is critical. Performance per dollar leadership is very critical as well and both of these translate into a lower total cost of ownership. And so when we engage with customers, really the key question is, can you deliver significantly lower cost of ownership, lower total cost of ownership. And that's where the advantage of Qualcomm comes in through performance. So what we think of as best-in-class performance per watt translates into a better TCO. And that's what the hyperscalers are looking for.

Okay. Akash, another follow-up on this is a lot of constraints on the industry from a supply perspective as you scale, you need to have visibility that not only that there's demand, but that you will be able to meet that demand with supply. What have you -- what actions have you taken on that front already to secure that.

Yes. And so I think one of the advantages of Qualcomm is just the scale we have, right? So we're obviously a very large player for TSMC, a very large customer for them. We work across 2-nanometer, 3-nanometer, 4-nanometer. We work across foundries as well. And so when you -- when a hyperscale customer looks at Qualcomm as a supplier, the scale that we have in the industry, the knowledge that we have of leading nodes, the experience we have in building complex chips becomes very important. And so I think given our scale, we have some flexibility across our product categories, across the nodes, and we're able to leverage that into having supply assurance.

Okay. Maybe let's move to the third one of that order of initiatives or sort of areas of growth that you're looking at custom silicon. How is it different from the other 2 opportunities, CPUs and accelerators that you defined? And maybe more than that, how much of it was an opportunity Qualcomm could have addressed organically relative to the Alphawave acquisition helping you address it?

Yes. So Alphawave has been in the custom silicon business for a long period of time. They also have key connectivity IP, SerDes, optical other things that become also important when you have some of these engagements. And so you kind of take that advantage, combine it with the technology portfolio and scale of Qualcomm and the combination becomes a very interesting alternative to an industry that has a couple of suppliers still now at scale, and we become an alternative to those suppliers. And what we're seeing is really the interest from the industry in having a new supplier that enters with our scale, with our technology competence. And the traction we're getting is a result of those factors coming together.

Okay. Got it. Before we move forward, just your competitors on the CPU front are pretty well known. When you think about your opportunities in custom silicon and accelerators, Who would you define as the competitive set that you'll be running up against just for investors to better visualize that you will have different competitors in these aspects, right?

I mean I think it's well known in the industry, but the custom silicon, the largest players who are probably are Broadcom and then maybe Marvell is the #2 player. When I think about the amount of assets that we bring together, especially between us and Alphawave and the scale that Qualcomm has, I think we compare very favorably to the existing players. And if someone is looking for one or two alternatives, I think we would make that cut.

Okay. On the accelerator front.

Sorry?

Accelerator?

On the accelerator side, I think the same framework applies, right, whether you're doing a custom CPU solution or custom accelerator solution, it's the same set of players who are competing for those sockets. And I think what I outlined just applies to this as well.

So now that you are taking a view on these 3 opportunities, you are expecting them to layer on over the next few years. But is there a view yet -- early view yet on which of these 3 will be the largest for the company medium term, if we can rank order them in terms of size of how they play out?

I think just generally, when you look at the SAM in the market, by far the largest, and this is not a largest opportunity for Qualcomm comment, this is just a TAM comment. The accelerator market is the largest market. Now the CPU market, especially with agentic workloads coming in, there's a significant change upwards in the size of that market as well and then you have the custom chip market. So those are the three.

But the way I think about it is Qualcomm has an opportunity to be successful in all three. And each one by itself would be extremely significant. So whether you believe we can be successful in three of them, two of them, one of them, each one would individually be very significant to our financials. So super excited. I think we always -- we've always believed technology always wins in the long term, and we're going in from a position of strength. We are going from a position of technology and excited about what we can do for our customers there.

Okay. Great. So maybe let's move to another end market since we've talked about data center for a while, autos, it's remarkable what the company has done in autos and position itself to be a leader in -- I want to say short time, but we know autos takes a bit long cycle. But relative to some of your incumbents, you've done really well. You guided to an acceleration in the growth into the June quarter as well. So maybe outline for us what are the incremental growth drivers that are benefiting the company? How sustainable is this acceleration that you're seeing?

Yes. I think, first of all, super excited about what's happening in the auto industry and our role in it now. We have 3 set of products that make up the Snapdragon digital chassis. The first set of products are connectivity chips. I would say it's clear that we are the global leader in that. The second is digital cockpit chips. So these are chips that empower the screens inside the car. And clearly, we are the leader in that as well. And the way this -- and then the third is autonomous driving chips as well and then stack that goes with it. The way this has layered in is very similar to the data center conversation we had.

We started with one area, which was connectivity, then we had Cockpit and now we have ADAS. And what you're seeing come through in the recent quarters is this take-up rate that is ADAS, that's now on top of the other 2 which are still growing in themselves. What is interesting in ADAS is as we go from 1 generation to the other, the industry is accelerating, right? Companies are looking to go in from Level 1 to Level 2 to Level 2++ to Level 3 and Level 4. And the silicon content increase across all of those is extremely significant, right? So Level 3 going to Level 4 is what is coming through in our financials right now. Later this year, we're going to start seeing OEMs, car OEMs in China start deploying Level -- sorry, our next-generation Gen 5 chips as well.

So a very strong portfolio of chips, very significant increase in content as you go to more capable chips and more capable autonomous driving capabilities. And then Cockpit still by itself continues to grow. So the setup is perfect. Our product leadership is, I think, very significant and clear. The entire industry has a very good understanding of it. But agentic AI now is coming on top of it, right? What used to be the Cockpit experience where you had to touch the screen to do everything in the car increasingly is going to become an experience where you're talking to the car and you're having an agentic AI conversation.

The car has the capability to understand the driver, understand the preferences, things like directional mic, being able to ask questions about the performance of the car, turn up the temperature, all kinds of things. And so I think transformation is continuing to happen, more digitization in the car is great for Qualcomm and you're seeing that come through in our numbers.

On the ADAS front, there's been a broader conversation about the impact of AI, right, on autonomous driving? And whether it makes it easier for new competitors the space with the help of AI on the software stack. Are you seeing anything change on that front? Where do you -- given your position now in terms of the pipeline that you have for ADAS wins? Where do you see the competitive moat that sustains your leadership here?

Yes. So I think our ADAS leadership is multifold. I'll say the first leadership is we have a single chip that does digital Cockpit and ADAS. So for lower-tier cars, that solution is optimal, and we create a sandbox environment where you can implement both at the same time. The second leadership vector is how strong our stand-alone ADAS solutions are at the premium tier and how -- when we go from 1 gen to the other, the content expands very significantly. And then the final, the third vector is bringing in an ADAS stack that we've developed working with our customers. And then we are advancing that towards now Level 3 and Level 4. That's the third vector of differentiation.

And it's really all of these things coming together in an end-to-end platform that if you're a car OEM and you're looking for a technology partner because this is not just a supplier relationship, this is a technology partner relationship. Our portfolio puts us in a great place and that's what you're seeing come through in the relationship with the customers.

Got it. Got it. Maybe let's move to talk about your opportunity in physical AI and how different or similar is it from what you're doing in automotive. And how, again, are you thinking about the market scaling in relation to physical AI?

Yes. So super excited about that as a long-term opportunity for Qualcomm, right? So new areas that we are pursuing in addition to what we've been doing in automotive and IoT is data center and robotics, right? Robotics is -- it's an incredible extension for us from what we are doing in automotive. In some ways, robotics is kind of the next step of ADAS. You were able to leverage all the technologies that we've developed for ADAS. But automotive is even more suited to Qualcomm's strength. Robotics is even more suited to Qualcomm's strength because you need to be -- have like very low power consumption, small form factor, wireless technology, very good camera, sensor fusion. All the things that Qualcomm does well comes together in the platform that's required for robotics and so very excited.

I think we're at the very beginning of what this market is going to turn out to be. It's going to be a massive market. There's no question. The debate is really just how long it takes to get there. And for us to be able to work with each of the leaders in the industry, both the U.S. and the global ecosystem pretty excited about it. I think we've talked about a relationship with Figure AI, with KUKA, with Booster, with Vingroup in Vietnam, with NEURA, there's a lot of different companies and then several customers, potential customers in China as well. There are several companies that are using our solution to build robotics at scale. And I think it's going to take a lot of different form factors.

You're going to have something that is manufacturing robotics, you're going to have something that is toy-like robotics. There's going to be a distribution center robotic solution. And then finally, humanoid. and each one kind of stresses on different things and requires incredibly capable chipset solutions. So very excited about it.

In aggregate, when do you see these opportunities starting to be material to Qualcomm?

I would say we expect this in kind of the next 3- to 5-year time frame, you're going to start with simpler robotic solutions. So think about home cleaning robots where we are already a very significant presence, they're going to become a lot more capable in the next couple of years. So that's one place to start. I think manufacturing robots now transitioning from fixed form factor to AI-based manufacturing robots that transition is going to happen faster. Warehouse robotics is probably a 2- to 3-year cycle, and you're going to see significant things at scale. And then finally, humanoid after that. So lot of different steps, and I think we get to participate in all of them.

Okay. And you start with the way it starts, you start with the lower content opportunities and as the volume grows and you get to humanoids, you have a higher front-end opportunity.

Yes. I think the amount of complexity in robotics is incredible because you need a solution for the locomotion, the movement. You need a solution for the brain, which is mostly AI-based. And then you need now, I think, distributed computing within the robot with chipset solutions in each limb, in arms and legs as well. So there's a lot of silicon content that will be required. And obviously, a great opportunity for us.

Is there any way to think about the content eventually on a humanoid robot like I remember the Automotive Investor Day, whether you outlined the automotive opportunity long term is like multiple thousands per vehicle, where does robotics opportunities say, related to automotive?

I think we had a range of, if I remember correctly, $300 to $2,000, something like that. I think actually, using a range like that for robotics is a reasonable starting point. I think it's going to evolve a lot might change. But using a very similar range to think of the range of robotics is probably a reasonable way of starting.

Okay. Great. So last 10 minutes, let's move to smartphones. What's your vision of what happens to smartphones with the agentic or what are agentic smartphones like...

Yes. I think in some ways a very interesting time to be in the kind of consumer device industry. If we just kind of quickly go through the history, you have the PC, the phone came in and the PC remained in its place and the phone drove a massive expansion of the SAM. There's this tremendous move towards a personal AI device. So what we used to think of as an XR device where you have virtual reality, augmented reality device that has evolved into a very different form factor now. I think -- there's a lot of companies who think of a device that is a personal AI device that can see what you can see, hear what you can hear and you can have a conversation with it. So we're seeing companies build glasses, companies build watches, earbuds with cameras, necklaces, broaches, just different form factors that people are experimenting with, but the core silicon content and the core function remains the same, it's this personal device that sits with you all the time and learns and understands about you and can be an assistant for you.

And so we see this trend as an incremental device that is on top of what we have in phones. And don't know yet which device form factor it will take in the end. We don't know if this is going to be a 50 million unit market, 200 million unit market or a phone size market. But any which way you look at it, it's a great expansion of our SAM. So that's one way of looking at the personal device market is an expansion of SAM through a different form factor.

Specifically to your question on smartphones. We're also seeing very interesting changes happening to smartphones. If you've seen a couple of the devices that have been launched in China, the idea is the interaction of the user changes from the classic app format to an agent format. And you can ask the agent to do an action, you could ask the agent to download an app or make a reservation, and the agent takes care of it from there.

And so you take this movement towards clause, combine it that -- combine that with agents and I think you put the 2 together and now you have a unique way of interacting with your device. I believe Google IO is happening today, and there's some of those discussions that outlines their vision on how the world will evolve. So I think super exciting time to be in smartphones in some ways from a technology perspective. This is not -- now putting those 2 things together, the personal device trend and the smartphone trend, this is not just something that the 5, 6 smartphone OEMs are working on, right? We have hyperscalers across U.S., across China, across other parts of the world, thinking about how to build this personal AI device.

And so we are very excited about what's coming up over the next several months. I think you're going to see a very large number of devices launched in the second half of the year. And it's -- each device is going to bring in a new kind of use case with AI. And I think it becomes a lot more clear about how different use cases are going to get implemented on devices with AI and then how the split works between the cloud and the device in each of those use cases.

Any more color on your engagement with either hyperscalers or Frontier Labs and helping them develop these devices?

I'd almost say pretty much -- everyone is working with us. We are in a very, very large percentage of those devices. The form factor of these devices are pretty challenging because if you are, say, building a chip for glasses, it needs to be very low power. It needs to have wireless connectivity. It needs to be very small. These are things that Qualcomm does well. And so I think we're very optimistic. We have a strong multiyear road map multi-design engagement with these hyperscalers. And I think just the beginning of a new category of devices, that's going to be super exciting.

From a content opportunity standpoint, in the past, you've talked about sort of this mid-teens content opportunity with some of your Android OEM partners, do you see the move to agentic AI driving a chipset capability that drives an inflection in that content increase?

Yes. So there's different architectures that people are experimenting with, right? There is a concept of having an AI co-processor in a phone that can allow agentic AI experiences that are persistent, that are always on at very, very low power. So if the architecture changes to something like that, that's a very, very significant increase in the content opportunity for us. There is also all these new devices that have different vectors of what it requires in terms of performance. And so there would be an expansion in some ways on a per user basis for us as well.

So a lot of conversations about how best to address this problem, but I think it follows the use cases. And when you think about a persistent AI use case where someone has to -- something has to stay on all the time, it will require a different architecture than what we have today, and that becomes an opportunity for us.

Got it. So maybe let's move away from the long term to the near-term dynamics. You did highlight confidence about getting to trough levels in F 3Q rate of the smartphone market and seeing growth beyond that with your Android OEM partners, and I think you specifically called out the China market or Chinese OEM handset OEMs. What is driving the confidence that there's no further inventory adjustment from their end?

Well, so there's 2 factors that can get -- that impact the revenue in the short term, right? First is just what is happening with the handset market given the memory industry dynamics. Second is the OEM decision to draw down on channel inventory. And so we have seen 2 quarters of drawdown, and we have seen this in history. Typically, the second quarter is the largest portion of the drawdown. We also have data on the channel inventory at this point. And so you -- at some point, you just get to a place where the channel inventory is too low and you cannot keep drawing down on it anymore. And so then our sales would reconcile to the size of the handset market. And that's what's reflected in our confidence. We also are going to have new devices launched in the September quarter, going into the December quarter, and that also helps us from a revenue perspective. So put those factors together, we feel pretty confident that June will be the bottom and we grow from there.

And what about the other consumer devices. So for example, when you look at PCs or consumer IoT or even maybe switching beyond that industrial IoT? Are they dealing with the same dynamic on the memory availability?

