Quanta Computer Inc. (2382) Earnings Call Transcript & Summary

Thank you all for joining us today for Quanta's First Quarter 2025 Earnings Conference Call. My name is Carrie Liu. I'm the research analyst covering Quanta at Citi and will be moderating today's call. Joining us on the call today are Mr. C.C. Leung, Vice Chairman and President; Mr. Elton Yang, CFO and Sokesperson; as well as Carol and Alai from the IR team. We will begin with the prepared remarks and financial overview from Carol, followed by a Q&A session. Without further ado, I will now turn it over to Carol. Carol, please go ahead.

Thanks, Carrie. Good day, and welcome to Quanta Computer First Quarter 2025 Earnings Results Conference Call. The call is hosted by Citi Global Markets Taiwan Securities. 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. We delivered another record-breaking quarter in first quarter '25 with sales, gross profit, operating profit and net profit reaching all-time high. This strong performance provides a solid foundation for the year ahead, which we anticipate will be shaped by both significant growth opportunities and evolving challenges. Revenue for the quarter was TWD 486 billion, up 16.4% quarter-on-quarter and 87.6% year-over-year, respectively. Thanks to slightly better supply conditions for the latest AI servers, along with stronger-than-expected demand for general compute servers and notebooks, we saw strong sales momentum throughout the quarter. Notebook shipment totaled 10.8 million units for the quarter, representing a 2.7% decline from the previous quarter but a 2.9% increase compared to the same period last year. This outcome surpassed our expectations of a double-digit sequential decline. We attribute the strong shipment to customers pulling forward demand at the end of the quarter due to the uncertainty surrounding U.S. tariffs. The company's gross profit reached a record TWD 38.5 billion, growing 23.8% quarter-on-quarter and 75.3% year-on-year. Gross margin for the quarter was 7.92%, representing 48 bps increase quarter-on-quarter and 55 bps decline year-over-year. The margin improvement was largely driven by strong demand for general compute servers and a lighter magnitude of margin dilution from high-end AI servers for slower ramp. Additionally, a reduced sales contribution from notebooks also helped margins. As servers outgrew other product categories, the sales contribution from servers reached 65% to 70%, while the sales shares of notebooks was further diluted to below 25% this quarter. Operating profit hit an all-time high of TWD 24.6 billion this quarter, representing a 62.3% increase sequentially. Year-over-year, profit more than doubled, up 110% from TWD 11.7 billion a year ago, making a strong TWD 12.9 billion increase. The operating margin of the quarter was an all-time high of 5.07%, up 143 bps quarter-on-quarter and 50 bps year-on-year. Operating expenses was TWD 13.9 billion for the quarter, went down from record level of TWD 15.9 billion in fourth quarter '24 or 12.8% sequential decrease. On a year-over-year basis, OpEx increased by 35.4%, primarily due to R&D expenses associated with AI development. Supported by strong sales momentum, OpEx ratio improved to 2.86%, down 96% quarter-on-quarter and 110 bps from a year ago, reflecting solid operating leverage. Nonoperating income for first quarter was TWD 669 million. The primary items include net ForEx gain of TWD 2.1 billion and net interest expenses of TWD 1 billion. Stepping down to the income statement. The net income after tax was a record TWD 19.5 billion, increased by 22.8% quarter-on-quarter and 61.6% year-over-year. The net margin after tax was 4.01%, which grew 21 bps quarter-on-quarter and declined 65 bps year-over-year respectively. The tax rate for the quarter was 22.1%. And all the above concluded an all-time high of earnings per share of TWD 5.06 for the quarter, representing a sequential increase of TWD 0.94 and a surge of TWD 1.93 from a year ago. Turning to the balance sheet. Cash and short-term investments slightly increased to TWD 215 billion from TWD 201 billion at the end of last quarter. Account receivables reached an all-time high of TWD 451 billion from TWD 371 billion in the previous quarter, reflecting the sales growth. Inventory rose to a record TWD 277 billion this quarter, mainly driven by customer request to pull in components early in preparation of accelerated AI server shipments in the coming quarter. The shareholders' equity attributed to the parent company was around TWD 193 billion, down from TWD 222 billion at the end of previous quarter due to the deduction of cash dividends of TWD 50.2 billion on an accrual basis. Let's now discuss the ongoing plans for responding to U.S. trade policy. As we highlight in previous calls, Quanta's business strategy is built on three key pillars: advanced R&D, precise marketing and manufacturing resilience. Among this, manufacturing resilience has proven to be a competitive advantage in today's volatile market environment. Quanta's manufacturing resilience is rooted in our extensive global manufacturing capabilities. We translate R&D breakthroughs into commercial products quickly and cost effectively, enabling customers to capitalize on emerging market opportunities. Our flexible and scalable production infrastructure allows us to adapt to shifting customer demand and evolving global trade dynamics. Our production network now spans nine major sites worldwide, one in Taiwan, two in China, Shanghai and Chongqing, two in Southeast Asia, Thailand and Vietnam, two in the U.S., Tennessee and California, one in Mexico and one in Germany. The global presence allows us to offer agile, regionally optimized manufacturing solutions. Specifically to address tariff-related challenges for AI production, our Mexico facility has the flexibility to realize Chinese expansion from focusing solely on manufacturing of automotive-related products to scaling up AI server manufacturing capabilities as needed. This adds extra flexibility to support our final system assembly operations in the U.S., alongside our existing AI production facilities in Taiwan and