Quanta Computer Aktie

Hello. Good afternoon, everyone, and thank you for joining us today for Quanta's second quarter results webcast. Quanta is a leading solution provider in notebook and cloud computing, and they also continue to innovate in the era of data economics, exploring new opportunities, including quantum computing as well as satellites.

My name is Howard Kao and I'm the coverage analyst here at Morgan Stanley. We are very honored to have Mr. CC Leung, Vice Chairman and President; Mr. Elton Yang, CFO and Spokesperson; as well as Carol and Alai from the IR team here with us today. And we look forward to their insights and comments on the company as well as the market.

Carol will first walk us through second quarter results and also provide some forward-looking commentary. And after that, we will open it up to questions. So now I would like to turn it over to Carol for opening remarks. Per, please.

Welcome to Quanta Comp Second Quarter 2025 Earnings Results Conference Call. This call is hosted by Morgan Stanley, and the presentation material for today's call is available to the public online and on Quanta's website at www.quantatw.com under IR section. This is Carol. I'm joined today by Quanta's management, Mr. C.C. Leung, our Vice Chairman and President; Mr. Elton Yang, Chief Financial Officer and Senior Vice President; and Quanta's IR team.

In a moment, we will start with the financial presentation. After that, the management will host a Q&A to share our view about the business and industry outlook.

During this call, we will be making forward-looking statements, which are predictions, projections and other statements about future events. These statements are based upon our current expectations and assumptions that are subject to lots of risks and uncertainties. As such, the actual results could differ materially from our forecast. We undertake no obligation to update or revise any forward-looking statements.

So now let's walk through the financials. Revenue reached a record of TWD 504 billion in the quarter, up 3.8% quarter-on-quarter and 62.6% year-over-year even as the sharp [ pricing ] Taiwan dollars has weighted on the top line by eroding U.S. denominated revenue. Despite the ForEx headwinds, we maintained robust year-over-year growth, anchored by elevating bottlenecks in high-end AI server rack production as well as stronger-than-expected notebook shipments. We shipped 12.1 million units of notebook in the quarter, up 12% quarter-on-quarter and 3.4% year-over-year, outpacing our previous forecast of high single-digit sequential growth.

That outperformance was driven by a pull forward in demand as customers accelerated orders to mitigate U.S. tariff-related price increases. This trend aligns with broader industry behavior observed since Q1, where OEMs globally have front-loaded notebook shipments, particularly into the U.S. as a hedge against expected cost pressure, resulting in 2 consecutive quarters of sequential shipment upside. The company's gross profit was TWD 35.5 billion, declined 7.7% quarter-on-quarter from the record of TWD 38.5 million in the first quarter and grew 33.5% year-over-year. Gross margin came in 7% for the quarter, representing 88 bps decrease quarter-on-quarter and 164 bps decline year-over-year.

We estimated that ForEx accounted for roughly 2/3 of that decline and the remaining margin dilution stem from product mix. While servers maintained a steady 65% to 70% revenue share, our mix shift toward higher ASP AI rack. AI sales continued to grow with higher ASP units scaling up and offsetting the decline of lower ASP models, though this mix shift put downward pressure on margin rate. Operating profit totaled TWD 20.4 billion this quarter, the record for the same period and the second highest level in the company's history. That reflects a decline of 17% from the record TWD 24.6 billion achieved in the prior quarter, yet an impressive 34.2% gain from a year ago.

The operating margin for the quarter was 4%, which declined by 102 bps quarter-on-quarter and 86 bps year-over-year, respectively. Operating expenses were TWD 15.1 billion for the quarter, grew from TWD 13.9 billion last quarter or 9% sequential increase. On a year-over-year basis, OpEx increased by 33.4%, primarily due to R&D expenses associated with AI development, continued talent recruitment and increased employee compensation, supported by strong sales momentum of 62.6% growth year-over-year. OpEx ratio stayed at 3%, reflecting that the OpEx increases remain disciplined and efficient to drive growth.

Nonoperating income for second quarter was TWD 1.2 billion. The primary items included net ForEx gain of TWD 668 million and net interest expenses of TWD 213 million. The remaining other income of TWD 759 million mainly came from equipment disposals. Stepping down to the income statement. The net income after tax came at TWD 16.9 billion, decreased by 13.5% quarter-on-quarter and grew 11.5% year-over-year. The net margin after tax was 3.3%, which declined 67 bps quarter-on-quarter and 155 bps year-over-year, respectively. The tax rate for the quarter was 21%.