I mean the PC industry dynamics are well documented. I think the tablet industry is very similar to handsets as well, a lot of the same components go into those devices like handsets. But I think you go up from here for those devices as well as the OEMs have drained as much of the channel inventory as they could. And then now you reconcile closer to the size of the market.

Okay. Okay. Good. So before we wrap up, maybe a couple of questions, how are you thinking about as you put all these opportunities together, data center, you have autos already going well and scaling, you're doing well in PCs as well relative to your margin targets to keep 30% EBT margin for QCT. How do you feel about that? And whether we, in the long term, are thinking about something higher than that just given the amount of opportunities we are pursuing?

I mean, at this point, obviously, we are not changing our target. We will address it at Investor Day. A lot of our op margin targets are tied to the revenue scale in the business, right? And so when you think about data center and if we significantly scale revenue in data center, that should help us on op margin. That's just kind of a logical conclusion. And that will be the primary dynamic on upside opportunities to our target.

The other thing to think about is as we enter into data center, we are investing incrementally to build our product road map. And so that is -- and it's already reflected largely in our OpEx run rate, but that does have an impact in the short term on our operating margin. But I think as you look forward and the revenue that we're expecting now in the data center area, we feel very confident that the 30% is a reasonable target to have.

Any gross margin implications as you build out?

I mean the way I think about gross margin is that when we enter new industries, our gross margin in those industries reflect what other players in the industry have. Of course, if you're the new player, initially, you have to take a lower gross margin to get in. But in the end, it will reflect the set of markets we are in. As we get into data center, I think one of the most important things for us is to look for opportunities where it's -- their operating margin accretive, independent of the gross margin impact.

Okay. And final one, how are you thinking about capital allocation, including any potential M&A or further M&A that you need to fill in gaps in the portfolio to address the data center opportunity?

I mean, our M&A strategy has been, I think, very consistent, very successful. We have bought companies that have allowed us to accelerate our organic diversification plan. Example is Alphawave now latest one. Before that, it was Nuvia, we bought an ADAS stack company before that. And each one of these were extremely important for us in accelerating our diversification strategy. And we'll probably stick to the strategy we've had for M&A.

Of course, we've looked at the larger transactions. And so far, we've chosen not to pursue those. Never say never, but the focus is not on those things. The focus is on executing on transactions, the strategy that we've outlined.

Okay. No, that is great. I'll wrap it up there. Thank you, everyone, for coming to the conference. Thank you, Akash, for coming.

Thank you. Thank you.

Ladies and gentlemen, thank you for standing by. Welcome to the Qualcomm Second Quarter Fiscal 2026 Earnings Conference Call.

[Operator Instructions]

As a reminder, this conference is being recorded, April 29, 2026. Playback number for today's call is (877) 660-6853. International callers, please dial (201) 612-7415. Playback reservation number is 13759551. I would now like to turn the call over to Brett Simpson, Senior Vice President of Investor Relations. Mr. Simpson, please go ahead.

Thank you, and good afternoon, everyone. Today's call will include prepared remarks by Cristiano Amon and Akash Palkhiwala. In addition, Alex Rogers will join the question-and-answer session. You can access our earnings release and a slide presentation that accompany this call on our Investor Relations website. In addition, this call is being webcast on qualcomm.com, and a replay will be available on our website later today.

During the call today, we will use non-GAAP financial measures as defined in Regulation G, and you can find the related reconciliations to GAAP on our website. We will also make forward-looking statements, including projections and estimates of future events, business or industry trends or business or financial results.

Actual events or results could differ materially from those projected in our forward-looking statements. Please refer to our SEC filings, including our most recent 10-Q, which contain important factors that could cause actual results to differ materially from the forward-looking statements. And now to comments from Qualcomm's President and Chief Executive Officer, Cristiano Amon.

Thank you, Brett, and good afternoon, everyone. Thanks for joining us today. In fiscal Q2, we delivered revenues of $10.6 billion and non-GAAP earnings per share of $2.65, with EPS coming in at the high end of our guidance. QCT revenues were $9.1 billion, with another quarter of record automotive revenues as well as growth in IoT. Licensing business revenues were $1.4 billion. Before I share key highlights from the business, I would like to provide some perspective on Qualcomm's current customer design cycles and the opportunities ahead.

We are in a period of profound change, and it may not yet seem obvious to the financial community. The emergence of agentic AI workload with [indiscernible] as an early example are fundamentally changing user experiences across connected edge devices and reshaping our roadmap in every platform we develop.

For agents to work efficiently, they must run continuously in the background, fuel sensor data into context, orchestrate multistep tasks reliably and deliver strong security. Today's installed base of devices were not built for this new capabilities, and it represents a significant upgrade opportunity and expansion of our addressable market in the coming years. Agent orchestration is predominantly CPU bound and Qualcomm has the world's best-performing CPU across smartphones, PCs, auto and soon the data center. Qualcomm's unparalleled connectivity solutions empower efficient NPU for local models will also be key assets to delivering agentic AI experiences.

No other semiconductor company matches the breadth and scale of our technology and product portfolio, which powers devices spanning milliwatts to kilowatts from smart wearables to data centers. As a result, we're seeing a step function increase in strategic customer engagement and is changing how we think about the broad AI opportunity as well as the speed of our diversification efforts.

Beginning with automotive, in Q2, we exceeded $5 billion in annualized revenues for the first time, and we expect to exit fiscal '26 at a run rate above $6 billion. This growth is driven by our fourth generation Snapdragon Digital Chassis platform, which comprises connectivity, telematics, infotainment as well as advanced driver assistance and automated driving. Notably, we have now enabled more than 1 million cars operating ADAS and autonomy on our Snapdragon Ride processors.

By the end of the fiscal year, we will begin commercial shipments of our fifth-generation Snapdragon digital chassis platform. This represents the largest generation-to-generation content increase in Qualcomm's history, delivering 3x higher CPU throughput, a threefold increase in GPU capability and 12x higher NPU performance while supporting in-vehicle agents and processing for Level 3 and Level 4 autonomous driving.

Looking ahead to fiscal '27, we expect continued share gains and increased content, particularly in ADAS. We're pleased with the performance of our automated driving stack with BMW, and we're seeing broad customer engagement from other leading automakers. Our recent announcement with Bosch and Wave are good examples of what's to come as we build on our proven platforms and self-driving stack and scale ADAS. In IoT, agentic workloads and edge AI are driving major product renewal design cycles. Overall, our pipeline is healthy, and there is clear momentum for Qualcomm solutions.

In personal AI, we expect a significant increase in the choice of new smart glasses starting in the second half of the year. We believe these launches, combined with the rapid progress in agentic AI will catalyze an inflection point in customer demand across this category. Our 2026 Snapdragon X2 PC platforms are currently in production and our world-class Orion CPU unlocks powerful always on agentic experiences, making it a true competitive differentiator. Agentic orchestrators such as Open Claw, Claude desktop, Claude Code, OpenAI Codex desktop, Perplexity Computer, Crew AI, Armes agent, Landgraf and Humane 1 running on Snapdragon X2 are early proof points.

A recent PC MAG review of the ASUS ZenBook A16, notes the Qualcomm is now a serious challenger in the PC space and states, "the generational leap from the original Snapdragon X Elite to the X2 series is particularly striking. Qualcomm hasn't just caught up to the industry. In some cases, is now helping to set the pace."

In addition, our [indiscernible] NPU is the world's fastest for laptops delivering up to 85 tops together with our industry-leading CPU, which has the best on-device token generation rate. Snapdragon X2 delivers the full agent experience end-to-end and outperforms Intel's [indiscernible] Lake by nearly 30%.

In physical and industrial AI. Our Dragonwing IQ10 platform has generated substantial customer interest since our launch at CES. This is a significant upgrade compared to IQ9, feature an NPU with up to 700 [indiscernible] of on-device AI performance an 18 core Orion CPU over 20 camera sensors in an integrated safety [indiscernible].

Building on our design win with Figure AI, we announced an exciting multiyear agreement with Nora reinforcing our confidence that we can become a significant player in the broad robotics market. Also during the quarter, we introduced [indiscernible] at Embedded World. This is the second Arduino platform built on Qualcomm silicon and we view it as a world-class prototyping engine for both robotics and industrial AI developers as we expand our ecosystem across key verticals.

[indiscernible] is purpose built to bring AI into the physical world, enabling fully autonomous AI agents in a wide range of Edge AI applications, including voice assistance and vision systems. Several new industrial AI products are also moving from design win to deployment across retail, utilities, oil and gas, agriculture and other verticals. In data center, the Alphawave integration is off to a great start, and we're pursuing multiple opportunities with large hyperscalers, cloud service providers, sovereign AI projects and other global partners. Building on that momentum, we're also entering the custom silicon space beginning our ramp with a leading hyperscaler and we expect initial shipments in the December quarter.

In addition, development of our leading data center CPU and high-performance AI inference accelerators is progressing well. We look forward to sharing more details and customer wins at Investor Day in June. Regarding handsets, I would like to underscore 2 key points: First, the quarter played out as we expected. Sell-through held up in our chip business materially undershipped consumer demand. We believe our China Android revenue is bottoming out in fiscal Q3, and Akash will provide more specifics in his financial update.

Second, we think agentic smartphone will soon begin to influence the premium tier and we expect this [ theme ] will only get stronger into fiscal '27 with examples like the [indiscernible] powered, agentic AI phone from ZTE Nubia. Xiaomi's recent announcement of [indiscernible] agent framework and other agentic [indiscernible] systems now in development across the Android ecosystem, we have a clear line of sight into how the AI upgrade cycle will unfold, and this is going to be an important tailwind for premium demand over time.

Next, I want to highlight a major strategic initiative and long-term growth driver for Qualcomm, 6G, the next generation of wireless. Design for the age of AI will believe 6G will present one of the most significant transitions for the wireless industry. From a connectivity perspective, 6G will enable new classes of mobile and personal devices such as smart glasses with enhanced uplink capabilities to support agentic use cases like see what I see.

Beyond connectivity, 6G will be an AI-native network where AI reasoning, learning and autonomous action are core functions. It is intended to act as distributed intelligent infrastructure that integrates communication in wide area real-time sensing. With these new capabilities, the network becomes critical infrastructure and provides the telecom industry an opportunity to develop completely new business and economic models.

He will make possible new AI-enabled services, ranging from context relevant data data insights and analytics low altitude like aerial, terrestrial and autonomous traffic management, drone detection and tracking and 3D mapping with telemetry to build dynamic digital wins at scale. QUALCOMM's leadership in connectivity, AI processing and high-performance, low-power computing position us to be one of the key architects and beneficiaries of the 6G transition.

In addition to the development of foundational technologies and standards, we're building end-to-end solutions for devices and the network from a genetic modems and compute platforms that power phones, PC, intelligent wearables and cars all the way to the network, including power-efficient next-generation radio units, wide area network sensing platforms and high-performance compute and AI accelerators for the RAN network edge, core and data center.

To help shape and accelerate the 6G road map at MWC, we launched a 60 company coalition spinning carriers, cloud infrastructure, AI native partners and auto OEMs. The engagement and feedback on our 6G vision and plans from our partners, customers and governments across the globe has been very positive and we look forward to working across the industry to deliver on this generational opportunity.

Before I turn the call over to Akash, I want to note that we will provide a broader update at our Investor Day to include our data center plans and our progress in other areas, including advanced robotics, next-generation ADAS, industrial edge AI, personal AI devices in 6G. We hope you can join us as we will be highlighting meaningful new avenues of growth to support our long-term diversification story. I will now turn the call to Akash.

Thank you, Cristiano, and good afternoon, everyone. Let me begin with our results for the second fiscal quarter. We delivered revenues of $10.6 billion and non-GAAP EPS of $2.65, with EPS at the high end of our guidance. QTL revenues of $1.4 billion and EBT margin of 72% came in at the high end of our guidance, driven by favorable mix with global handset units approximately flat on a year-over-year basis. QCT revenues of $9.1 billion and EBT margin of 27% were in line with our expectations. QCT handset revenues of $6 billion came in as anticipated as OEMs remain cautious on handset bills due to the impact of challenging memory industry dynamics.

QCT IoT revenues of $1.7 billion were up 9% on a year-over-year basis, driven by growth across consumer and industrial products. In QCT Automotive, we delivered another record quarter with revenues of $1.3 billion, representing 38% year-over-year growth driven by accelerating demand and increasing content per vehicle due to the transition of new digital cockpit and ADAS launches to our fourth-generation chipsets.

On a combined basis, QCT automotive and IoT revenues grew 20% year-over-year, underscoring the continued diversification of our business, consistent with our long-term revenue targets. We also returned $3.7 billion to stockholders during the quarter, including $2.8 billion in share repurchases and $945 million in dividends, reflecting acceleration of our capital return program.

Lastly, we released a previously recorded tax valuation allowance, resulting in a $5.7 million noncash GAAP tax benefit in the second fiscal quarter. This benefit is excluded from non-GAAP results. This reversal reflects new guidance on corporate alternative minimum tax issued in February by Treasury and IRS permitting taxpayers to deduct previously capitalized domestic R&D expenses.

Before turning to guidance, I'd like to provide an update on the continued impact of memory industry dynamics on our business. Last quarter, we highlighted that the increasing demand for memory and AI data centers was driving uncertainty in memory supply and price increases to handset OEMs. And as a result, the handset OEMs, particularly in China, were taking a cautious approach by reducing build plans and drawing down channel inventory. These dynamics played out as expected in the second fiscal quarter and are also reflected in our third quarter guidance. As a result, in both quarters, our China QCT Android shipments are meaningfully below the scale of end consumer handset demand. We now estimate that QCT handset revenues from Chinese customers will reach a bottom in the third quarter and return to sequential growth in the following quarter.

Now turning to guidance. In the third fiscal quarter, we are forecasting revenues of $9.2 billion to $10 billion and non-GAAP EPS of $2.10 to $2.30. In QTL, we estimate revenues of $1.15 billion to $1.35 billion and EBT margins of 67% to 71% with sequential decline primarily due to the operating assumption of weaker low-tier handset units. In QCT, we expect revenues of $7.9 billion to $8.5 billion and EBT margins of 25% to 27%. We are forecasting QCT handset revenues to be approximately $4.9 billion as a result of the impact of the industry-wide memory dynamics I just outlined.

We anticipate QCT IoT revenues to grow by high single digits versus the year ago period, driven by industrial and consumer products. In QCT Automotive, following another record quarter, we expect year-over-year revenue growth to further accelerate to approximately 50% in the third fiscal quarter. Lastly, we forecast non-GAAP operating expenses to be approximately $2.6 billion in the quarter.