Thailand. If the policy conditions are clear and customers are ready to move quickly, we can complete mass production readiness within just a few months. This positions us to effectively support downstream production at our U.S. site and respond quickly to market demand. Reflecting the continuing capacity expansion and relocation, our CapEx for first quarter reached TWD 4.7 billion, up from TWD 2.6 billion in the same period last year. In line with this momentum, we revised up the full year CapEx budget to TWD 20 billion. We'd like to provide some commentary here on several factors that will impact our business performance in second quarter and the remainder of 2025. Driven by customer pulling in orders during the period of short-term tariff costs and ongoing manufacturing relocation, we expect that the strong notebook shipment momentum seen in first quarter to carry into second quarter and result in a high single-digit sequential growth with upside potential. However, this demand is somehow front-loaded, we could see a more muted seasonal pattern in the second half of the year. At this point, visibility beyond 90 days remains limited due to ongoing macro and industry uncertainties. We'll provide more color on full year guidance when additional clarity is available. For auto business, we are shifting our focus and investment away from low-value commoditized and volume-driven products like small modules and sensors. Instead, we are prioritizing resources for the development of high-value and content-rich products such as advanced ADAS, car computers and [indiscernible]. To navigate the current auto market uncertainty. This strategical move of resources relocation will help bolster our competitiveness for the upcoming recovery. Considering the evolving revenue mix, driven by the accelerated shipment of new generation AI server racks with higher ASP, relative to lower ASP AI models, general compute servers, notebooks, auto and other product categories, we anticipate some degree of margin dilution moving forward. The new generation of AI servers went to mass production in late first quarter, while still in a slower ramp pace due to supply constraints. On top of that, the demand for the lower ASP AI models remain strong and contributed to the overall AI shipments as customers are urgent for timely deployment of their AI infrastructure. We are seeing strong demand and limited supply on AI. Given that revenues are tied to the timing of shipments of respective projects, we could see the reliability in server revenue growth rate, depending on ASP mix and individual customer delivery schedule. Servers was the largest contributor, driving the sales momentum in the first quarter with robust growth from both AI servers and general compute servers. AI server sales exceeded 60% of overall server sales and their contribution will further increase in the first -- in the second quarter. The sales of AI server grew more than double from a year ago, in good shape to realize a substantial full year growth for 2025. Currency exchange fluctuations are another uncertain factors that could impact our quarterly performance. Over the years, we've been establishing a hedging mechanism to manage this risk as a regular course of business, monitoring our net position exposed to the exchange rate fluctuations on a daily basis. In general, most of our account receivables and comparables are denominated in U.S. dollars. The exchange rate risk tend to offset each other. Even during the recent turmoil in the ForEx market, our hedging strategy has proven effective for our growing business and the risk remains manageable. Overall, we believe the product mix has a greater impact on our margin than currency exchange fluctuations. To close, we are pleased to present another remarkable quarter to our investors, and the first quarter results provided a great start to the year. Though the outlook for the second half remains unclear, we are set for a better second quarter. Thanks to our employees and management team for their dedication in achieving the unprecedented outcome that enables Quanta to consistently make our way for a balance between delivering satisfactory shareholders' return and investing in our long-term goals. Recently, Quanta's Board of Directors has reserved a record cash dividend payout of TWD 50.2 billion for fiscal year 2024 earnings distribution. The management also executed several rounds of capital investments in our manufacturing sites in North America. The investment helped expedite the expansion of capacity and product -- production relocation, specifically for AI server manufacturing. The expansion is providing us additional flexibility to respond to the U.S. tariff policy. We are seeing CSP customers maintain heavy investment in AI compute and data center infrastructure. The recent CapEx guidance from top U.S. hyperscalers confirm that they are either maintaining or rising CapEx for AI spending in 2025 despite from a high base last year. As such, the accelerating AI shipment presents a major tailwind for the year. We keep working closely with our customers in an agile and nimble manner to effectively make the necessary adjustments to adapt changes. Over the decade, Quanta has had solid track records in facilitating prudent strategy, smart investments and excel operational execution to navigate through fluctuation environment. As we scale our global presence, we remain focused on driving improvements in productivity, efficiency and utilization to help offset the growth in expenses and manage associated risk. Overall, we remain cautiously optimistic of our business for the rest of the year. We are confident in our ability to stay on a growth trajectory with profit expansion continuing to outpace cost increases. We believe that Quanta has positioned well to capture value through continued investment, innovation and strategic partnership in the rise of AI. And that concludes the review of the first quarter results and business outlook. We will now open the call to Q&A. Please limit your questions to one and one follow-up. Thank you for standing by. Now may we introduce the first question, please?