All of the above concluded earnings per share of TWD 4.34 for the quarter, representing a sequential decline of TWD 0.69, but an increase of TWD 0.45 from the same period last year.

Please flip to the next page. The first 6 months of 2025, we delivered record-breaking results. Our revenue, gross profit, operating profit and net profit both hit the all-time high. Notebook shipment reached 22.9 million units, up from 22.2 million units a year ago, representing an increase of 3.2% year-on-year. Revenue for first half came in at TWD 990 billion, grew by 74% year-over-year. Gross profit for first half reached its record of TWD 74 billion, grew from TWD 48.6 million in the same period last year, implying 52.4% up year-over-year.

Operating profit was TWD 45 billion, showing a surge of 67.3% year-on-year and the net income after tax was TWD 36.4 billion, which grew 33.7% year-on-year and EPS for first half was TWD 9.43. In terms of margin rate, gross margin was 7.5%, declined by 106 bps year-over-year. Operating margin was 4.5%, drop of 18 bps year-on-year, and net margin after tax was 3.7%.

Turning to the balance sheet. Cash and short-term investments stood at TWD 194 billion, decreased from TWD 215 billion at the end of last quarter. Account receivables reached TWD 395 billion, down from the record level of TWD 451 billion in the previous quarter. Inventories increased further this quarter, reaching a record TWD 286 billion. The rise was primarily driven by ongoing AI server shipments and the preparation of new generation AI project development.

The shareholders' equity attributed to the parent company was around TWD 188 billion, decreased from TWD 193 billion at the end of the previous quarter. During our last financial report, we announced an increase of full year CapEx budget to TWD 20 billion, while the CapEx for first half totaled TWD 7.6 billion in cash basis. We fully expect a significant acceleration in the second half to meet the planned target. This anticipated increase reflects our commitment to strategic growth and the continued execution of our investment road map.

We are pleased to announce that our Board of Directors today approved several significant investment projects. This includes a capital injection of USD 170 million into our 100% owned U.S. subsidiary, QSM, reinforcing its operational capabilities and accelerating AI server capacity expansion in our U.S. site. To address tariff-related cost impact on AI server production, our Mexico facility will kick off its expansion later this quarter upon customers' requirements. The infrastructure and facilities are adjusted as needed and scaling up for AI server manufacturing. This adds extra flexibility to support our major final assembly operations in the West Coast and the middle of the U.S. alongside our existing AI production facilities in Taiwan and Thailand.

We expect that the AI production in Mexico will start pilot run at the end of this year and begin mass production in early 2026. Additionally, the Board has approved a USD 50 million investment in Quantinuum Inc., a U.S.-based company specializing in quantum computing technologies. This strategic move aligns with our long-term objective to explore and invest in next-generation computing innovation. By engaging with industry pioneers at an early stage, we aim to position ourselves to capture emerging opportunities as the commercial potential of quantum computing evolves.

Turning to our server business. The supply chain has successfully achieved commercial ramp-up for the GB200 platform in the quarter. Looking ahead, the next-generation platform is just around the corner and is set to enter the market soon. For third quarter, we expect AI server sales to maintain a similar pace as Q2. This is primarily driven by the initial small volume shipments of the new GB300 platform, which is expected to start towards the end of the third quarter.

Initial feedback from the supply chain on the new generation has been positive. suggesting that the ramping process from the pilot production to mass production for GB300 is likely to be smoother than with the previous generation. This has prompted some of our clients to adjust their project plan, slightly slowing down their intake of the prior generation products. These customers are keen to move quickly to the new generation of servers to take advantage of the higher computing density and improve the capital expenditure efficiency that the new platform offers. We believe this quick migration is a healthy sign of a strong sustained AI demand for the adoption of the latest technology and set to a solid foundation for accelerating AI server business in the coming quarters.

On the other hand, momentum in general compute servers softened in Q3 after a strong shipment in the first half. Overall, general compute servers are returning to healthy growth for the full year, reflecting a recovery in demand compared to last year. In the notebook segment, we are seeing stable shipments in the third quarter, slightly ahead of what we expected a couple of months ago, following better than seasonal demand in both the first and second quarters, we now anticipate low single-digit sequential growth in the third quarter. However, we expect a softer fourth quarter as some demand was pulled forward earlier in the year. Lastly, we are maintaining our full year total shipment guidance of single-digit year-over-year growth.