In closing, while our near-term revenues are impacted by memory industry cyclical dynamics, we're confident in the underlying fundamentals around Snapdragon product leadership and content growth opportunities, including the adoption of agentic AI technologies. We continue to execute on our secular growth opportunities in automotive and IoT and remain confident in achieving our long-term revenue targets. In addition, we are very excited about the progress in our data center products and customer traction. We now expect initial shipments for our custom silicon engagement at a leading hyperscaler later this calendar year.

We look forward to providing an update on our growth initiatives, including opportunities in data center and physical AI at our Investor Day on June 24. This concludes our prepared remarks. Back to you, Brett.

Thank you, Akash. Operator, we are now ready for questions.

[Operator Instructions]

Our first question comes from the line of Joshua Buchalter with TD Cowen.

Obviously, I'm not sure you're going to be able to front-run the AI day you plan to host in June. But any details you're able to share or context on what the custom silicon engagement, what the scope is, what the magnitude is? Is this a CPU? Is it an accelerator? Is it a networking chip? Just any help you can give us beyond the press release and prepared remarks, I think would be helpful as that's where investors certainly want to dig in today.

Josh, this is Cristiano. Thank for the question. Look, I can't provide a lot of details, as you said, we don't want to front run, I think, June 24. But here's a couple of things I can tell you, which is alongside what we said in the script. I think we have spent the time, I think, building assets and we've been building our CPU. We have accelerator. We have a different solution for memory in the accelerator. We have added a lot of capabilities for custom ASIC with the acquisition of Alphawave and Connectivity, we have been pursuing customer ASIC. We talk about have an engagement with a number of companies and pleased with the engagement several quarters ago. And I think given the capabilities that we're developing and what's happening in the market, that's accelerating. So we're very excited. The only thing I can tell, it is a large hyperscaler and we're really thinking about a multi-generation engagement. But I think that's what we can say at this point.

Okay. I guess we'll stay tuned. For my follow-up, can you maybe walk through why you're confident that -- I assume it's fiscal third quarter that you were referring to with the third quarter, but why you're confident fiscal third quarter can be the bottom for Android sales -- Android QCT sales into China. I mean that's typically -- September quarter is usually a typically down seasonal quarter. And just given how low visibility is right now overall in the handset market, I'd be curious just what inputs you're seeing that gives you the confidence it's going to bottom in the June quarter.

Sure, Josh. It's Akash. So if you think about the impact to from Chinese handset OEMs as a result of the memory dynamics. It's really 2 parts. The first part is the scale of the handset market. And there, we have seen some small decline in it, especially in the mid, low tiers. But by far, the larger impact have been the OEMs making a decision to slow down their bills and draw down on channel inventory. So both of these factors have been in our March quarter results and they're also represented in our June quarter guidance. And so what we end up doing in both quarters is really significantly under-shipping the end consumer demand for handsets as a result of the channel inventory drawdown factor.

So as we look forward, we feel confident that the third quarter is now the bottom. And so as we go forward, the revenue is going to be much closer to the scale of the handset business versus the inventory drawdown factor continuing into our forecast.

And Josh, this is Cristiano. I just want to add one thing because -- there's another way to look into this. As you know, because of our licensing business, we do have visibility of what happens in the market. So we know [indiscernible] true. We know how the [indiscernible] market is behaving even with increased price on the handsets. So it gave us a real good idea on activations and customer demand versus what we're shipping. So that dynamic outlined by Akash, is kind of very clear to us that Q3 becomes the bottom.

Our next question comes from the line of Samik Chatterjee with JPMorgan.

Cristiano, maybe just going back to the data center opportunity and just trying to think about if you can sort of help us think about the competitive landscape here. I mean you've had, which is the IP provider now announced they want to vertically integrate and make chips. You had NVIDIA announce that they are going to focus on the [ inferencing ] market as well. How are you thinking about the competitive dynamics where they are related to maybe 3 months ago or 6 months ago that you're now sort of going and trying to deliver these wins? And I have a quick follow-up after that.

Great question. Look, I'll give our perspective. And I think we have now, I think, more clarity than we ever have about where kind of we are in the AI space. So a little bit of maybe at a very high level, right? In the beginning, it was all about training. It was all about creation of AI, a lot of GPU, very GPU-centric deployment. Infra started gain scale and then the conversation changed to -- I'm going to use my GPU from training on the cluster that I build and when I'm not training, I'm going to use that for inference.

As inference starts to gain scale, we started to see dedicated solutions. The data center becomes more disaggregated. You have separate computing solution, some for compute bounds, some from memory bound. And now we're entering the, I will say, the next phase, which how AI is really going from infant generating tokens how do you generate demand for tokens, which all those agentic experience and those orchestrators, they run into a lot of the devices.

So when you think at that landscape and you look at our IP, in the places that we could be very differentiated, I will start by our CPU. I think when you think about agents, CPU becomes very important. And I will argue, we were one of the companies that have a pretty good CPU asset. We've proven that CPU performance with leading performance on the markets that we are right now, such as PC, smartphone and Auto.

And we have built and we'll provide details on Investor Day a dedicated CPU for agentic experiences in the data center. We're going to show the metrics, we're going to show how it performs. People will be able to compare. As you know, we have an architecture license, and we have a very, very high-performance CPU. So that's one of the assets.

The other asset is how you think about the scale of a semiconductor company like Qualcomm. We're not small. And the ability to combine the IP with the ability to do custom silicon, make sure that, that yields make sure it's delivered with quality and combine a lot of the connectivity IP, which I believe Alphawave because it was a licensing IP company has a leading IP, the more you license, the better your IP becomes.

The number three is how we think about the accelerator. You're going to need high compute density, low TCO. And we think that we have something unique, which is focused on a cluster that is disaggregating very specific function especially like [indiscernible].

I think the activity you've seen with companies like Grok and [indiscernible] just prove that you have opportunity for a dedicated inference accelerator. And the last point is, I would not discount the position that we have on the edge. If you actually track what's happening with Open Claw in all of the different desktop and co-work solutions, you rely a lot on a high-performance CPU device, which is also causing an upgrade cycle for us.

So we look at this whole landscape, and that's how we feel so good about the agentic transition of AI, what it means to Qualcomm. And hopefully, on June 24, we'll show the details on the road map and an investor will be able to see where we stand, and please reserve a seat.

Yes. No, please look forward to talking about that more at the Investor Day. Maybe for my follow-up, just going back to the handset business. Can you just remind us of the multiyear agreement framework that you have with your primary premium smartphone customer, Samsung. You did have some sort of changes in the market -- in the share with them this year. I think there are some more indications for the step-down in share and more use of the in-house SoC sort of next year? Or can you just remind us sort of how you're thinking about that engagement long term? And what does the multiyear agreement sort of capture at this point?

No, absolutely. I love answering the question. So this is a very, very stable, I think, a relationship with Qualcomm. I want to remind you all that we have reset the framework of this relationship. Historically, we always had a business with [indiscernible] that was in the 50% share between us and their own in-house silicon. That has changed to greater than 70%, as you know, and that has been the framework. And sometimes we get more than that, but we plan our business in greater than 70% share, which is exactly what we have said. You should expect that that is the framework of this year, and that is also the framework for next year. I would say that, that's probably one of the most stable relationship that we have, and we have visibility of what that entails. And we feel good about the position of Snapdragon. And I'll argue, I think given what's happened with agents, we have an opportunity to actually have a positive bias on that share.

Our next question is from the line of Chris Caso with Wolfe Research.

I guess the first question is just returning to the data center briefly. And to clarify what you mentioned in terms of the hyperscale engagement for the December quarter. Is that an engagement for an accelerator or a CPU? I understand your targeting both, it sounds like, but what's the particular engagement for December?

This particular engagement, which we're going to have shipments in December is a custom product. We're working with a hyperscaler.

Okay. So no other specificity past that okay. Just with regard to QTL, and it looks like that's modestly down and likely due to what you've been talking about with regard to what's going on in the handset market. What's the right way to think about the QTL business as we go forward into the second half of the year? Do you think that we kind of maintain these levels and adjust for seasonality as you get to the end of the year? Or do you expect the impact on QTL to be more significant as you go in the second half?

Yes, Chris, it's Akash. So as you saw in our results for the second quarter, year-over-year handset units were flat for the global units. And this was really kind of impacting our guidance as well, our actuals as well. As we look at third quarter, what we're guiding is some weakness in the mid, low tiers in the market. I mean this is obviously something that we are projecting forward, and we're going to track closely. But what we're seeing is the premium high tier of the market is continuing to hold and weakness in the lower tiers. And that's what's reflected in our guidance. That's a reasonable way of thinking about the market going forward as well.

The next question is from the line of Stacy Rasgon with Bernstein Research.

So if the China handsets bottom in Q3 and then they grow in Q4, that September quarter, September quarter, I think, is when we're supposed to get the Apple step down, which may be an offset. So I guess just how are you thinking about handset seasonality in the September quarter, given those kind of competing dynamics? How should we be thinking about that?

Yes. Stacy, it's Akash. You're right. I think in terms of kind of handset revenues for QCT from Chinese OEMs, we do expect that the June quarter is the bottom, and you will see sequential growth from there. And Apple, you're right as well that typically, it's a growth quarter for Apple product revenue, and we do not see that at this point given the share assumption change. So those are the 2 factors you would use to forecast September quarter.

I mean do you think handsets growing sequentially in September or not given those two factors?

Stacy, we are not specifically guiding it at that point -- at this point, but I think those 2 factors would be the input into the forecast.

Okay. And for my follow-up, you talked about like agentic devices and agentic smartphones like driving a shift and things, I guess, into '27. Do you think the memory issues are going to be done by -- like how much memory does an agentic smartphone need? And is that something that's going to continue to be a headwind do you think on this as we go into 2027? How do we think about the broader dynamics around memory as we go forward?

Thank you for the question, Stacy. Look, it's a little early to talk about '27. I think 1 thing to see is I think the pace of change of AI is getting scale. When I think about the framework that I talked about it before, which you go from inference to now, you know how you generate demand for tokens with a lot of agents. And I think what we see is 2 things. One is the devices are changing the requirements in the design and the players. We see interesting associations now starting to form between smartphones and AI companies. We're starting to see some very interesting dynamics there which is changing the nature of designs. We see designs moving towards products they have much more capable CPU to run those type of products. And there's a lot of noise in the memory environment right now. I wish I could make a prediction on '27 is a little early, but we see a combination of some of the same companies that want a lot of demand for data centers and also getting involved with some of the devices at the edge as well. And we see new memory players coming and building capacity. So we're going to have to monitor the situation and see what happens in '27.

The next question is from the line of Timothy Arcuri with UBS.

Thanks a lot. I wanted to ask also about this custom that is going to ship in the fourth quarter of this year. I know you brought a team in from Alphawave. It seems a little fast to get something to market that includes your IP by the end of this year, given cycle time. So I think they had some chiplet stuff and maybe some custom DSP stuff. Is that the sort of thing you're talking about? Or is this truly something that includes a big portion of your IP that you've been able to turn around since the deal closed?

Look, I think 2 answers. So we have -- I'm pretty positive for the past several quarters, we've been talking about engaging with customers in the data center. So I think when we start engaging and talking about some of the Qualcomm capabilities, it's probably, I think, even before the acquisition of Alphawave. I think the acquisition of Alphawave increase our execution capabilities in the portfolio of IP. I think you should expect that we're going to have a longer multi-generation agreement with those companies that brings a lot of Qualcomm capabilities to the table. I wish I could provide more details, but like I said, I don't want to front run what we're going to do on June 24. But I think we will provide a detail of everything we're doing. Our customers win in our roadmap and our IP.

And then I guess just as a follow-up, is the assumption still the same that like around your share in Apple. I know there are some signs that there's going to be a little more aggressive displacement. Is the assumption still that it's going to be 20% for the new launch?

Yes. No change, Tim, to our assumption there. We've set of the 20% of the -- 20% share of the phones that will launch in fall this year and no product relationship beyond that. And this is -- this assumption has been consistent for the last couple of years. In terms of Apple product revenue for fiscal '27, we've seen sell-side models in the range of a little over $2 billion in terms of QCT product revenue in the year, and we think that's a reasonable place to model the business.

Next question is from the line of Joe Moore with Morgan Stanley.

You alluded to the weakness being more in the medium tier and the premium tier as being stronger. Is that something you can see where you're sort of taking limited memory allocation and that sort of drives just to sort of limit that puts it in the heart here. Just what are you seeing in terms of what that's doing to your mix going forward?

Yes. I think the actions from the OEMs are obviously very logical. If you had to choose between which devices you put your memory allocation to, you would pick the premium and the high tier, that's where the profitability sits, and that's what you're seeing happen in the market.

Okay. And you talked about 6G in 2029. I mean is that -- what does that time frame represent? Is that sort of introduction of technology? Is there a shipments then? Just anything you can tell us about when 6G starts to become relevant?

Yes. And look, thank you for the question. The reason I brought it up is because 6G is going to feel, I think, very different than the other Gs for Qualcomm. I also believe that 6G creates some very interesting, I think sovereign AI and data center opportunities, I think, for Qualcomm as well. You should be thinking about our time line, and we've been consistent with it. We will have prototype base, I think, demonstrations in 2028. Likely, we're going to have first silicon in '28, and we want early launches in 2029, and then we expect that to get scale by 2030.

Our next question is from the line of Ross Seymore with Deutsche Bank.

I want to go back to the handset side. And if you could just level set us to whether it's the fiscal second quarter or fiscal third quarter, what percentage of handsets is China. And you mentioned that a lot of the dynamics is how far you're under-shipping versus true demand, are you believing that true demand, whether it's China or elsewhere, is truly weakening still? Or is that side of the equation stabilizing despite the memory fairs?

Yes, Ross, when you think about the total handset market, and this is more of a QTL comment, right, that's where we are seeing that the -- there's a slight decline in mid-low tiers, but the overall scale of the handset market has not changed much, at least in the March quarter, and we're going to obviously closely monitor that going forward.

In terms of QCT shipments to Chinese customers, it's a factor, as I said earlier, of 2 things. It's not just the scale of the handset market, but the OEM decision to drawdown on channel inventory. And so my comments earlier were about the drawdown of channel inventory will end soon, and that's us calling the bottom on the quarter. And then really, our shipments will reconcile to the size of the handset market.

Okay. And I guess for my follow-up question, shifting gears to the automotive side of things. You mentioned about the ADAS side starting to ramp and growing 50% this year, so -- or at least in the next quarter, so doing very, very well. As you go from more of a cockpit business to the ADAS mix increasing, how does that change the revenue trajectory and perhaps the gross margin trajectory in your automotive business?