Yes, sure. Thank you very much, Carol, for the financial review. We'll now begin the Q&A session. [Operator Instructions] Let's go to our first question, Randy.

My first question, I wanted to ask on the second quarter outlook. First, after the strong first quarter, you had a 20% dip to start April. Curious if you could give a view as we go through if you're seeing an acceleration, how you view overall growth for second quarter? And then if you could split that between the AI, the higher ASP AI racks versus lower ASP AI server and then also for expectation on continued strength from general server.

Thanks, Randy, for your questions. As Carol touched on in the prepared remarks, the new generation of the high ASP AI server will kick in, in this quarter. So we're expecting the percentage will keep growing among the split between the AI server to general server will rising from 60% in the first quarter to keep on by growing to 70% as we set target for this year. So we keep seeing more high ASP product contribution into. So that will -- that's what I mentioned about there are some dilution for our gross margins on the same direction. This is our AI server situation for second quarter.

Okay. And I'll just ask the follow-up, and it will be a two-part follow-up. But if you could give a view of the general server, what drove the strength in first quarter? It's been kind of a stable market, but how much was share gain? And then how much you see that market continuing? And the second part of my follow-up, I actually wanted to ask about the GPU supply demand. I think you mentioned tight supply of GPUs, but we've seen a lot of delays on the racks, but actually chip production smoother. So I'm curious where the constraints are on GPU, like if you're getting enough availability whether for the AI server, high ASP AI server or constraints with prioritization there waiting for those racks, the low ASP server. Just curious why supply constraints given the delays and where you're seeing those constraints?

Let me put another way around for you to answer your question. First of all, customers to deploy the data center buildout keep going down and demand is very strong. So they pick up when is available in the market. So in the first quarter, due to the limitation from the high ASP of the server, the customer employee more current versions of the AI server. And definitely, the allocation between the current order between the high ASP and low ASP rather talking about the share gain or share loss. Let me put the other way around for your reference because the demand is still there, demand is still intact. Customers just do the reallocation between the GPU, the readiness, so they can move on to next version or current version, okay? So that's a dynamic. And moving to the second quarter, the supply of the high ASP is getting better and better, but just subject to the availability from the chip vendor rather from the customer demand. Customer is still over there, still intact, still very strong, okay? So it kind of dynamic to relocate different split between the high ASP and low ASP AI server.

Okay. And if you could comment on the general server outlook, and then I'll drop back to the queue.

They keep going, but in this quarter, the high ASP server, the growth rate should be better than the low ASP server relatively.