Turning to our auto business. We shift our focus and prioritized resources towards higher value and content-rich products like advanced ADAS, car computers and inverters. As a result of this strategic pivot, auto sales slowed down year-over-year in first half. However, margins remain comfortably above the corporate average. With the AI server outgrowing, auto contribution will be dilutive to low single digit of total sales for the year.

In the third quarter, we expect to see some pressure on our margin. This is primarily driven by a shift in our revenue mix with major shipment momentum from higher ASP AI server models, offsetting a more moderate sales contribution from other product lines, including general compute servers, notebooks and auto-related products. On top of that, we expect some residual margin pressure from second quarter's ForEx volatility to carry into third quarter, primarily resulted from the material [ pooling income ]. Forex, FX, often flow through the financials with a delay, causing a lag impact due to timing differences between revenue and cost recognition.

In terms of operating expenses, in view of new customer acquisition and more project wins, we anticipate ongoing increases in our R&D expenditure as we continue to invest in the upfront cost of new project development. During recent earnings calls, hyperscaler customers have consistently reaffirmed their commitment to continued investment in AI. They are focused on solidifying their growth in both AI infrastructure and application market. This is a high step to competition, one with [indiscernible] ground fight but still nonetheless, centered on the computing power, talent, technology, capital and energy resources.

Our demonstrated strength and track record have earned us the trust of more customers and secure new projects, providing us with a critical opportunity to widen our leadership position in AI server market. Experienced and high-skilled R&D engineers in AI server design and manufacturing are a scarce resource globally. The competition for AI talent is intensifying, and we believe our people are the ultimate key to our success. Over the past few years, we've been steadily strengthening our R&D pipeline for AI, a critical pillar of our long-term strategy.

We are working on multiple fronts to deepen our talent bench, retain our top performers and attract the best people in the industry. This includes expanding recruitment efforts, significantly enhancing compensation, benefits and bonus program and optimizing resources allocation through internal mobility. This initiative has not only boosted our competitiveness in the talent market, but also help us maintain a high retention rate among our most critical R&D personnel.

We are seeing a lot of promising opportunities for growth, but that also means we need to tackle some internal and external challenges and make sure we have the right resources in place to support that growth. As our business continue to scale, we are seeing a corresponding increase in our working capital needs to support it. To proactively maintain strong financial flexibility and a healthy balance sheet, our Board of Directors has approved USD 1 billion ECB issuance and a refinancing program of USD 1.5 billion syndication loan today. The strategic combination allows us to fund our robust growth while minimizing shareholder dilution and capitalizing on favorable market conditions.

Following the successful distribution of a record high of TWD 50.2 billion in cash dividends or TWD 13 per share in July. We remain committed to providing long-term sustainable and above-market shareholders' returns. Despite an external environment that continues to present challenges from U.S. tariff policies and currency fluctuation to geopolitical factors, our management team has delivered on its commitments. The solid performance of the second quarter results is a powerful testament to our operational excellence and the resilience of our business model. We will build on this momentum and continue to deliver superior long-term value for our shareholders.

And that concludes the review of the second quarter results and business outlook.

[Operator Instructions] Now may we introduce the first question, please?

Yes. Thank you, Carol, for your prepared remarks. [Operator Instructions] Before we get to the questions in queue, Carol, can I quickly ask you, maybe there's some audio issues on my side, but your prepared remarks on AI server comments, can you please run through them again?

Repeat all or...

Just on the AI server part in your prepared remarks.

The overall AI server shipments will remain at a similar path as it was in the second quarter as there will be some project migration from prior generation to the new generation. And the pilot ramp of the new generation is expected to start at the end of third quarter.

Got it. And so before we get the questions in the queue, just 2 questions from me. The first question is on AI servers. I think last quarter, Elton, maybe you mentioned that there is some supply constraint in terms of GPUs for these GB200 racks. Is this still the case? Is that the reason why maybe some of your competitors are seeing stronger revenue momentum while you are not? Can you just kind of touch a little bit on that versus what you saw 3 months ago?