Yes. So this is Cristiano. I think what you see is it accelerates revenue dramatically because it's a lot more silicon content. If you -- and that is true actually on both sides. I think what you saw is when we went from generation 3 to generation 4 in digital cockpit. We -- I think we keep mentioning the car, it's really becoming a computing surface. We saw a step function increase in the capability [ silicon content ]. You expect another 1 when we go from fourth generation to fifth generation. And as we add processors and you started to see more and more development of L2++ in direction towards Level 3, you're starting to see the amount of computing power going up. So for us, it's basically a significant revenue accelerator within automotive.

And specifically on your question on gross margin, I'd highlight maybe 2 additional factors to what Cristiano said. I think we have the we're transitioning from a chip sale to a [indiscernible] sale. And so as we go to a module, it increases the revenue opportunity for us as well. And then in addition, we have software opportunity on top of the the chipset, which also helps our margin profile. So net debt, we still model the business in line with our corporate average, but it really is a business that has several vectors of growth as both of us outlined.

Our last question is from the line of Vivek Arya with Bank of America Securities.

For the first one, Akash, you mentioned Apple product sales, I think, $2 billion plus for fiscal '27. What about the royalty contribution? How does that evolve as you approach the date for kind of renegotiation negotiating that business?

Yes. So pending the renegotiation, the royalty, we don't expect it to change, right? It should be in the same scale that it's at, and it's an independent business separate from the chip business.

And for my follow-up, Cristiano, back to the hyperscaler discussion, I realize you'll give more details at Analyst Day. But was -- is Qualcomm's intention to approach this from an ASIC perspective, I thought you plan to enter the data center from a merchant perspective. But are you saying that now the goal is to approach it from an ASIC perspective? And if that is the case, what impact does it have on margins? Like are you really going to compete head on with the other ASIC suppliers that are out there. So this is going to be more kind of a one-on-one right type approach to the market as opposed to approaching the market in a broader merchant. So just what is kind of the broader strategy, go-to-market strategy that Qualcomm has in this business?

Very good -- great question. I think the answer is all of the above. Look, we -- first of all, as a new entrant, I think we we're very flexible. But we also look at the reality of what's happening in the hyperscalers. You can see that the majority of the revenue for semiconductor companies is heavily concentrated in a few number of very large companies. And those companies have now had indicated very clearly, they have different -- as the data center gets disaggregated, you have different approach to compute to connectivity. And you should assume that Qualcomm will play on merchant, on custom and it's going to be a combination of how we're going to configure our IP and different IP blocks for different solutions is going to be a bespoke business.

And specifically on your question on this custom engagement we talked about, we do expect that to be accretive at the operating margin level.

Thank you. That concludes today's question-and-answer session. Mr. Amon, do you have anything further to a before adjourning the call?

I think the obvious thing I want to say is please ask everyone to attend our June 24 [indiscernible] Investor Day. We intend at the Investor Day to really highlight not only, I think, everything that is happening with the new Qualcomm, but also I think the details of the products and technology we have been developing for the data center space, provide an update in how physical AI is transforming our business and provide the clarity that we have today, how really agents and agentic experience exactly has a broad implication in our entire business. And I'm looking forward to speak to all of you, and I'd like to our partners, our employees for a great quarter as we continue to transform Qualcomm. Thank you very much.

Ladies and gentlemen, this concludes today's conference call. You may now disconnect.

Ladies and gentlemen, thank you for standing by. Welcome to the Qualcomm First Quarter Fiscal 2026 Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded, February 4, 2026. The playback number for today's call is (877) 660-6853. International callers, please dial (201) 612-7415. The playback reservation number is 13758127. I would now like to turn the call over to Mauricio Lopez-Hodoyan, Vice President of Investor Relations. Mr. Lopez-Hodoyan, please go ahead.

Thank you, and good afternoon, everyone. Today's call will include prepared remarks by Cristiano Amon and Akash Palkhiwala. In addition, Alex Rogers will join the question-and-answer session. You can access our earnings release and a slide presentation that accompany this call on our Investor Relations website. In addition, this call is being webcast on qualcomm.com, and a replay will be available on our website later today. During the call today, we will use non-GAAP financial measures as defined in Regulation G, and you can find the related reconciliations to GAAP on our website.

We will also make forward-looking statements, including projections and estimates of future events, business or industry trends or business or financial results. Actual events or results could differ materially from those projected in our forward-looking statements. Please refer to our SEC filings, including our most recent 10-K, which contain important factors that could cause actual results to differ materially from the forward-looking statements. And now the comments from Qualcomm's President and Chief Executive Officer, Cristiano Amon.

Thank you, Mauricio, and good afternoon, everyone. Thanks for joining us today. In fiscal Q1, we delivered record revenues of $12.3 billion and non-GAAP earnings per share of $3.50. Within QCT, record revenues of $10.6 billion were driven by strength in flagship handsets. We also saw another quarter of record revenues in automotive and positive momentum in IoT across industrial, EDGE networking applications and smart glasses.

Licensing business revenues were $1.6 billion. While global consumer demand for handsets, especially premium and high tier exceeded our expectations with healthy sell-through observed through fiscal Q1 in the first few weeks of 2026. In the coming quarters, the handset industry will be constrained by the availability and pricing of memory, particularly DRAM.

As memory suppliers redirect manufacturing capacity to HBM to meet AI data center demand, the resulting industry-wide memory shortage and price increases are likely to define the overall scale of the handset industry through the fiscal year. Given the current environment, several handset OEMs, especially in China, are taking a cautious approach in reducing their chipset inventory.

This is reflected in our guidance for the upcoming quarter. We will continue to work closely with our customers and suppliers as the situation evolves. Akash will share more details on the memory impact in his prepared remarks and now some key highlights from the business.

We are pleased with the continued expansion of the premium and high-tier smartphone segments and traction of Snapdragon platforms, including broad OEM adoption for dual flagship product strategy. For Samsung's upcoming family of premium tier devices, we expect approximately 75% share, consistent with prior expectations. It's important to note that during the quarter, ByteDance launched their first Agentic AI smartphone powered by the Snapdragon 8 Elite.

This is a significant milestone in the transition toward AI-native smartphones and the precursor to the Agentic experiences shaping the future of mobile. With the development of agents and AI becoming the new UI, intelligent wearables are evolving into personal AI companions and quickly emerging as the next mobile computing category.

Our early investments in this area, including powerful and power-efficient chipsets, advanced connectivity, including micropower WiFi as well as ambient sensing and perception technologies position Snapdragon XR, Wear, and Sound as the platforms of choice for the industry. We're pleased to be working with 7 of the 9 largest cloud companies globally and more than 40 personal AI devices are in production or development.

In PCs, we introduced the Snapdragon X2 Plus, an expansion of our second-generation platforms purpose-built for the enterprise and commercial segment. The X2 Plus is powered by the third-generation Qualcomm Oryon CPU, which delivers up to 35% faster single core performance and up to 3.5x faster multi-core performance compared to the competition in previous generations.

Our Hexagon NPU provides up to 5.7x and 3.4x faster inferencing versus competitors' NPU and GPU, respectively. 18 Snapdragon-powered PCs debuted at CES from ASUS, HP, Lenovo and Microsoft. The ASUS Zenbook A16 was one of the standouts, featuring the Snapdragon X2 Elite Extreme and is the fastest Snapdragon-powered laptop to date.

It features our 18 core third-generation Oryon CPU and 80 TOPS Hexagon NPU for AI workloads in an Adreno GPU, delivering up to a 2.3x improvement in performance per watt versus the prior generation. X2 Elite Extreme enables desktop class performance, advanced graphics and more than 21 hours of battery life in an ultralight 16-inch form factor. We remain on track to commercialize 150 Snapdragon X-powered PCs this year.

Demand for our Snapdragon Digital Chassis solutions remains incredibly strong, and we announced several collaborations with top automakers, OEMs and service providers during the quarter. We signed a letter of intent for a long-term supply agreement with Volkswagen Group, which spans many brands, including Audi and Porsche. Under this intended agreement, we would provide advanced infotainment and connectivity capabilities powered by our digital chassis across multiple vehicle segments, price tiers and markets. We would also serve as the group's primary technology provider for its software-defined vehicle architecture developed through its joint venture with Rivian Automotive.

In addition, we're collaborating with the group's Automated Driving Alliance formed by CARIAD and Bosch to accelerate development of highly automated driving systems. We're very proud that the newly launched RAV4, Toyota's top-selling vehicle globally and one of the best-selling cars worldwide is powered by our Snapdragon Cockpit Platform, delivering premium AI-enabled in-vehicle experiences.

We also announced new and expanded collaborations with Hyundai Mobis, Leapmotor, Li Auto, Zeekr, Great Wall Motor, NIO and Chery, bringing our total design wins for Snapdragon Elite platforms to 10 programs. In industrial IoT, we continue to expand our portfolio of advanced computing, connectivity and AI solutions for an increasing number of verticals.

With the recent acquisition of Augentix, we augmented our Dragonwing vision portfolio in Qualcomm Insight platform with its AI-based low-power image signal processing solution. At CES, we also introduced two new Dragonwing processors delivering on-device intelligence for security-focused drones, smart cameras and industrial vision, AI TVs, media hubs and video collaboration systems.

Additionally, the launch of our new Dragonwing IQ-X series marked our entry into the industrial PC space with best-in-class compute performance and efficient Edge AI engineered for PLCs, advanced HMIs, Edge controls and panel and box PCs. This quarter, we formally announced our expansion into advanced robotics and introduced a full suite of robotics technologies and solutions, including the Dragonwing IQ10 series.

Our general-purpose robotics architecture supports advanced perception and motion planning using models such as VLAs and VLMs, allowing robots to perceive, reason, adapt and act in real-world environments. As part of a complete hardware to software stack, IQ10 is designed to accelerate commercialization of household, industrial and humanoid robots. It combines heterogeneous Edge compute, safety-grade SoCs and end-to-end AI.

In a short period of time, we have engaged with Advantech, APLUX, AutoCore, Booster, Figure, Kuka Robotics, Robotec.AI and VinMotion to help define the compute architecture for their robotics and humanoid platforms. The physical AI in robotics space is experiencing rapid growth driven by advances in Edge AI and sensor fusion, and Qualcomm is one of the best positioned companies to enable this next frontier of AI.

We will do this by leveraging our strengths in high-performance, power-efficient computing, connectivity and Edge intelligence as well as our experience in ADAS and autonomy, industrial and safety-grade silicon and perception and sensing technologies. Many of the drivers of our leadership in automotive are applicable to advanced robotics.

Finally, we continue to develop our data center solutions and engage with leading hyperscalers, cloud service providers, sovereign AI projects and other global partners. We remain encouraged by the positive feedback on our CPU and innovative AI processing and memory architecture for next-generation inferencing data centers.

Additionally, the recent developments in the industry validate Qualcomm's view of the importance of specialized and power-efficient AI platforms as inferencing becomes the key driver of data center growth. In fiscal Q1, we completed the Alphawave Semi acquisition, adding high-speed wire connectivity technologies to further strengthen our platforms.

We also acquired Ventana Micro Systems, reinforcing our leadership and commitment to expanding the RISC-V standard and ecosystem and development of our high-performance RISC-V CPU for data center workloads. We look forward to providing more information, including an update on our road map at our next investor event. We'll also share our progress in robotics, automotive and next-generation autonomy, industrial IoT and 6G. I will now turn the call to Akash.

Thank you, Cristiano, and good afternoon, everyone. Let me begin with our strong first fiscal quarter results. Total revenues of $12.3 billion and non-GAAP EPS of $3.50 were both records with non-GAAP EPS coming in at the high end of our guidance. QTL revenues of $1.6 billion and EBT margin of 77% were at the high end of our guidance, driven by higher units and favorable mix. We delivered record revenues in QCT of $10.6 billion including strong year-over-year growth across automotive and IoT.

QCT handset revenues reached a record $7.8 billion, reflecting the benefit of recently launched flagship smartphones. QCT IoT revenues of $1.7 billion grew 9% year-over-year, driven by demand across consumer and networking products. In QCT Automotive, we delivered another record quarter with revenues growing to $1.1 billion, up 15% versus the year ago period on increased demand for our Snapdragon Digital Chassis platforms.

QCT EBT margin of 31% came in line with expectations, exceeding our long-term target of 30%. Lastly, we returned $3.6 billion to stockholders, including $2.6 billion in stock repurchases and $949 million in dividends. Before turning to guidance, I'd like to address the impact of the memory industry dynamics on our financial outlook. The fundamentals of our handset business remain favorable. With a stable global economic environment, total handset shipments exceeding expectations in the December quarter, especially in the premium and high tier and a strong design win pipeline for our Snapdragon chipsets.

However, increasing demand for memory solutions in AI data centers is driving near-term uncertainty in memory supply and pricing for handset OEMs. As a result, the handset OEMs are taking a cautious approach in planning their business. We've seen several OEMs, especially in China, take actions to reduce their handset build plans and channel inventory.

Our guidance for the upcoming quarter reflects the latest signals from these customers, which includes reduced chipset orders aligned with their scaled back expectations for build plans. We expect to return to our prior run rate and growth trajectory for QCT handset revenues when these conditions normalize.

Now turning to guidance. In the second fiscal quarter, we are forecasting revenues of $10.2 billion to $11 billion and non-GAAP EPS of $2.45 to $2.65. In QTL, we estimate revenues of $1.2 billion to $1.4 billion and EBT margins of 68% to 72%, reflecting normal sequential trend. In QCT, we expect revenues of $8.8 billion to $9.4 billion and EBT margins of 26% to 28%.

We are forecasting QCT handset revenues to be approximately $6 billion. As a result of the impact of memory constraints I just outlined. We anticipate QCT IoT revenues to grow by low teens percentage versus the year ago period, driven by growth across industrial and consumer products. In QCT Automotive, following another record quarter, we expect year-over-year revenue growth to accelerate to greater than 35% in the second fiscal quarter.

Lastly, we expect non-GAAP operating expenses to be approximately $2.6 billion in the quarter. The sequential increase is driven by typical calendar year resets for certain employee-related costs and completion of our acquisition of Alphawave to further strengthen our platforms for next-generation AI data centers. In closing, we are pleased with our strong first quarter performance, delivering record results across the following metrics: Total company revenue, non-GAAP EPS, QCT revenues, QCT handset revenues and QCT Automotive revenues.

While near-term QCT handset guidance is being impacted by memory industry dynamics, the underlying fundamentals around consumer demand for handsets and Snapdragon product leadership remains strong.

Our second quarter guidance reflects the continued revenue acceleration across automotive and IoT with their combined growth outpacing the run rate required to achieve our long-term revenue targets. Our product announcements and strong customer engagement at CES 2026 further demonstrated our momentum across multiple growth vectors. In automotive, we have reinforced our technology leadership with 10 design wins for Snapdragon Ride Elite and Cockpit Elite, 8 global programs for Snapdragon Ride Flex and continued success in building an automated driving stack ecosystem for our customers.