And then the next question will be coming from Howard.

Congratulations on the quarter. My first question is on notebook. So I wanted to ask, I was a bit surprised that notebook was less than 25% of total revenues in the first quarter. So I guess my first question is, are we seeing a pretty strong kind of recovery in terms of Chromebook shipments in Q1?

No, it's not allocated to the Chromebook. Chromebook still come back to the regular pattern, only 10% plus/minus something only. So it's not a key portion of our total notebook shipments.

Got it. And then real quick, just in terms of the 2025 kind of guidance by segment, just to touch back on what you guys guided for the last time. You guys said that AI revenues will grow triple digit year-on-year and general server will be flat year-on-year and automotive will be up single digit year-on-year. Are these guidance still unchanged 3 months later today?

By far, the AI contribution to over 70%, over 3 digits is still on track, okay? And the general server looking slightly better from flattish to slightly up, for year-on-year basis.

Got it. Okay. So one last question for me. And I just wanted to clarify maybe your answer to one of Randy's questions earlier. So looking at the lower ASP AI server product, you mentioned Q1 demand was quite strong. Is it fair to assume that, that demand will continue to remain that strong for the remainder of the year? Or are you expecting demand for the lower ASP AI servers to fall off in the coming quarters?

As you can see it's still growing, as I mentioned, there are two dimensions for your reference. First of all is about the allocation between the -- split between the high ASP and low ASP [indiscernible] of the server to the availability. And we still see the low ASP, a general server still keep growing at a percentage maybe below to the high ASP servers for the quarters.

Still growing, but growing at a slower pace.

Yes.

And the next question will be coming from Angela.

Just a follow-up question regarding to the AI server. Can you give us a definition of the low ASP and high ASP of the AI server? Not that sure what that means low ASP AI server demand is strong in the first quarter. That's my follow-up question first.

Yes, that's the different ASP. [indiscernible] No, that's the current version of the GPU or next generation of GPU due to the different infrastructure, so they have a different price tag. And the next generation now kicking in this quarter is a high ASP the current GB200 stuff.

Okay. So it means Hopper is a lower ASP AI server and Blackwell GB is a high ASP AI server. Am I right?

Yes. Your comment is saying yes...

Okay. Thanks for clarification. And within your AI server is adopting buy and sell model. And based on accounting principles, if you buy the GPU from clients CSPs and assemble and sell to them, will you book your sales in the value, including GPU CPU or you will have a net deducting the major chip with the other net value only? That's the question follow-up.

That's the mix depends on the customers' requirement. It's a consignment model, they will offset the key components on the cost. In the buy-sell model, then we kind of everything. So that's a mix among our customers. But the major lion part of our business is still buy and sell models.

Okay. So based on your clients and the current business model, is the mix of the consignee higher or the buy and sell higher?

I just touched about the lion part of business is buy and sell models.

Okay. Last question is that regarding...

The major part...

Major part. Buy and sell.

Yes.

Okay. And the last question is regarding to the first quarter financial about the non-op. It's quite rare to see that your interest net expense in the first quarter. Can you explain the major reason? And will this condition last into the following quarters?

Yes. The business keeps growing. So you can see the flowing of account receivable, inventory, there is working capital hungry. This is also one of the strength of advantage, our balance sheet, our income to accommodate. But to win the business also expense you're sowing working capital needs. As a result, you need to have additional interest expense to cover the sowing working capital needs. So customer require you to pull in the GPU, that means you need repair more working capital to accommodate. So only a few players -- that's one of the key reasons only few players can in this industry to working with the CSP for this type of product for this type of business.

Okay. Makes sense. And so do we need to expect that this situation will continue when your GB or AI, high ASP AI sales growing up the we will see this kind of situation continuous.

That could be continuous, but as how we do funding this working capital needs, okay? You can pay for any low funding cost vehicle, maybe can lower down the overall interest expense going forward, okay? But I think they have a peak season in the first half. When ramping up giving this business model, then that definitely comes with more working capital needs, okay? That as a result with a higher interest expense.

Next question will be coming from Albert.