Thanks, Howard. Okay. That's -- first of all, it's not exactly a good comparison between the key supplier for ODM for the GB200, 300 because different customer mix and product mix. As we touched on in the prepared remarks, the GB200, the yield gradually ease. So it's getting smoothly worse. But during the time, customers try to enter into a transition into the next product cycle, the GB300. So different customers different adoption curve and schedules. So some of them are shifted gear to access the adoption of the next-generation products. But anyhow, we still see the demand is very strong. But different customers different schedules, just kind of mixed ones.

In terms of the different ODMs, some of them have different customer mix, so more comfort customer products, some different product mix is more comfort [indiscernible] so different profile. So kind of hard to do peer-on-peer -- head-on comparison. But in terms of year-on-year, we still see over 20% growth year-on-year. So we still exciting about the evolvement. And also exciting about the customer migration progress and we're still excited to see new emerging customers to exciting about our performance in the GB200. So they are more than happy to work with Quanta. So we'll see more new project wins and emerging customer to work with. [ That's a ] more color about the evolvement for our AI business.

Yes. Very helpful. Elton, one follow-up from me. Within your AI server, especially these GB rack projects, I think previously, you have mentioned that most of your projects are still buy and sell where your ASPs will include the price of the GPUs. Is that still the case? Or have some of these projects now transitioned to more of a consignment model where the GPUs are not included in the final ASP?

Almost our customer business model still on the buy and sell model, a very few projects on consignment basis. So still keep on the buy and sell model for now on and keep going on that kind of business model.

Got it. And also...

Sorry, Howard, just wanted to explain why the margin to be [ diluted ]. We wait.

So now we'll move to questions on the queue. The first question will be [ Tim-Sin Lee ].

So first of all, I think congrats on the solid quarter. I'm also asking on behalf of Randy. So I got a question and a follow-up. I guess related to AI service in July, do we see the July softness in sales mainly due to transitional period of AI service? Or do we see other reasons for this?

Okay. Yes. As we touched on in our prepared remarks, we see the customer actually has entered into the transition migration. Some of them moved migration phases into next generation. So that's what explain customer different agenda and the schedule production curve. So that explain what month-on-month and we expected by year-on-year still double 20% over growth rate. So that's why I'm trying to explain our business profile, customer profile and the product profile is different with others. So it cannot be directly in comparison, okay.

Okay. It's good color. Could we expect that July and August to be more of a transitional period ramping up the next-gen platform and in September. I guess how should we think about the ramp and the throughput there?

As we mentioned, they were starting from the end of third quarter and started more shipment out from the early fourth quarter. So the revenue will gradually [ accelerate ] progressively quarter-on-quarter.

The next question will be [ Kai Huang ].

My first question is about the margin profile. As we have mentioned in the previous meeting, we mentioned that maybe the higher percentage of high ASP or AI server may have a dilutional effect on our gross margin. But my question is about whether or not this trend will continue into the future, since some of our competitor has mentioned that we may have some ability in the corresponding component of AI server, which may yield a better gross margin in the future. So my question is about the margin profile about the impact from AI or high ASP AI server in the future.

Yes. Thanks for your question. By nature -- by principle, when we enter into similar high ASP product by nature the margin rate should be dilutive, okay? But on the other hand, to lower down our total cost is our mandate. So we always keep moving on to provide a lower TCO product for the customer and for the product as well. That's why our migration have more automation for our product, okay? So we have our affiliate to provide kind of a robotic arm to help us. So that's a twofold and the mix one. By natural, we migrate to next high ASP product, it should be dilutive. But our mandate, our regular cost, to lower down the production cost. And again, we get the learning curve for our GB200, we are expanding to our GB300 yield rate. So it also helps the cost as well. I would say it's a mix one.

Yes. Understood. I appreciate that. And my follow-up is about the overall server revenue forecast for this year. As you have mentioned, we still expect triple-digit growth from AI server this year. But what I'm curious about is that since the revenue seems to be like flat quarter-over-quarter in the third quarter for the AI server, you had mentioned about some like transition problems. Do you still see the same size of revenue contribution from AI server as you see in the -- at the beginning of this year? Or is there any difference from you see in the first quarter in this year?

Okay. By far it's still on track. We still target for that. [indiscernible] Kind of about the allocation, the distribution between the AI server and non-AI server. Okay. So we still target that and still aim for that. And by far, it's still on track.