In robotics, we announced a full suite of technologies, including the industry-leading Dragonwing IQ10 chipset platform and engagement with several players in the ecosystem to drive commercialization of our products. In industrial, we showcased our ability to serve a wide spectrum of customers from global enterprises to local developers with an expanded portfolio that offers advanced Edge computing and AI solutions across industry verticals. This concludes our prepared remarks. Back to you, Mauricio.

Thank you, Akash. Operator, we are now ready for questions.

[Operator Instructions] First question comes from the line of Joshua Buchalter with TD Cowen.

I wanted to start with the handset outlook. Any other factors that are driving the weakness beyond the memory pricing? It was good to hear the reiterated Samsung share. But I think most importantly, how should we think about the TAM for the year? And do you feel like this inventory correction is sort of the last shoe to drop in the March quarter that you're seeing?

Thanks, Josh, for the question. I will start, and I'll ask Akash to add more color. It's 100% related to memory. Actually, I'll say the macroeconomic indicators have been strong. If we look, the handset demand has been strong. I think because of our licensing business, we have a good understanding of the overall demand. We look at sell-through data, also very strong. But unfortunately, I think what we saw in Q1, as we guide to Q2, is 100% sized by the availability of memory.

So as we all know, as all the indications show, the DRAM availability for consumer electronics, especially handsets, is actually down based on year-over-year because of the prioritization of HBM for data centers. I think the market is going to be sized by that. And I think we saw the reaction right away from our customers who are adjusting. I think they build production through the memory they have available. And I don't know, Akash, you'd like to add some more color to that?

No, I think that covers it.

Okay. I guess to follow up, I mean, just backing into the guidance you just gave on QCT, I mean, that auto number is implying a pretty sharp acceleration sequentially. Is this some of the ADAS wins that you've talked about previously layering in? And maybe you could speak to both the drivers and the durability of the high watermark that you're guiding to.

No, thank you very much. As we have said consistently, I think the pipeline we have built in automotive, it has continued to translate into revenue, especially as new cars ramp and new cars launches. And I think that's why we continue to see record revenues in automotive. We don't move with the industry, we move primarily with our share gains. I think we're very excited about the trajectory. I will say, we feel good about all the projections we have made about the size of the revenue. When you look at our targets for fiscal '29, it's all going in the right direction. And we continue to have more design wins.

I think our position in the industry becomes stronger, I think, with the platform. We're seeing traction with Flex, which both the ability to bring ADAS and digital cockpit in the same chipset across other tiers. We're seeing now some of the major, I think, volume drivers achieving SOP. We did announce a very broad partnership with Volkswagen Group. And to your comment, it is correct. We're getting more traction with ADAS. Once OEMs were able to see the stack that we launched with BMW that was an option for them, we're seeing interest and those things are progressing very well.

The next question is from the line of Samik Chatterjee with JPMorgan.

I have one on data center and one on the smartphone side. Maybe on the data center, Cristiano, if you can give us an update in terms of the progress with your customers on that front. And given sort of the volatility we are seeing in memory, is that sort of being more disruptive to making progress with your customers? Or instead, is it sort of augmenting some of the pace of the discussions given sort of a big focus on that side of the sort of bill of materials as well? And I have a follow-up.

Thank you so much, Samik. So let me start with the data center. I think everything is going in the way we have planned. I think the only public, I think, customer announced today is HUMAIN. That is progressing well. We have started shipping. We have been working with them in ISV on third-party workloads. We're encouraged about the progress our teams are doing on our road map.

We continue to get very positive feedback, I think, from broad engagements you would imagine that a company our size will be engaged in conversations with some of the largest hyperscalers and cloud service providers in the industry. We have something very unique. We always said we have a dedicated platform for the disaggregated data center. We do very, very well in certain workloads such as decode with our different approach to compute and memory. If anything, I think the transaction of Groq kind of validates that when you think about disaggregated data center, you have specialized hardware versus just a GPU that would do everything. And we're getting good traction.

Where we're really focused right now is on execution. I think we had identified some of the milestones. We're executing on 2 fronts. It's CPU. We added a RISC-V CPU now to our road map in addition to Oryon, which is ARM compatible. And we're executing on AI250 with our new memory architecture. And we will provide details of our road map in our investor event. But so far, everything is on track. We still restate that we expect '27 to start showing in revenues. And we feel good. We're just going to keep executing that.

And I don't know, Akash, you want to add anything before I go to memories.

No, I think the only thing I'll add on data center is, we've mentioned previously that we expect this to be a multibillion revenue opportunity in a couple of years. And so everything that Cristiano outlined kind of just reiterates that opportunity for us.

Okay. So the memory thing, and look, I think we're going to see how this thing played out. I'm going to give you maybe a little bit of the dynamics. When we step back and we look at the business, we're very, very happy with everything in the business. We just wish there was more memory. And the handsets get hit the most given its scale and its cycle time. So we expect that the impact is going to be more muted in other business. For example, automotive is a little bit less sensitive to memory price increases. As you pointed out, the impact on handsets and the BOM.

Having said that, when we go back to situations that we saw in the past, I think the best proxy is what happened during the pandemic. The premium and high tier has proven to be more resilient to price increases. And we think that, that may be a factor that plays out. But the most important thing is that issue is not just the price. The issue is just availability. So I think the memory availability will determine the overall size of the handset market.

OEMs are very likely to prioritize premium and high tier, how they have done in the past. That could be less impacted, and we will see the reaction on consumers as their price increases for the finished product. I do stand by what I said. I think the whole fiscal year mobile handset size will be determined by memory availability, and we're just going to monitor this on a quarter as the phones get repriced, tiers kind of shift towards high-end premium, and we'll see what happen in the marketplace.

Got it. If I just can quickly follow up on that, Cristiano. On the OEMs prioritizing the higher tier, I mean, within that higher tier, do you expect them to downshift in terms of the tiering of the chipsets or the SoCs that they go for just to be able to manage their overall cost in relation to what they need to pass on to consumers? And that's it for me.

So as a general trend, and I wanted to emphasize what we saw in the quarter. Yes, there's a memory shortage, but when there was memory, we saw the result was very good. The consumer demand was very good. And what we have seen, which has been going on like for years now, the premium tier continues to expand. In a market that has been relatively flat, which is the handset market, we have seen growth in the mix with the premium tier expanding. So I think that's a factor that is likely going to drive OEMs to continue to be focused on the premium tier.

I did mention one thing in my prepared remarks, which is a dual flagship strategy that we have adopted, and that has been also very well received, I think, by the market. You probably see that when you think of different OEMs, how they have like ultra or different categories, and they have multiple tiers of the premium tier, I expect that's going to play. But overall, our hope is that the premium tier will be more resilient. Granted, the memory that is available is the memory that's available.

Next question comes from the line of Ross Seymore with Deutsche Bank.

You mentioned a couple of different things on the handset side, for my first question. But I guess what it comes down to is, what percentage of your handset business do you think is in China, considering that you cited them as being especially hit? And do you think normal seasonality is likely to occur after the step-down in the March quarter? Or is that too difficult to tell?

Yes. I think on your first question, Ross, we don't really kind of break down by regions. But if you think about the percent of volume that is driven by the Chinese OEMs, but then adjust it down for the tiers that they play in, so our exposure would be less than what you would just see based on the units.

And the seasonality side?

The seasonality on the handset side. I think you should think of the seasonality in the demand from the consumers is going to be consistent with what we've seen in the past. I think consumers wait for premium tier launches and there is significant purchases that happen when that plays out. I think to Cristiano's earlier point, it's really a question of how supply aligns against the demand. We don't have a demand issue, as we said earlier. The demand continues to be strong. Our design win pipeline continues to be strong. And then it's just a question of supply alignment with it over the next few months.

And I guess just for my follow-up, on the OpEx side of things, you gave a good explanation why it's popping up a bit in the March quarter. After that, are there any adjustments given what you're seeing in the memory side? Or are you guys kind of investing right through this?

I think it's -- the way we've guided the March quarter is a reasonable way of thinking about the rest of the year. I think our focus, as we've said before, is the following framework on OpEx, really kind of reduce the investments in mature businesses and use it to fund the diversification priorities. And then we have these acquisitions, including Alphawave kind of driving incremental expense and investment in data center. But it's really just focused on those things. As we've been extremely disciplined over the last several years and grown OpEx significantly slower than revenue and gross profit, that framework for our operating plan doesn't change going forward.

The next question is from the line of Stacy Rasgon with Bernstein Research.

For the first one, I want to ask that seasonality question a different way. I think it was really getting at June. Like usually just seasonally, I know your revenues step down in June. So you're guiding $6 billion on handsets. You're guiding $6 billion in March quarter, which is down about 13% year-over-year. Are you expecting, just given what you're seeing in the memory market right now, a similar -- what can be supplied as similar type of year-over-year growth like for handsets in June? Or do you think this like $6 billion number, given it is sort of supply constrained, is like a good number to have given the current supply that is out there until things normalize? Like just how do we think about June in the context of March, given the March decline in the context of the memory situation?

Yes. So Stacy, given the uncertainty in the market, we're obviously not guiding beyond the second quarter at this point. But as Cristiano said earlier, when you think about the demand fundamentals, they're strong. And really, it's a question of how supply aligns against it. And we expect that supply will really define the financial forecast for the year -- for the rest of the fiscal year. Specifically kind of between quarters, you should think of March as a reasonable way to model June as well. It is really kind of similar seasonality profile that you would have seen in other years.

Got it. And for my follow-up, I want to ask just about QTL. So again, it sounds like the demand is there, but we just don't know how many handsets are going to be able to be built. I guess in that context, how are you thinking about sort of like just the typical QTL run rates in the various quarters through the year. Do you think they're similar to what we've seen in the past? I think your guidance is maybe in line to maybe slightly below what we typically see for March? [indiscernible] and below? I mean, how do you think about that?

Yes, Stacy, it's Akash. So let me try to address it a couple of ways. I think first is just strong performance in December quarter. We saw handset units higher than expectation in the quarter. I think as you go into the next quarter, we are guiding QTL just slightly below what we did last year. So pretty consistent with trend. But of course, that's subject to supply considerations. As you think about the full year, at this point, given the supply, we have a negative bias on units. But really, we're going to have to see how it plays out as we go through the next several months.

Got it. So maybe a touch below, though similar to what we saw in March seems reasonable given what we know right now?

I think that's the framework that I outlined is the way we are thinking about it.

The next question is from the line of Timothy Arcuri with UBS.

Akash, I wanted to ask about the op margin guidance in QCT. The drop-through is more than 100%. I mean, it's not surprising that margins would come down, but they seem to be coming down pretty quickly, like faster than I would have thought. Is there something else going on there? I know that wafer costs are going up and MediaTek said on their call that they're still gaining share at the high end. Is there something going on to make the drop-through more than 100% on the op line for March?

No, there isn't, Tim. I think we're expecting gross profit margin to be largely in line with the December quarter. And so it's just the scale of the revenue coming through and the OpEx guidance that we've provided.

I just want to add one thing. No, look, we saw how I think the other company reported as well, very consistent view, I think, on what we're seeing sequentially on the quarter. It's just the whole market has kind of been adjusted to the new build-out reality. So we actually don't see anything other than that. And remind you of the seasonality that we always have, regardless of this memory issue, a lot of the premium tier launch in Chinese New Year. So you actually normally see some of the Chinese go down on a sequential basis, because they just build for the premium launches.

Okay. And then do you have any update on the Huawei license? I know we're still waiting for it. And maybe what's the sticking point? And is there a risk -- we talked about this before, but is there a risk and precedent for the big customer if you don't sign a license with Huawei?

Thanks for that. This is Alex. Really no update on the Huawei discussions. The discussions are still underway. In terms of sticking points, I really can't get into what are confidential discussions. I see these 2 sets of negotiations as fairly distinct, actually significantly distinct, operating on different paths. And as you know, with the other company, whenever we see a renewal date on the horizon, we start discussions very well in advance. And so that's underway, and we don't have any update on that.

The next question is from the line of C.J. Muse with Cantor Fitzgerald.

I guess curious, obviously, the DRAM makers have been talking about satisfying only 50% to 70% of the demand, and they're highlighting shortages into 2028. So curious how you're planning for a situation where this could be sustained. Are your Chinese customers looking to design in CXMT? And could you get qualified inside of that? I would assume your business with Samsung would be strong given their internal supply from DRAM and as well as your supply chain in terms of your wafer commits to TSM. I guess how are you managing all of that given all this great uncertainty?

Look, very good question. And I'm going to -- I know it's obvious, but just in case, I'm going to use this opportunity to make clarification. For handsets, we don't buy memory. I think -- I know there is some memory that gets stacked on modems, but the majority of the memory is purchased directly by our customers. You should expect, given our scale, we're probably among the first to be qualified with every memory provider. Every single memory you can imagine, CXMT and other smaller companies, we have been qualified.

And also, we have flexibility versus some of the other companies. If you actually double-click, you're going to see we have flexibility about working with new versions of memory as well as older version of memory on our platform. So we have multi-generation memory controllers. So from a platform perspective, we're going to work with whatever is available. I think that's kind of the approach we always took when you have shortage.

So the second part of the question, which is the bigger question. Look, the trend, I think, of growth in the data center continues, and it's pretty obvious. I think the memory vendors have prioritized the build-out of HBM. And I think some of the data that you just provided is kind of what we see. As I said before, it's a very clear indication that as of today, the availability of memory for consumer electronics year-over-year has been below the demand, and we've seen that in handsets. You start to see commentary on gaming consoles and other consumer electronics devices.

We can't really predict if this will continue for '27 or '28. I think there's capacity build-out in plans. It all depends also how much the trend on data center continues to accelerate. It is fair to assume at this point that for the fiscal year, the size of the handset market, which is one that is probably getting the blunt of the impact in our business, is going to be defined by the availability of DRAM.

And C.J., on your second part of your question on wafers for leading nodes. As you know, kind of leading nodes are constrained on the wafer side as well. But we have great relationships with our suppliers, and so we're confident that we'll have enough wafers to address the demand.

And I guess as a follow-up, curious, if we do see a mix shift higher Snapdragon, but unit volumes lower, how should we think about that impacting your QCT EBT margins?

Yes. I mean, as you know well, CJ, we do very well in the premium and high tiers. And so as the volume shifts up, that is usually a benefit for us.

The next question is from the line of Ben Reitzes with Melius Research.

I wanted to just kind of keep going on the memory side. As we kind of look at Apple and their propensity for double-digit growth, maybe even the whole year, it just seems like they are going to continue to get disproportionate share of the available DRAM. Is it possible -- how do you kind of navigate that with all your partners? And I guess the question would be, does that add to some of the uncertainty that could linger into the next fiscal year with one vendor getting disproportionately this kind of unit growth and obviously, the kind of allocation they get?