Congrats management on the strong results. So my first question is based on your -- what you're saying, the CSP customer may bought a lot of Hopper HGX in first quarter. And now we are ramping up the Blackwell server into second quarter. So my question is, how would you -- how will Quanta see customer preference in Blackwell HGX versus GB server configuration? Is it fair to say GB is a preferred solution by CSP in Blackwell generation? And if customers may face some issue in GB server, is it possible for CSP customer to switch back and buy all Blackwell HGX or it's not doable because their design is mainly for GB.

Thanks, Albert, for your question. Customer -- firstly, customer has different AI adoption plan, and they have a huge AI demand. So they can accommodate different types of compute needs ranging from H200, GB200 or GB300. So there are different type of adoption plan. And from a product perspective, we cannot speak for the customer or speak for the vendors. So what we can do is best to meet the customers' requirements at the right timing, right cost and right scale. So that's what I mentioned, we are do the allocation between the high or low end ASP AI server. Quanta, we can say we are ready for every type of product and customers subject to how to roll out the AI adoption plan. So we cannot speak for them. Well, for sure is we can accommodate what the requirement. So whether they can scale back or not, which we don't have ideas. But what we can do is we get ready at the best price for that best TCO leadership product for them.

Understood. That's very clear. And my follow-up question is, Carol mentioned notebook will grow like high single digits with [indiscernible] shop in second quarter. And you also mentioned that the Hopper or HGX server will grow, but probably at a slower rate versus GB, which means AI server will accelerate into second quarter. And with these two, that means probably we will see more than 10% quarter-on-quarter revenue growth in second quarter. Is it a fair argument? And if that's the case, second quarter can be easily like TWD 55 billion, but April revenue is only TWD 15 billion. So why there's a big shortfall in April? Is there a specific reason and we are expecting a huge jump in May and June?

Albert, again, thanks for your question. First of all, in our customer -- we don't guide about the revenue growth rate and that granularity. What we can say is we see the notebook is going better and high ASP server will kicking in this quarter. Again, this is still an unknown swing factor about how the customer to allocate between the high ASP and low ASP as well as the readiness of the chips from chip vendors. So that's a swing factor over there. Anyhow, we're expecting...

Elton?

Carrie?

Yes. Elton, I think we lost you for a couple of seconds.

Really. Okay. Just touch on about for the -- we don't provide granularity for the revenue growth rate based on our customer targets. But what we can say is we see the better notebook pulling from the customer for the second quarter, and we're expecting witness the high ASP server will kick in this quarter. Again, it's subject to the readiness of the chip vendors and the customer plan to allocate between the high ASP and low ASP server, okay? But we are looking for that and what the innovation we have now for the second quarter.

Next question will be coming from [indiscernible].

I just want to know that many automakers and companies are collaborating with on automotive projects. I'd it typically takes 3 to 5 years to go from project initiative to mass production. I'd like to ask, is there a chance that our automotive business could reach TWD 2 trillion level in 3 to 5 years later.

As Carol touched on in the prepared remarks, the EV business probably facing some certain hiccup in this year due to the macro economy, due to the tariff issues, due to the inflation issues. So that's why we mentioned we changed our strategy to migrate from the low EV product to high value added product, okay? And we keep -- we're expecting the revenue growth still flattish for this year, okay? But we're still expecting -- we still are excited about this outlook for the EV business onwards. So we still think that should be in the next 1 or 2 years, another growth driver for Quanta onwards. Hello?

We will move on to next question from Randy.

My follow-up question, you mentioned some of the high ASP server subject to the chip reliability availability. Could you discuss more on the hardware? I mean there's been comments about yield ramp for the past few months. But if you could talk about what you're seeing from the ability to manufacture and improve yields, how smoothly you see that ramp up? And as we go towards GB300, originally, there was optimism at GTC about some of the design changes on the new architecture, but it looks like we'll keep with Bianca. So how you're seeing the transition to GB300 if that ramps up, say, later third quarter or fourth quarter?

Firstly, we don't comment about the rumor. Secondly, if that's a whole truth, we are more than happy. That be easier for the ramping schedule and improve the yield rates. But again, that we don't speak for our vendors. But what we understand is that most of the key issue has been solved for the GB200. So but what the customer will change their migration plan, we don't know. But as I mentioned, customer has their own adoption plan, and they will ever change it. So we always have flexibility to meet the customers' requirements, okay?