Next question will be Albert Hung.

This is Albert from JPMorgan. So in the opening remarks, I think Carol mentioned that third quarter AI revenue growth will be similar to second quarter. But could you help us understand more what kind of like second quarter AI server revenue growth was in like, say, second quarter? How much is like AI server revenue in percentage in the second quarter? And a follow-up on that is, is it fair to say GB200 still have a lot of like yield issue? And in that case, why customers decide to migrate to GB300, while there's a lot of like GB200 backlog?

Okay. Yes, you may be led to the earnings call. As Carol touched on, the yield issue is eased off. But the issue is the GB300 is around the corner. So customer -- now is kind of migration phases into that, okay? So it's not a yield issue. It's kind of customers' production schedule issues. And by the way, we don't provide that kind of granularity by quarterly about AI server contribution. But our still target AI server kind of total server about 70% for a yearly basis, still on track.

Could you help us -- maybe it will be better if you can help us understand how much is the GB200 order backlog at the moment? Because, I mean, it looks like when we transit into like GB300, there's still a lot of like strong demand for the legacy product. So is it fair to say even when we talk about like slower transition into like third quarter, we are going to expect a very big jump in fourth quarter?

Okay. Unlike notebook, okay, in the server business, every CSP for the new emerging customers, they keep a multi-platform in the data center. So they have a H, they have a B, they have a GB200 series, the 300. So they're moving to GB300 doesn't mean GB200 is gone. So what I'm trying to touch on is some customers, some -- not every customer moving to GB300. So different customers, different agenda, depending on how their application and their agenda for their data centers. So we see some customers -- some new customers also target on the GB200. Some customers looking for GB300. Again, it's kind of blended one for us. So we still see both of them going on forward. And our view is for the AI server, still very strong.

The next question will be Alex Wang.

This is Alex from Bernstein. Just want to confirm what I've heard. So do you mean that for the third quarter, you expect your AI server revenue will be the same as the second quarter? Or do you expect there's like double-digit growth as in the second quarter?

We touched on about the [indiscernible] top line for everything in total, okay. Again, during the third quarter, it's kind of mixed. So some of the customer shift gear to move to GB300. So kind of the AI -- total server contribution kind of slightly increased again in the second quarter.

I see the revenue mix will increase quarter-on-quarter.

I mean AI server percentage will slightly improve during the third quarter. And again, we target to maintain 70% AI server to overall servers.

I see. That's very clear. And second question is about the gross margin. I think Carol mentioned that in the second quarter, there is some FX impact. How much is that? Can you remind us? And if we move from GB200 to GB300, is there further margin dilution between these 2 platforms?

Okay. In the second quarter, about 2/3 is coming from the FX impact. Let me kind of elaborate about that. It depends on when you acquire the material. So it could be 2 months or 3 months ahead due to the tightness of the supply or customer required to acquire. So it's a different FX profile, okay? So that had an impact about our gross margin, okay? When they migrate to the third quarter, when you move to a high ASP GB300, definitely, that by nature, there was dilution about gross margins. And it depends on how you're digesting currently material, which acquired 2 or 3 months ahead, there probably still some FX impact, but maybe not significant in second quarter, but still some minor impact in the third quarter. We just touched on about in the prepared remarks.

I see. And without FX impact, let's just consider the gross margin for GB200 and GB300. Do we see more margin dilution to that front?

Yes. As I mentioned, high ASP product by nature have margin rate dilution, not margin dollars, margin rate.

The next question will be on Chen.

So what is the Quanta's progress in AI ASIC servers? And is there any guidance from the company on how the current sales contribution of ASIC AI server? And what would it be in 2026. And the first.

Sorry, go ahead, sorry.

Sorry, so what is the major customer for AI ASIC server for the company? And in terms of the sales contribution, what -- how can we evaluate that? And is it only also in the buy and sell model.

Firstly, we don't -- we still stick to the buy and sell model for all the AI business -- of server business firstly, okay. Secondly, we do have a ASIC project out here now, okay? But in the past, due to the high bandwidth, we cannot fulfill customer requirement. But as Carol touched on in our prepared remarks. As our complete R&D bench and our retention and recruiting our R&D, we are inherent, undisputed of current SP customers to ASIC server provider because we know customer well. We know how to meet demand, and we are then new projects, and we're going to wrap in 2026, okay. We don't provide granularity for the speed and the customer particularly for the AI ASIC servers, but we're happy to see the AI ASIC server to win and to land as we have more R&D resources.