Look, it's hard to make a prediction, but I'll also probably remind you that we have another large customer that also have the memory division as well. So I think as a general statement, I think it's probably a fact that OEMs with larger scale will have probably better ability to have enough memory, and they will make priority calls than OEMs with smaller scale. But I think this problem is probably going to be industry-wide. I don't think any OEM has been immune. In general, I think the statements we have seen broadly in the industry is not a demand issue, it's all supply constraint.

Okay. Well, look, there's been a lot of questions on that. Just my next one, I just wanted to double-click on the data center. And I know you got asked about whether there'll be memory available for that. But just in terms of the recent events that validate, I believe, the decoding aspect of your solution. I was wondering if you could just provide a little bit of an update there. What's happened since Groq with NVIDIA? And how are discussions going beyond HUMAIN, and just that overall trend and your ability to play?

Look, here's what I can say without I think front-running our investor event. First of all, I think we are -- I would describe it like this. I think there's a lot of companies right now that recognize, I think, the technology and the technical capability of Qualcomm. I think our track record on technology execution has been very successful. And I think we also understand some of the dynamics on compute and memory, I think, given the breadth of our IP road map. We're probably one of the few companies that go from sub-5 watts to now all the way to 500 watts.

And we have said in the past, as we're going to enter this market, we needed to kind of intercept where the market is going, and we're going to be really, really focused on inference and especially the disaggregated. I think you pointed to the right way. I think, for example, decode applications, we believe we're incredibly competitive, not only from a power consumption, but also from an overall TCO, compute density, memory density.

And we're really focused on execution. The feedback we have been given from a lot of the large companies on the technical side and on the product side is very positive. Now the ball is in our court to execute, have hardware available and kind of show the results and/or just kind of continue doing that.

That concludes today's question-and-answer session. Mr. Amon, do you have anything further to add before adjourning the call?

The only thing I want to add, look, it's unfortunately, I think that the whole sector is impacted by memory, but we remain incredibly encouraged about, I think, the foundation we set up of the company to be relevant to many industries. We are on track to the commitments we made on the diversification revenues for the company for fiscal '29. We have, in a record time, I think, been having a very good traction in the future opportunity of robotics.

Physical AI in a robot is the best example I can provide other than autonomous driving of what edge AI is. And we believe that we are creating really a completely different company with relevance in many, many markets, and what we'll just continue to execute, I think, on our road map. I would like to thank all of our partners, suppliers, they're dealing with us in this memory shortage, and our employees, and we look forward to talk to you next quarter.

Ladies and gentlemen, this concludes today's conference call. You may now disconnect.

Okay. We're going to get started. Good afternoon. I'm Tim Arcuri. I'm the semiconductor and semi equipment analyst here at UBS. We're very pleased to have Qualcomm, we're very pleased to have Cristiano Amon from Qualcomm, who's the President and the CEO. So thank you, Cristiano.

Thank you, Tim. Good talking to you.

Great. So let's start talking about the business that everyone wants to know about, which is your data center business.

I'm surprised.

Yes. So you're attacking the low-power inference market. You announced the AI 200 and AI 250. You still haven't told us very much about the specs and the road map. What can you say?

Okay. Look, maybe let me start with the very top and I'll walk to some of the details. I think we look at the market, we look at what's happening with AI. And I think I'm assuming is the hope of everybody in the room that eventually, you go from training to very large-scale inference and you start doing inference, you're putting AI to work and you have a lot of customers.

And I think what's happening is we see this as a one of the entry point for us, as AI is going to go into inference, you're going to be able to build large inference-focused clusters, the data center is going to go to the next phase of this aggregation. You're going to have dedicated hardware tooling inference. And we think that creates an opportunity, I think, for us to enter. We expect there's going to be competition, everybody is playing to win.

And at some point, I think tokens per dollar will matter, tokens per watt will matter. And we have an opportunity to come up with something that is very competitive for inference. The second, I think data point, I think to drive the entry is how we see AI evolving.

Eventually, it's been kind of the as you put this into work and you started to see the combination of mix of experts and distillation, chain of reasoning. You see that the AI is really becoming mature to the point that it's not going to be farfetched that you're going to have small appliances that it could be doing multiple hundreds of billions of parameters of model.

If anything, you look at NVIDIA DGX that's kind of what we do, and that trend will continue. And I think you're going to end up in a situation that architectures are going to be available for inference in the data center is going to be competing, I think, for the efficient architecture. With those 2 things in mind, we look into some of the assets that we can leverage in the company.

And we have 2 assets, and that's what we'll be focused on. 1 is we're doing a CPU, which is the head node of an inference cluster. And the other one, which is the largest opportunity is leveraging on our NPU architecture, which we believe has very high computer density, a completely different approach to how we think about the compute and memory, I think, together and thinking about developing a very efficient inference solution.

The good thing about Qualcomm, we don't have to get a lot. I think we only need to get a small portion of this very large tenant is very significant for the company. We also like the fact that a lot of the market is concentrated. You have a few customers that buy at scale. The different thing is the market welcomes competition. Qualcomm is not a small semiconductor company. We can do things at scale. I think we have like a proven track record of executing in a number of different industry at scale.

And I think the market has been very intrigued about what we're doing. We're doing something different. We're thinking about the next disaggregation of the data center. And we're excited about it. We will unveil details of the road map, both the AI 200, the AI 250, what we're going to be doing after the AI 250. We announced it a little bit prematurely because we had a customer who had to announce it.

I think the first customer that we have is 200 megawatts of data center with the Saudi national AI company. We're very pleased with the progress they made with the license and everything that's moving to execution. We are in conversations, as you would imagine, with all the hyperscalers. And we're very pleased, I think, with the feedback we're getting so far.

Great. Is it fair to characterize so that 200 seems more evolutionary from what you currently have and the 250 seems more bottom-up purpose built, and that seems more ramping in [ 2020 ].

100% correct.

Is that fair? And then I think you're trying to attack the decode portion of inference workloads. When I hear that, it sounds a lot like what NVIDIA is doing with CPX, Rubin CPX. Is there really a window of opportunity for you? Or is it tide just rising so fast that you think you can get some of that market as well?

Look, I think -- like I said, -- the interesting thing about the inference clusters is you ended this aggregation. I think as other, I think what's happened with the decode. I think when -- some of the discussions we have has been fascinating to see how much more performance that you get on how much you manage cooling or even like a few percentage increase in performance, how it changes, I think the total TCO.

So I think there's definitely an opportunity for Qualcomm. It is driven by all those things. This is moving very fast. There is going to be competition. People need more compute that they can deploy. I think the demand is real. So any improvement it makes a lot of sense.

And I think we have a good technology. I think just if you look at our track record, all of the new things that Qualcomm have done, even things that were new to scale, our IP is a leading IP. So why would you not bet the Qualcomm can do a competitive solution? I think that's kind of the feedback we're getting.

Great. And relative to your financial model, this seems like it could be pretty significantly incremental actually?

Absolutely.

Your $22 billion for fiscal '29, that's non-handset. I mean this is a huge market. So you don't have to get a very big share of it to be pretty incremental to your financial models?

I agree. It's 100% incremental, not model in our $22 billion of no handset. Those are for the other markets we've been executing right now. And we did say in the last earnings call, I think we feel confident to -- we historically have been very conservative with some of those assumptions. We feel comfortable pulling in by 1 year, where we originally set.

And how does the Alphawave deal fit into the strategy here? Does it intersect the road map for 250, which ramps in 2028?

Yes. The Alphawave provides, I think, important, I think, connectivity IP that allow us to really scale the solution. So -- and they also have a team that has been doing custom SoCs for the data center. So I think it's both provide the scale, connectivity IP as well as additional resources to execute on the opportunities.

Let's just talk about the general -- the adjacencies in general. These have been growing very, very nicely. They're up to 30% of revenue. Obviously, more if we exclude Apple, which of these efforts are you most excited about between auto, IoT, there's a lot of submarkets within IoT. But auto, IoT and PC, which of these are your most...

Look, we -- auto has been a great success story for the company. We continue to be bullish on it. I think when we look at the opportunities we have on the horizon, we see a lot of opportunities to even expand our design win pipeline, which is very robust. I think so far, you're seeing this pipeline converted into revenue. And I think we're tracking very well, but we see opportunities to expand the design win pipeline.

And that's happening because the same thing that we see on personal AI devices, which I'll talk to you in a second, we're going to see that happening in the digital cockpit of the car, a lot of GenAI and Gentech experience coming to the digital cockpit of the car that's going to increase the value and the silicon opportunity there. The other thing that we see is we're super pleased with the stack that we launched with BMW.

This is an OEM-friendly stack 3.5 years into making -- we have a lot of inbound for OEMs really interested now that they'll be able to see the KPIs, the cars are in the road. And I think that could create expansion opportunity for ADAS stack. And we like the ongoing transition to software-defined vehicles. I think the architecture of the car is going more towards center computing.

So Auto will continue to see growth of the design win pipeline, and we're very pleased. I think the Snapdragon digital chassis became like an industry platform. The second one is the one that I mentioned briefly personal AI devices. So I think the best way to describe this, and I -- this is a topic that we spend a lot of time thinking about this, the evolution of mobile.

Phones will continue to stay on this trajectory that they are right now. They're going to require more and more AI compute. Phones are not going anywhere. Phones are like laptops. Laptops continue even after we all bought smartphones, but phones will do a lot more processing, but there's a new class of devices that are going to drive agentic experience are going to be personal advice.

Glasses is the 1 that is the most promising, but there are others. I think you see all of those companies that have models designing different types of devices. The good thing we're designing all of them. And we think that's going to drive new agentic experience that could be a significant opportunity. And it can grow dramatically. We talk about $2 billion in our $22 billion projection for that we said in the last 2024 Investor Day, we said that $2 billion will be XR devices.

We're well ahead of it when you think about what's happening with personal devices and glasses. I'm very bullish on that opportunity. And I think this is going to play out like this. When the phone is at the center of our digital life right now. So all of the wearable devices, they have been around the phone. They extend the functionality of the phone, like a watch get the phone sensor data, give you back notifications. They actually, the first project we did with Meta was to -- for you to do Instagram stories, extend the functionality camera.

It doesn't matter. Now the agent is at the center. And those devices, they get better every month. Every month, they are new use cases. I think we see customers for example, even completely new markets, we have a customer in India, they are doing glasses. They integrate it with the national payment system. You can look at the QR code and you can pay a bill. And I think that's going to be -- those new agent experience are going to develop this category can be very, very big, and it's going to change a lot the relationship that we're going to see happening in the mobile industry.

The phone will continue to do phone things. The phone will do more processing for those devices. But then those devices are going to be developing around the model. Once you connect to a model, it doesn't matter how you connect to the model. The model will understand the human intentions, and to take action. And I think that's going to drive a combination of connectivity and processing of those new devices and it may change the dynamic of the industry.

And here, I'm going to make -- provide an opinion. Humans already decided what they're going to wear. Humans will wear bracelets, watches, jewelry pendants, glasses and if the model understand what we see, what we hear, what we say, it's closer to our senses. That's why glass is very natural. You turn your head, that's the camera is seeing what you're seeing, close to your mouth, close to your ears. And this is going to be a combination of technology and fashion, different countries, different regions, different brands that is more conducive to a horizontal platform than to a vertical platform.

So I think we're very optimistic about what Google is doing with Gemini, what Meta is doing and what OpenAI is doing. And this could be an interesting new category. So that's the second one, and I'm excited about it. And then the third one, I think we recently closed the acquisition of Arduino. We've been saying that we've been building a platform for industrial.

I think that's the ability to bring high-performance compute and AI to industrial replacing microcontrollers. We were building a development platform, Arduino and Edge Impulse are big drivers for that and just build more confidence on our $22 billion revenue.

And so in that example, the model started on the phone. In your world, the models on your phone?

No. See, I know that there's a lot of questions like this. And before allow me to be, I think, very precise. Before we used to get a lot of questions about this, is a cloud or is edge. It's a cloud or it's edge. Now we get a lot of questions. Is the model running on the phone? Is the model running on the last, where is the compute is. Let me take a step back.

The smartphone today is the most cloud connected device in the world. If you put your phone in an airplane mode, you're not going to use it. You're going to be very frustrated. You're just going to put it back in your bag in our pocket. But having said that, there's a lot of processing that happens on the phone. I think you need to be thinking about this a little bit different.

First of all, the answer about models, where does the model understands what we see, what we hear and what we say are located. They are located in the device because latency is unforgiving. So anything that has to do with a user interface is located on the device. So we announced the first chip in a glass in a frame that does 1 billion parameter model. The next ship is going to do much more every company, things like voice to text, the ability to quickly annotate and image and do things, they're asking us, I needed to have that into the device.

At the same time, those devices are going to have to need a lot of connectivity because a lot of those things are going to be happening in the cloud. The other thing you're going to see is with mix of experts with chain of thought reasoning, you have smaller models, there are certain things and then the bigger model doing some other things. And you can easily see how this play out, whether you're looking at a QR code, for example, you look at an image, you're looking at a person and how this thing is going to transaction.

So at the end of the day, I will go in and make a statement that the work product of foundational companies right now. They're being designed in the way that they have a cloud component and an edge component. I think if you look about your road map of Google Gemini, you see that. You saw that the open weight OSS model from OpenAI, you see that.

So I think the models are being designed. It's the evolution of computing and I think what you're going to see is those are going to be a combination like a phone. There are going to be certain functions they're going to run on the glass, on the watch, on the phone. Certain functions are going to run on the cloud. And the last data point. You're going to see from us a big change in the computing architecture for a smartphone.

We're working on it for a couple of years -- we're going to announce, I think, as we head into 2026, maybe in the -- in regular announcements in the second half of the year. But there's a new architecture. One thing that is happening context is super important for agent experience. And even when you think about advertisement, right, if you remember, I'm going to quote when there was this incident that unmet that they lost access to some information on the iPhone and they use a lot of AI to compensate for.

So -- when you think about context that happens around you, that becomes incredibly important for an agent, incredibly important for you to create a personal graph. So we're seeing a lot of demand models that they run on the phone and they run in a very pervasive way at a very high performance or very low power just to get context. Just to basically get sensory data.

And then I think a lot of people want to do that on the phone because that's how it's going to scale and the phone in real time has real-time knowledge about what's happening -- that's also true for the devices. So that's another change of computing that we're going to see unfold. Sorry for the long answer.

No, it's great. So how much -- in all those cases, how much R&D dollars, like how much can you repurpose versus how much do you have to spend from an incremental perspective because all those markets seem to be peripherals of what you've already -- of your existing sunk costs?