Okay. Great. That's clear. Okay. And I wanted to ask a follow-up on margins. I know you can't guide explicit, but you are coming off a high base, both gross and operating margin. So maybe a couple of things around it because the -- if you could give a direction for OpEx growth and then also whether it's kind of an ability to maintain operating margin, even though gross margin comes down, you can, with the scale up, maintain that. And sorry, the second part, I wanted to ask for FX, is there a sensitivity on revenue and cost from that NT dollar recent appreciation? How much -- if that would have any impact or that's kind of a natural hedge to the gross and operating margin?

Okay. Thanks, Randy. For your first question about OP margin, maybe [indiscernible] because every quarter has different product mix. So there are different leverage. So may be up and down. But for the overall OP expenses, we are managed to make sure the growth -- the OP expenses, the growth rate is below our sales growth rate. Secondly, it's probably slightly above of next year total about TWD 49 billion something, okay? So that's up and down. In the second quarter, we still need to put more effort to ready for the GB300. So we're expecting the R&D expenses should be raised as well. So more than for conservative sake, we're still willing to see some of the pressure for the OP margin in the second quarter, even though they have a leverage on our top line. For the second part of your question about FX, I also love to have some magic to come on a linear forecast how many percentage of NTD appreciate, they will come out how many percentage impact our margins. But this is different way around, okay? So there's a lot of [indiscernible] factors among that. So that's what I mentioned we have a natural hedging for AR, AP. Secondly, we have a dynamic hedging mechanism. We don't live in a negative position. We will always margin on a daily basis. So we hopefully, we can manage well. But [indiscernible] anyhow, we still see kind of manageable and controllable based on our regular basis of the hedging activities.

Our next question comes from Kaylin.

I'm asking on the margin side. For OPM, I think we mentioned that in 2Q, our R&D expense could increase on GB200 ramp. Does that mean our second quarter OPM will also have pressure quarter-on-quarter gross margin pressure because of product mix?

Yes, I just touched on that, yes.

And on the overall long-term view when high ASP AI servers like GB200 reached a certain scale, and we also have certain experience ramping that, would you say those AI servers OPM is higher or lower than current OPM level?

This is a dynamic industry. We keep the next generation of the new CPU in the market. So we keep the new R&D expenses to make Quanta ahead of a competitor to come out set out a product. And that's what I mentioned, we more likely to provide the guidance overall OP expenses for the full year basis for your modeling your forecast. So what I'm saying next year, we have TWD 49 billion something for overall OpEx. And this year, probably about the TWD 49 billion. But again, our goal is to make sure our revenue growth rate is below our expense growth rate is below our revenue growth rate. That's our overall goal. The next goal is to increase absolute margin dollar since the product mix can change from time to time and subject to the product mix and subject to the new product ramping out. So there's a lot of volatility about the OP margins. So probably I give you another dimension for you to model in your forecast.

Okay. Just a quick follow-up on previous question about GB200 assembly yield. Are we seeing any issues with that? Or it's mainly the chip supply issue that we are seeing to ramp GB200?

Thanks for your question. I don't want to comment about what the vendor performance, okay? What we think about is what the amount and the timing, our customers -- we don't buy directly from the chip vendor. We buy from our customers. So customers give out a schedule and we discuss what the schedule. That's what I mentioned. We are not other module maker. The customer is a chip maker. Our customers will buy from -- based on the buy-sell model, we buy from CSP. So the key thing for us is how to discuss what the timing and the quantity they can provide and what to allocate our resources allocate our product among the H200 or GB200 or GB300. So total different dynamic. When you talk about the vendors, which we are facing about instead of the chip makers.

Due to time constraint, our last question will be coming from Kevin.

Congratulate on the outstanding quarter. And my first question is about the visibility. As you have already mentioned, the visibility for like [ MVPC ] for those consumer products remain very limited for the second half. And I want to ask about what is your visibility? What is the order visibility of your server, especially for AI, high ASP or low ASP server business for the second half? And what do you think the future will be for this industry? Like it has been a very strong growth during the past year in 2025. Do you think the momentum will remain for the year to come? That's my first question.