Okay. And just a quick follow-up. Can you give us a recap of the sales contribution of AI server in the first half of 2025.

So far, over 60.

Okay. Great, Chi-Yuan. Elton, maybe just one question for me before I go back to the queue. I think last quarter, you mentioned HGX systems, these lower-priced AI server order momentum, you still see it growing every single quarter throughout the rest of the year. Is that still the case from what you're seeing in terms of the order forecast? Or has that changed within the past 3 months.

Okay. Firstly, I think we never touched on about specific amount SGX split and kind of momentum. What I'm going to say as long as -- let me say one thing, So as long as the AI industry keeps growing, every sector, including the GB200, 300 high ASP server, AI server or [indiscernible] versions of server or ASIC server, they are all growing, but didn't go through a growth trajectory and AI GB20 high ASP server, AI server should enjoy the highest growth rate, but the risk to keep growing.

Okay. Whether we can have -- whether we should lend more order just clearly because our R&D resources, once we fix our R&D resources, we just have to fulfill all the customers' requirement, but every CF customer, they have their own design, they have different models for the AMD model, for the NVIDIA model, for ASIC model, S86 model, they have unfulfilled demand for us. So we keep seeing the strong demand and we can [indiscernible] grow our business opportunities even from the same CSPs some new emerging plants. So we are still exciting over AI demand going forward.

Got it. Very clear. And the next person in the queue [ Tim-Sin Lee ].

It's Tim-Sin from UBS again. I guess kind of curious on for the high-end AI server form factors. We have 36 and 72 and also the 1 GBU form factor. I guess going into GB300, should we be expecting more of the 72 form factor taking the larger mix? Or I guess, how should we be thinking about this?

Yes. Moving to the next generation one where both 32 -- sorry, 72 [indiscernible] worse. And when migration to the next one, they were double that one as well, okay. So we are seeing normal 72 one is the mainstream product. And moving forward, the 144 maybe next one -- for the next one mainstream products.

Okay. My follow-up on a different question, maybe on Section 232 -- 232. I know that Carol touched upon our CapEx plans in the U.S., but could you touch upon your sort of more detailed view the importing components into the U.S. assembly, on tariff exposure or potential semi tariffs. Do you see like signs of exemptions? Or yes, what is our [indiscernible] strategy into doing this?

Again, the major CapEx will be more focused on par Mexico and the U.S. and we will keep exploring our process, our capacity, in the state onwards.

Okay. Great. Elton, so just a follow-up on the capacity question. On Mexico, if I remember correctly, in the past, it was only automotive that was produced there, but you guys are now investing and expanding, I guess, server and AI server production capacity there that will ramp up sometime next year in terms of mass production timing. Is it L6 that you guys are expanding in Mexico? Or is it rack-level assembly? Can you give us a bit more color on that?

I don't want to explore more, but more than L6, let me put this way.

And how do you ...

You need to make sure you fulfill USMCA. So highly unlikely only L6.

Right. Right. Okay. So is it -- is there a possibility whereby a very, very large portion of your server production will shift from the U.S. to Mexico to -- because of USMCA compliance?

So that will depend on how the evolvement of the 232 investigation, reciprocal tariff and customer commitment as well. So that's why we are -- we have over 9 production sites to provide a flexibility to work with the customers. So what the best way to get to -- to get the lowest tariff in the customer's intention. And when you go along with that, the key thing is the customer's decision, and we have a flexibility and resilience and the speed to help them to meet the target. So I cannot tell you exactly what will happen, but we were subject to how the Trump administration decision.

Got it. Very clear. And a question on yield rate. I think you guys mentioned in your prepared remarks and also in a bunch of the Q&As that GB200, you have said your rates are improving. Is there any additional color you can provide in terms of what the yield rates for GB300 looks like at this point in time versus 200. Is there a meaningful improvement? And are you expecting GB300 to ramp up much more smoothly versus GB200 a couple of months ago?

Okay. It's a different design from GB200 to GB300 and as the learning curve of GB200, we are being appreciated by our customer as well as GB vendor of our performance with the GB200 also help the customer to improve. So by far, for the GB300 ramping, the yield rate learning curve is better than GB200.