Look, if you look at the company financial results, with the number of bets that we're doing in parallel, right, we are -- we are in the phone business and we are like clockwork. We execute on our Snapdragon premium every year plus other Snapdragons and the rest of the road map. And we have maintained the performance leadership.

And now on top of that, we added our own CPU. We don't license anymore. We design our own CPU. On top of that, we're in the PC market, we are into the broadband market we are into the industrial market right now, automotive, including the STACK and you have now the data center and robotics. Those are the things we're all doing.

And if you look at the financial performance of the company, we have been increasing efficiency a lot, if you look at the number of bets that we're making. And that is because we leverage a lot of IP, and we create an ability to scale our IP from 5 watts to 500 watts.

We're probably 1 of the few companies. They have the engineering capability. If you think about how broad our portfolio is to scale from 5 watts to 500 watts I think that muscle we develop in the company as we set ourselves by necessity to diversify is really helping us and enabling us to do more things that have become efficient and accretive to the results of overall Qualcomm.

Let's shift to the handset business. So you had a great quarter last quarter. You're doing very well in handsets as especially in Android. If you exclude Apple, your trailing 12-month handset revenue is up nearly 10% in a market that's basically flat. How are you doing this? And how long can you keep doing that?

Okay. This is an important thing that we have been seeing playing out for several years now, several years, quarter after quarter after quarter. Actually, most people don't realize, the phone market right now is still smaller than it was before COVID. Still smaller. We have not yet recovered in units, the size of the phone market that was before the pandemic.

But how are we growing on continuing to grow into a relative flat market? We have seen this trend that has been very consistent in the phone industry, especially as you get fully penetrated that the premium and the high tier expense. If you go to the United States market today. The United States market today, there's only 2 phones to buy. You have an iPhone or have a Galaxy phone. That's a 2 premium phones or you go to Walmart and you're buying a prepaid phone. There's nothing in the middle, right?

And I think China is now becoming like that. you have an expansion of the premium and the high tier, you have a low-end market. The middle is contracting. I think if you remember back when I became CEO, I think I said in 2021, I said our strategy and phone is going to be very simple.

We're going to go after share wallet, we're going to concentrate on technology leadership in premium and high on Android, and we significantly increased the up margin of the business because the premium here is more resilient, I think, to the cost of technology, once more compute, once more performance and it's been in expansion.

Even markets that are -- like India, which has been historically very price sensitive, we see an expansion of the premium tier that is happening across the board.

I wanted to ask about India, since you mentioned it, can we actually talk about it, it seems like China maybe 10 to 15 years ago. How much of that -- how much of a driver of your interbusiness can India be?

Look we see continued healthy growth for that market across multiple categories. We're designing and all the Indian automotive companies. We were doing -- Snapdragon continues to gain share -- we're very happy about our Snapdragon brand position in India. It's probably as high as in China right now on consumer preference. We see a healthy expansion of the premium tier.

And here's what you need to think about it. I think I may give numbers that are not entirely precise, but they're directly correct. The phone market today, it's about 1.2 billion phones get purchase every year. $200 million is iPhones, $1 billion is Android.

So -- and a market like India, like your comparison is correct. It's a very large market. So as that android market starts to drive towards a high tier or a premium phone. That's the most important, I think, electronics purchase. That continues to drive quarter-after-quarter, year-over-year, a positive outcome of our handset business.

Great. And then can we talk about -- there's kind of a strange dynamic where some of the major Android OEMs have their own internal efforts for modems. And yet these same customers seem to be more dependent on you than ever. Samsung, we used to say 50% baseline share, now it's more like 75% baseline share Xiaomi, you just signed a big deal with them yet they too also have their own internal efforts. How do you look at that?

Look, first of all, Tim -- and by the way, I don't mind we love doing this. But I've been answering this question for about 25 years. 30 at Qualcomm 25 years remember, it was managing CDMA and people said, the our customer just designed their own CMA chip. And I think historically, there has been an ongoing thing in the industry.

I think the answer to that question is you need the scale. You need to have the ability to change to technology transitions very fast. Every year, we have a new Snapdragon 8 that's kind of designed to set the pace of performance. You have -- it's a very competitive market. Maturity of silicon design is very important.

The mobile market is super unforgiven because what happened is you design on a brand-new process node, brand-new IP across your CPU or GPU, you have to bring up a whole new technology. You have to ramp from 0 to $100 million in 2 quarters. You have to make the selling season. Then you kind of tail off, you prepare to ramp the next one on an annual cadence.

Look, we're called Qualcomm. It stands for Quality communications. Some of the things we had to develop as being part of this business, some of the other semiconductor company, when they find out about it, they think we're crazy. Sometimes we go into mass production. Mass production on a design before we actually get our chip back. We haven't even got a chip back. We don't know if it has any bugs. We do a mass production because of the speed of mobile. So I think that has maintained our ability to stay in front to continue to have the designs that matter for our customers and have a leadership in IP.

And there's another thing that most people don't think about that, don't think about Qualcomm. We actually -- ourselves, we're very focused on our speed of our CPU or NPU or GPU or modem, but Snapdragon brand is very powerful. And I point you to do one thing. We had a Snapdragon Summit, and we do it every year when we launch it, 0.5 billion views impressions on the media. We had a simultaneous event in China.

Every single one of our customers decided to launch their phone at our event at the same time. If you look some of their advertisements, you see a bigger Snapdragon than you see a picture of their phone. And we didn't know. I think through the process and knowledge, like we're 39 on Interbrand 100. So Snapdragon also has a big community effect that continues to drive our customers to put Snapdragon in their flagships.

Great. Let's talk for a moment about Apple. You've been very clear about how that's going to come out of your business. baseline being that the fall launch this year or sorry, next fall will be 20%, and it's fully out in the fall of '27 launch. However, if you look at how the internal modem is selling, not selling as well, maybe as they hoped, and there are -- there's some talk that they're kind of coming back to you again, had in hand, maybe wanting to extend a little bit longer. Can you just talk to all that?

Look, first and the most important thing in this conversation, I think we have been planning our business, assuming that this customer is going to go away in '27. I think that's our planning assumption. That's how we've been managing the company. That's how we're accelerating our diversification. We had provided metrics, which we're very pleased with it, how the non-Apple business in the company is growing. It showed that our strategy is working, our engine is working, and we're going to keep doing that.

Having said that, I think, look, -- we are a high, I think, quality provider of modems to them. I think it is -- it is a very reliable modem, and they can continue to use our modem as long as they want to use our modem. We just are planning our business on the assumptions that we told you, and that's what we're marching forward.

Got it. And maybe I'll ask you the question that is still a debate. So how do you think about the lack of when they don't need about them? How do you think about the precedent that they will continue to pay a royalty and just the precedent that they say, well, Huawei is not actually paying a royalty. Now by then, they might be, but today, they're not.

Look, the licensing business, the oldest business of Qualcomm, I think, has been probably battle harder by at least 4 generations of wireless. And I think look, we have one of the largest IP portfolios in the world. I think the -- it's -- we're the #1 company in -- across all American companies and patterns applications in 2024. Let's see, '25 didn't close it yet.

I think we have a very strong patent portfolio, not only in cellular. We have that on video and WiFi. I think our model has been proven to be fend, been tested in any government or every regulatory agency. We battle tested that with Apple.

So look, we're confident in our position. I think like -- like every other licensee. I think I point you to -- I know you mentioned about Huawei, but every other Chinese company has renewed their agreements. I remind everyone, I think, we're licensed with Samsung, including 6G. So we think that's a very stable business, and we expect to continue to be that way.

Great. Well, thank you for the time. We're out of time. Thank you.

Thank you so much. Great taking to you. Thank you.

Ladies and gentlemen, thank you for standing by. Welcome to the Qualcomm Fourth Quarter Fiscal 2025 Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded, November 5, 2025. Playback number for today's call is (877) 660-6853. International callers please dial 201-612-7415. The playback reservation number is 137 5609. I would now like to turn the call over to Mauricio Lopez Hodoyan, Vice President of Investor Relations. Ms. Lopez Hodoyan, please go ahead.

Thank you, and good afternoon, everyone. Today's call will include prepared remarks by Cristiano Amon and Akash Palkhiwala. In addition, Alex Rogers will join the question-and-answer session. You can access our earnings release and a slide presentation that accompany this call on our Investor Relations website. In addition, this call is being webcast on qualcomm.com, and a replay will be available on our website later today.

During the call today, we will use non-GAAP financial measures as defined in Regulation G, and you can find the related reconciliations to GAAP on our website. We will also make forward-looking statements, including projections and estimates of future events, business or industry trends or business or financial results. Actual events or results could differ materially from those projected in our forward-looking statements. Please refer to our SEC filings, including our most recent 10-K, which contain important factors that could cause actual results to differ materially from the forward-looking statements. And now to comments from Qualcomm's President and Chief Executive Officer, Cristiano Amon.

Thank you, Mauricio, and good afternoon, everyone. Thanks for joining us today. In fiscal Q4, we delivered another strong quarter with revenues of $11.3 billion and non-GAAP earnings per share of $3, both of which exceeded the high end of our guidance range. QCT revenues of $9.8 billion, up 9% sequentially were driven by strong end customer demand for Snapdragon power premium tier Android handsets continued traction for automotive Snapdragon digital chassis and strength in IoT across industrial WiFi 7 Access Point, 5G fixed wireless and smart glasses. In addition, all 3 QCT revenue streams exceeded our expectations, including record automotive quarterly revenues in excess of $1 billion. Licensing business revenues were $1.4 billion. Fiscal '25, non-GAAP revenues of $44 billion were up 13% year-over-year, with record QCT annual revenues of $38.4 billion, up 16% year-over-year, including automotive and IoT revenue growth of 36% and 22% year-over-year, respectively.

We delivered 18% year-over-year growth in total QCT non-Apple revenues above our prior estimates. We remain on track to achieve our fiscal '19 long-term revenue commitment as outlined at our 2024 Investor Day. I will now share some key highlights from the business. At Snapdragon Summit in September, we introduced our Snapdragon 8 Elite Gen 5 mobile platform for next-generation flagship AI smartphones. This platform is equipped with our custom-built third-generation Orion CPU and the fastest mobile CPU ever, along with an upgraded NPU and GPU with the Snapdragon 8 Elite Gen 5, we continue to set the pace of innovation in mobile processors. This year marked our tenth Snapdragon Summit with simultaneous events held in Maui and Beijing, validating the strength of our Snapdragon ecosystem leading China OEMs, including Xiaomi, Honor, Vivo and Oneplus announced their flagship phones at our event.

More than 1,100 partners, analysts, tech influencers and press attended in person and our key notes capture over 26 million unique views across both events. Together with our announcement, Snapdragon Summit generated over 547 million social media impressions. In addition, our Snapdragon in ciders community of tech enthusiasts, developers and fans has grown to more than 20 million members worldwide. Our highly differentiated technology continues to drive increased brand visibility. And during the quarter, Qualcomm debut at 39 on the Interbrand Top 100 Global Brands list for 2025, reflecting the strength of Snapdragon. And for the first time ever, Kantar's brands the most valuable global brands list included Snapdragon, where we ranked #38. Also at Summit, we unveiled our newest platform for premium laptops, the Snapdragon X2 Elite and X2 Elite Extreme. Once again, our industry-leading processors continue to outperform competitors surpassing Intel and AMD in both speed and power efficiency.

Our latest NPU sets a new benchmark as the world's fastest AI engine for laptops also exceeding Intel and AMD and performance. And with the new Orion Gen 3, we have the world's first 5 gigahertz CPU for the ultra mobile laptop category with extended battery life. We now expect approximately 150 designs to be commercialized through 2026 and remain optimistic about the continued momentum for Snapdragon powered AI PCs. As AI transforms HUMAIN computer interactions, intelligent wearables and specifically smart glasses are evolving into personal AI devices that can connect the user directly to an AI agent or model. This emerging category is growing at a remarkable pace and has reached an inflection point fueled by very strong demand for smart glasses from Meta. This quarter alone, Meta introduced several new Snapdragon power styles, including the Ray-Ban meta second-generation glasses, the Oakley Meta Vanguard performance classes in the Meta Ray-Ban Display and neural band.

In addition to Meta, our leadership in this space is reflected by the 30 designs in production or development with our global partners. They include Samsung, which recently launched Galaxy XR, a truly multimode AI headset in the first device for Google's new AI native operating system, Android XR. We achieved a significant milestone in automotive with the launch of Snapdragon Ride pilot, our first full system solution for L2+ automated driving, developing close collaboration with BMW, a debuted in the automakers BMW iX3 EV SUV. Powered by our advanced self-driving software stack, Snapdragon Ride Pilot sets a new standard in automated driving. It's designed for universal compatibility and seamless integration with automakers unlocking L2+ driver assistant features like hands-free highway driving and urban navigation for vehicles worldwide. Snapdragon Ride Pilot is currently validated in 60 countries and extends to 100 in 2026. The broad interest from leading automakers globally is exceeding our expectations.

At IAA Mobility, Qualcomm and Google announced an expanded partnership, including the integration of Google Gemini models to our suite of Snapdragon digital chassis solutions. Together, we will enable automakers to build and deploy personalized AI agents to act as an in-vehicle assistance, bringing multimode edge to cloud AI to next-generation software-defined vehicles. In industrial IoT, we completed our acquisition of Arduino, a premier open source hardware and software company with an IoT development ecosystem of more than 30 million users worldwide. This builds on our acquisitions of Edge in Pulse and foundries I/O and accelerate our plans to provide a comprehensive edge development platform for a broad set of applications. With these new assets, we're expanding our portfolio to a wide range of customers and verticals, further cementing our position as a leader of AI for the edge. Additionally, we recently released the Arduino no single board computer, powered by a Dragonwing processor. This full stack edge AI platform enables the rapid development of solutions for applications ranging from smart home automation to industrial robotics, drones and more. AI data center growth is moving from training to dedicated inference workloads, and this trend is expected to accelerate in the coming years.

The mass adoption and continuous use of AI applications is driving the industry to look for competitive alternatives that prioritize power-efficient performance and cost. We announced our entry into this market and recently unveiled our AI inference optimized AI 200 and AI 250 SoCs and associated accelerator cards and racks. We are very pleased to have you Main as our first customer for these solutions with a target deployment of 200 megawatts starting in 2026. Looking ahead, we're executing on a multi-generation road map with an annual cadence I would like to share that we're looking forward to providing an update in the first half of 2026 on our data center plans, including our road map performance and differentiated memory and compute technology. We will also highlight our progress in other areas, including advanced robotics, next-generation ADAS, industrial edge AI and 60 devices and AI-powered RAN. As we execute on our strategy and expand our IP and capabilities, we believe we are 1 of the best positioned companies to lead the expansion of AI to the edge to cloud hybrid AI and develop a power-efficient cloud inferencing solution. I will now turn the call over to Akash.