First, definitely, the visibility is getting better. That's why we kick in, in this quarter. And again, on [ war ] is still subject to the availability for the chip makers. And whether for what outlook for 2026, actually, all the CSP's earnings call already tell you they keep their investment, they keep a lot of CapEx, and they are also looking forward to keep a growth rate for 2026, maybe growth rate is lower, but 2026 still keep growing. This is what I keep talking the demand there, the demand intact, but subject to what availability about the chips, they will come up. So customer will decide what to allocate the product because they want to keep to provide AI service, they want to keep deploying the low of the data center. They don't hold the pace, okay?

Yes. Understood. So can we assume that we have full visibility for server business for like second half this year?

We target for that. We're looking for that [indiscernible].

Yes. Okay. Understood. And my final follow-up is about the automotive business. As you already mentioned, it still will be our momentum for like the next year to go. Yes, I just wanted to understand the like percentage contribution we are expecting for the auto business for years to come because we have mentioned it maybe like high single, maybe reach double digit, yes, but we all see that the growth momentum for the high ASP server is very strong, the percentage-wise, it seems to be diluted. Any new comments on this percentage contribution for the years to come for auto business?

Again, it's still single digit for [indiscernible] the year, okay? And if we all agree, 2026 is still a growing year for the server. So we are expecting this -- the EV business will outgrow the server business in 2026. So I will say 2026 is still single digit for the AV business for the next year, okay.

Yes. Understood. So maybe we'll see it in like 2027, when we will see the percentage may be larger?

I totally agree because there's a long certification process. So what we can do is to make sure we are keep going -- keep go along with the key buyer, key brands, key every player. We are still working with them. And that's why today, we are touched on about the expansion in our German, one of the key purpose to provide a service enough customers in the Europe continentals. So this -- again, all this stuff, no matter it's EV or server, we need to prepare upfront and taking time to work with the customer to win the customer trust, to come out the cost, list product for them. That's why you can get a low-hanging fruit when market come out, when products come out in the market. That's part of our DNA. That's part of our strategy.

Yes. Understood. Sorry, may I have the final follow-up, like our dividend policy. We have very good dividend for this year, but you still have already mentioned that the CapEx seems to be much higher than what we expected for this year. So I'd like to ask about the future dividend policy.

It depends on how the business grow. Again, if you look at OpEx about what we're doing in the dividend policy, we're always looking to balance between the cash reserve for the CapEx, for the R&D and also to return to our shareholders. So it could be some of the volatility. But if you look, we at least pay 80% of our earnings to our shareholders. This is our track record.

And Elton, I know we are running out of time, but maybe just one very quick question from me regarding GB200. The system complexity have obviously raised the bar for ODMs. So order allocation has been highly concentrated on leading suppliers, including Quanta. But as we migrate to GB300 or even next generation, do you expect to see any competitive landscape changes? Or do you expect to gain more market share?

Again, again we work with the customer, closely working with the customer and timely provide the design for them and provide the most -- again, competition is always there, and we get used to that. We also assume there's a competition over there. So we don't fear about any competition, but who can provide a fixed cost, who can provide the leadership product, who can provide the TCO product, they can win the orders and Quanta get used to them. And Quanta keep working with key chip makers, make sure we go along with them. And also fully unstack what the CSP need, no matter how they change the strategy. If you know customers well upfront, they even know how to design product for them, then we can keep winning the orders from that. Again, this is a market practice and Quanta in this business for decade, will migrate from the notebook for the regular server to the AI server for the EV or even for the next new product, the same DNA, same strategy. We see -- we get used to the customer from a decade ago to make sure we know the product, we know the customer, we come out and the computing design still our DNA. So we make sure we can keep winning the order from the customers. That's our strategy that's our DNA.

Understood. Before we close, I'd like to thank the management team for their time and insights today. C.C. and Elton, do you have any final remarks to share?

Thank you for the questions about our Quanta. And as time come then we will still continue to do our best on the [indiscernible] development and in technology and to work closely with our customers grow any possibility that comes.

Great. Thank you again, C.C., Elton, Carol and Alai. Thanks to everyone who joined us on the call today. If you have any additional follow-ups, please reach out to Quanta's IR team or me Carrie from Citi. That concludes today's call. Thank you, and have a great day.

Thank you.

Thank you.