Got it. Very clear. And we have a question on the line from Claire Wang. Claire, please go ahead..

I just want to simply understand that we have a lot of CapEx going on, but could management update the investors like the shareholder returns, maybe in terms of the dividend payout ratio or dividend yield.

Okay. It's kind -- kind of always -- to make the decision between the page for meaningful investment and investment in high growth yield and high return to our shareholders. Based on our experience, we almost provide over 80% in payout ratio. And even nothing changed, we try -- we target to maintain that percentage.

Albert, please go ahead.

Yes. So for the margin side, Carol mentioned that 2/3 of the sequential gross margin decline was due to FX, right, sequential. So if I add that, say, 0.6% back and assuming that normalized gross margin is, say, 7.6%. And the gross profit dollar actually remained just flattish quarter-on-quarter. So my question is when notebook is growing like 12% quarter-on-quarter, general server, as you say, there is some point demand and AI sprinting up and why the gross profit dollar only remained flattish quarter-on-quarter. Was there any like other one-off expense, not just CapEx, maybe ramp up costs, et cetera. Could you help us understand more about that?

Okay. You know, firstly, for [indiscernible] GB200, the margin rate [indiscernible]. So not only FX on the product mix as well. We keep emphasizing again and again this other thing is a different product, product mix profile, the mix is different. And I appreciate your math. But anyhow, that's a reality. When we migrate to a high ASP product by nature, there is a dilution about gross margins, definitely, okay. I just want to give you more color about how the other impact from FX.

Got you. And second follow up is, if I look at inventory, actually, our inventory will control very, say, we were well under control [ ASIC ] in second quarter, it didn't increase that much. So my question is usually inventories are leading to cater for the, say, sales, right? So if like AI demand is so strong, why shouldn't we see the big jump in the inventory to existing second quarter? And also, if the inventory can sustain at current level, then why we need more like working capital, say, help from the ECP or syndicated loan?

First of all, syndication refinancing, okay. And to have the ECB, they were 2-fold. First to have more financial flexibility. Secondly, to cap the low interest rate with low dilutions. This kind of financial tool, financial department, we need looking for the best financial vehicle to funding our needs. In terms of inventory, depending on when the input yield you will have. So some of them are already in our warehouse. So that's why they have a double impact.

When you're early in your warehouse, you have an FX impact, but you also secure your orders. Okay. So that's -- we move to the next one. Okay. The only -- some of them already are in our warehouse. So we don't see significant grow about the inventory. So that's a dynamic, okay.

Maybe one last question from Kai Huang.

And my final question is the notebook business. Could you please share with us the percentage of contribution from notebook in the second quarter? And also, what is the dynamics for this segment in this year, like from Chromebook, the gaming or commercial? And what is the outlook for this business for 2026 to go.

Notebook now account about 20 to 25% of total revenue. And Chromebook in the second half were back indiscernible] represent percentage as regular cost, okay? And we still have a console [indiscernible] high, okay, in console. So back to the normal -- in the second half, back to the normal distributes between the regular notebook and Chromebook.

Yes. Understood. So do we expect like a 50 to 50 distribution of notebook business for this year or still some higher demand for the second half?

I will say 50-50 split between the first half and second half.

Is that for like revenue base or like a shipment-based to you.

Shipment-based. Okay. In our notebook, we see a different customers, different profile, different price profile. But anyhow, the shipment will be more a proxy to evaluate that. Otherwise, we fully increase margin dilution issue as well.

I think in the interest of time, we are getting closer to one hour mark. So maybe let me turn the call back to Elton or CC to see if you guys have any closing remarks.

Okay. This is Elton. In summary, our better quarterly results reflect the momentum we are building across our business. While we are facing some headwinds from dynamic macro regulatory environment, we believe they are more than offset by the powerful tailwinds from our leadership product portfolio and resilient manufacturing capability. And the outstanding performance shows our grit and determination by quick adoption to a dynamic market, focus on what we can control, extending our business strategy into those high growth opportunities, and we are exciting about that. Thank you.

Perfect. Thank you, Elton. Thank you, CC. Thank you, Carol, and Alai, for your time, and thank you, everyone, for joining us. We will see you all next time. Thank you.