Thank you, Cristiano, and good afternoon, everyone. Let me begin with our fourth fiscal quarter results. We are pleased with our strong non-GAAP performance with revenues of $11.3 billion and EPS of $3, both of which were above the high end of our guidance. QTL revenues of $1.4 billion and EBT margin of 72% were above the midpoint of our guidance driven by slightly higher handset units. QCT delivered revenues of $9.8 billion and EBT of $2.9 billion, with year-over-year growth of 13% and 17%, respectively. QCT EBT margin of 29% was at the high end of our guidance. QCT handset revenues of $7 billion increased 14% on a year-over-year basis. reflecting increased demand for premium Android handsets powered by our Snapdragon 8 Elite Gen 5 platform. QCT IoT revenues of $1.8 billion grew 7% year-over-year, driven by strength across industrial and networking products and increased demand for AI smart classes powered by our Snapdragon platform. In QCT Automotive, we surpassed $1 billion quarterly revenue milestone, delivering 17% year-over-year revenue growth as the adoption of our Snapdragon digital chassis platform continues to accelerate. With the recent enactment of the One Big Beautiful tax bill, we now expect our non-GAAP tax rate to remain in the 13% to 14% range going forward, and we anticipate lower cash tax payments relative to prior expectations.

This new legislation resulted in a noncash charge of $5.7 billion in the fourth fiscal quarter to reduce the value of our deferred tax assets. This charge is excluded from non-GAAP metrics, but impacts our GAAP results. Before turning to guidance, I'd like to take a moment to highlight our strong performance in fiscal '25. We are incredibly pleased with our execution with non-GAAP revenues of $44 billion and EPS of $12.03 representing year-over-year growth of 13% and 18%, respectively. In QCT, we achieved 16% year-over-year revenue growth driven by double-digit increases across all revenue streams with IoT up 22% and automotive growing 36%. We also delivered QCT operating margins of 30%, in line with our long-term target we have previously outlined. Over the past 5 years, our non-Apple QCT revenues grew at a 15% compounded annual growth rate. Similarly, over the last 2 years, our non-Apple QCT revenues grew by 17% and 18%, respectively.

Lastly, we generated record free cash flow of $12.8 billion. And consistent with our commitment, we returned nearly 100% to stockholders through repurchases and dividends through the year. Now turning to guidance. In the first fiscal quarter, we expect to deliver record results with revenues in the range of $11.8 billion to $12.6 billion and non-GAAP EPS of $3.30 to $3.50. In QTL, we estimate revenues of $1.4 billion to $1.6 billion and EBT margins of 74% to 78%. In QCT, we expect record revenues of $10.3 billion to $10.9 billion and EBT margins of 30% to 32%. We anticipate record QCT handset revenues with low teens percentage growth sequentially, primarily driven by new flagship Android handset launches powered by Snapdragon. Following our outperformance for QCT IoT revenues in the fourth quarter, we expect a sequential decline consistent with last year, driven by seasonality in consumer products. In QCT Automotive, following a record fourth quarter, we estimate revenues in the first fiscal quarter to remain flat to slightly up on a sequential basis.

Lastly, we forecast non-GAAP operating expenses to be approximately $2.45 billion in the quarter. In closing, as we approach 1 year since outlining our growth strategy at Investor Day, I'd like to provide an update on the progress towards our $22 billion fiscal 2019 revenue target across automotive and IoT. In automotive, we've established ourselves as the most strategic silicon partner for OEMs globally. The accelerating adoption of our StapDragon digital chassis platform and 36% year-over-year revenue growth in fiscal '25, puts us on track to achieve our $8 billion revenue target. Across IoT, the increasing importance of artificial intelligence, high-performance, low-power computing and connectivity and validated by our 22% year-over-year revenue growth in fiscal '25 reinforces our confidence in achieving our $14 billion revenue target. In Industrial, increasing customer engagement and growth in design win pipeline, combined with our recent acquisitions to unlock access to 30 million users, underlines our confidence in strong revenue growth through the end of the decade.

In XR, we are exceeding prior expectations on strong demand for AI smart glasses, and we remain the platform of choice for smart glasses and mixed reality devices across leading global OEMs and ecosystems. In PCs, we extended our technology leadership with the recent launch of Snapdragon X2 Elite and X2 Elite extreme platforms, which deliver multigenerational performance increases across CPU, GPU and AI. Given our strong pipeline of approximately 150 design wins, we're optimistic about the growth potential for Snapdragon powered AI PCs as we expand our presence across global consumer and enterprise channels. In networking, our continued innovation and leadership in WiFi, 5G, edge processing and AI combined with our integrated platform approach positions us to drive content growth and adoption globally.

As Cristiano outlined, beyond our revenue target, we're also pursuing incremental opportunities across data center and robotics. Finally, I want to thank our employees for exceptional execution and continuing to deliver industry-leading technologies and products. This concludes our prepared remarks. Back to you, Mauricio.

Thank you, Akash. Operator, we are now ready for questions.

[Operator Instructions] First question, which will come from the line of Joshua Buchalter with TD Cowen.

Congrats on a stellar set of results in a bumpy backdrop. I wanted to start with the data center business. I realize you're going to provide more details in the first half of 2026 and my questions might get pumped as a result. But -- maybe you could spend a few minutes talking about what you see as Qualcomm's right to win in the data center space? And any details you can provide on the specs of the 200 and 250 beyond what you were able to offer in the press release when the Humana engagement was announced. And then lastly on this topic. Last quarter, you called out a, I believe, a hyperscale engagement. I assume that's distinct from the HUMAIN engagement and any details on timing there?

Josh, thanks for your question. And thank you. Yes. Look, we're very excited. I think this is the next chapter of I think the process we have been in Qualcomm to changing the company, diversifying the company, spending our IP. I think that's 1 of the reasons I think we made acquisitions such as Alphawave. We think there are 2 areas that we outlined that we can participate into the data center. We are incredibly excited about the size of the opportunity in the next phase I think, of data center build-out, where there's going to be real competition, we go from training to inference. We have been focused in 2 areas. One is we believe we have one very strategic asset in the industry, which is a very competitive power efficient CPU that is both for the head node of AI clusters as well as general purpose compute.

And then we also have been building what we think is a new architecture dedicated for inference. I think the focus has been increased computer density and and simplify the architecture for the data center in terms of increase, I think, performance per watt. I think it's all going to be about generating the most amount of tokens with the least amount of power, and that's our right to play. We're excited about what we're doing. That has been in development. It's something that we're actually doing in a very disciplined manner. We spend a lot of time. I think with our early experimentation with AI 100 to develop the software and then we're now building AI 200, 250 both the SoC, the card direct solutions. And I think we're pleased with what we're seeing. We will provide more details on that as we outlined early next year.

Specific to your questions, I think we were in discussion with a hyperscaler. We're very pleased with the outcome of that conversation, and that's going to be part of our update when we provide details on the road map, the performance, the KPI we'll be able to show details of the solution as well as our customer engagement. We are in conversation with a lot of companies. It's clear the market wants competition for this. But in a typical Qualcomm way, we're just going to be focused on executing and show the products performing. Like I said, this is exciting in the new chapter of our expansion. And alongside robotics, those are kind of new opportunities for us.

Appreciate all the color there and looking forward to the update. For my follow-up, I wanted to ask about the handset market. So you highlighted your ongoing momentum in the Android space as driving growth in the fourth quarter -- or excuse me, calendar fourth quarter in your prepared remarks. There's been a lot of noise, I think, about at your lead Android customer potentially looking to use an internal modem more than they have in recent years. Could you maybe just talk about your visibility into your share at that customer? And any sort of share that you would expect to -- how that you would expect that over the next year or so?

Look, thanks for the question, I want to spend a little bit of time on this because I sense that there is potential for a lot of noise when noise is actually not required. I think first of all, there is one thing that is happening with our Snapdragon and our premium tier Snapdragon Android, which has been very consistent, and this is going to happen over the past few years. The premium tier is expanding. I think if we look at the overall market, we have this trend that is very healthy and the premium tier is expanding and is adding more compute. That is the reason why our Android business, even on a market that is relatively flat, which is a handset market. We continue to grow content, ASPs and earnings because we see premium tier expanding. A lot of the upside we have in the handsets is primarily driven by the Android premium tier.

Second part is our relationship with Samsung. We have said for a number of years -- a number of reasons. And that's been true in the past, I think, several years what used to be a normal relationship at a 50% share, the new baseline is about 75% share. And that is always going to be our financial assumption. When we out-execute, sometimes we get more than 75%, on Galaxy S25, we got 100%. Our assumption for any new Galaxy is always going to be 75%. That's our assumption for Galaxy S26.

The next question comes from the line of Samik Chatterjee with JPMorgan.

Cristiano, you mentioned the -- on the data center side, starting with that. You mentioned the price performance for the inferencing performance that you're trying to deliver. I mean most of the training clusters that we've sort of seen the other incumbents sort of talk about the ranges of installation cost is somewhere in the sort of $30 billion, $40 billion per gigawatt that we're hearing off. Can you just rightsize us in terms of when you're thinking about the deployment on a gigawatt basis, what kind of cost performance or price performance are you thinking of relative to these inferencing workloads that you can support on the AI 200 or AI 250. And I'm also trying to get to sort of what revenue implications are for HUMAIN when you sort of deploy 200 megawatts with them? And I have a follow-up, please.

Okay. I'm going to try to give as much color as I can without getting ahead of the update we're going to provide next year. So first, let's just have a broader discussion about revenue. before, what we said before that we expect data center products to start leading to a revenue ramp beginning in fiscal '28. I think as a result of the HUMAIN engagement and our progress on the AI accelerator, I think we're pulling this forward into fiscal '27. So you should expect now what we said before, I think data center revenue is going to start to become material in fiscal '27. So I think that's the extent of what I can provide at this moment is about a 1-year pull-in.

The second thing is we are getting interest. You should assume the companies that are having to deploy as much compute as they need in the data center for inference, especially now that you see the constraints you have on power, the constraints that you have on the amount of computer density I think we have a lot of folks interested. We will not have in conversations if we didn't have a solution that is competitive, but we will show the KPIs of the platform, I think, when we have a road map update earlier next year.

Okay. Okay. Got it. And maybe the second one, similar to Josh's question, I think, Akash, if I'm interpreting the market's reaction to your strong numbers, there seems to be that concern about what March looks like with the change in share at the primary Android customer. Typically, on the handset side, your quarter-over-quarter decline into March has been sort of this high single-digit pace. Is that still a good run rate with sort of the lower level of share? Or would you sort of guide us otherwise? Because I think that's really what the market seems to be sort of concerned about at this point.

Yes. Samik, thanks for the question. We're not guiding beyond first quarter at this point. But when you look at our strong business momentum exiting fiscal '25 you see the benefit of that showed up in our results also showing up in the December quarter guidance. And so that carries forward into the rest of the fiscal year. The only additional thing I'd note is just a reminder that we expect to close our Alphawave acquisition in the first calendar quarter of 26%. But otherwise, I think the business momentum is strong and just a couple of factors that you outlined.

The next question comes from the line of Timothy Arcuri with UBS.

Akash, when you talked about September, you said that the beat was driven mostly by premium Android, but it seemed like it came a little more from your top customer because before you were saying to take like 30% of units out and that was like $500 million roughly, but it seems like nowhere near that much came out from that customer. So I mean it was kind of barely down year-over-year. So can you just square that? And then also as part of that, can you speak to how much that customer is as kind of a baseline assumption for December. I think we've seen the model that has their modem and it's not really selling very well. So I would assume that that's a tailwind for you also in calendar Q4. And then I had a second question.

Sure, Tim. So as we had said earlier, we expect it to be in 3 of the 4 models of the phone that was launched. And so that is exactly what happened. And share, of course, is based on what -- how sell-through plays out. Specifically on the September quarter question, we already had kind of demand from the customer that was factored into the guidance we gave. So the upside we saw was not from Apple. It was really driven by Android customers and primarily premium tier with the launch of our new Snapdragon chip. When you look at the sequential trend as well, as I mentioned, the -- we are forecasting approximately low teens sequential revenue growth in the handset revenue stream for QCT and primarily driven by Android as well. So there is some benefit from Apple, but the primary driver for the growth quarter-over-quarter is actually Android premium tier shipments.

Okay. And then is there any update on the negotiation with Huawei for a license? It seems like it's kind of dragging on a little bit. Can you just talk about that?

Yes. This is Alex. Thanks for the question. No, we actually don't have an update now. Discussions are still underway. Really nothing substantive to say beyond that.

The next question comes from the line of Stacy Rasgon with Bernstein Research.

So you noted the non-Apple QCT revenue was up 18% year-over-year. And even if I take out the auto and the IoT, it's clear that the Android piece was up like pretty strong double digits year-over-year. So am I right in assuming that's all content or primarily content given, I don't think units grew that much? And is that the right sort of pace of like further content increase that we ought to be thinking about as we go forward?

Yes. So Stacy, you're doing -- obviously, doing the math right. There are 2 primary drivers on this. One is just the mix shift of units up. And so this is a trend that we've seen over the last several years. And it's -- sometimes it's thought of as a developed market trend, but that's not true. It's across all developing regions as well. The devices that are purchased continue to move up and so that shows up in the benefit to our revenue stream. The second trend is within premium tier. Content continues to grow as we deliver more and more capable chips and more capable handsets are being delivered as a result of it. And those are the 2 primary drivers of kind of the long-term trend of our handset business.

Got it. And if I could have a quick follow-up. Just the Snapdragon Android strength in September and December. Is that primarily China? Is there any concerns? I mean, is that just the timing of the launches? Like any thoughts on pull forward or anything like that, anything we ought to be thinking about there?

Yes. No, there's no pull forward there. I think what we've seen is all of our -- most of our China customers, actually all the major customers have already launched devices and the initial reception to the devices have been very positive. We'll see a lot of our global customer launch devices as well later this quarter going into early next year. And so it's just a reflection of kind of normal purchase patterns around the launch of these devices and the great initial consumer reaction to the launches.

The next question is from the line of Chris Caso with Wolfe Research.

And a question again on AI data center, and I realize you're going to provide some detail coming up, but there's some spend out there, so which is why we ask. But for we've seen what was in the press release was perhaps a different architecture than what we've seen others attack the market with DDR memory, PCI Express in that. Should we interpret that as sort of a first approach by Qualcomm with more to come? Or is this rather a different sort of philosophy for attacking the market? You talked about being more efficient on power consumption. Is this sort of a different -- attacking the market differently than what's in the market today.