Advantest (Tokyo:6857) Q1 2026 Earnings Call

0:06 We'd like to take the first question from Ms. Shikana from JPMorgan Securities. Thank you very much. This is Mia from Mia Shikana from JPMorgan. First of all, con congratulations on a very strong learning. I had a a question around around the SOC tester market. How has your view of the SOC tester market changed with your prior outlook in terms of which applications are is it GPU, ASIC, or is it CPU? And how should we think about the relative contribution from higher chip volumes, longer test times and increase in test insertions? Like, I guess it's a bit early, but are you thinking about the sustainability of the demand into next year? This is Doug. Thanks, Shikanae san, for the question.

1:15 To answer on the SOC market, we're seeing very strong demand across all of those areas, GPU, custom ASIC, and CPU. And I would say most of the increase in the SOC TAM that we just announced is coming from more inference applications. And and so that's what's really driving that change more than anything in particular with the the CPU business. We do see that to be sustainable throughout this year and into next year. As to kind of a test content, I think we're continuing to see additional test content increases through test times and additional insertions, in particular for the accelerator market where three d packaging is is driving more complex devices.

2:13 So that's gonna continue to be a big driver for us, but a lot of the unit volume increase on the inferencing side and CPU side is driving some of the additional market TAM increases. Thank you. And just a quick follow-up. How

3:53 does CPU testing compare with GPU in terms of test time intensity? Yeah. The the test time, particularly for the package test on CPU, is gonna be quite a bit lower than what we'd see on the accelerator just because the complexities of a accelerator in in a three d package. At the wafer sort level, if you're testing a monolithic chip, you get they're relatively similar. You may see actually a little bit higher on on the CPU. But in the in the back end, the final test package, definitely, it's higher. I'm I'm not at liberty to give you clear numbers on that, but it's just it's fair to say that the accelerator side on the final test is higher.

5:20 Very clear. Thank you. Thank you very much, Ms. Tsukanae. And next question is from Mr. Yamamoto of Mizuho Securities. Go ahead, please. This is Yamamoto from Mizuho Securities. I'm sorry for the very similar questions. I have a question for Page 12. This is regarding the desktop market outlook for SoC.

6:46 So now as for midpoint, you have revised your projection from 9,100,000,000 yen to 11,000,000,000 yen 2,000,000,000 yen increase. So I'd like to know the breakdown of $2,000,000,000 among ASIC, CPU and GPU. How much impact for each one of three application have and how much upward revision is made? Also for the SoC tester sales, again, you have revised the sales projection. So I want to know the breakdown for ASIC, CPU and GPU for that after acquisition. Yes. That's a good question. The upward the change is mostly, as Doug said in the beginning of the call, due to the growth in inferencing and the compute segments.

7:33 I think it's too complex, in my opinion, to divide it up into three different segments because a lot of these devices go together. And some of the applications are also interchangeable in terms of how they are used in the end. But the upward projection, obviously, as Doug said, is due to the changes that we are seeing in the compute or the growth we are seeing in the compute and the inferencing market. We can also safely say that the custom ASIC market, the compute market, as some of the leaders in the industry have been saying, are going to be significantly larger than the GPU market. So right now, I would say GPU market is still the largest.

8:29 But as the time goes, we will see that the other markets are as large, if not significantly larger. And hi. This is Doug Yamamoto. So let me add just, you know, one comment, which is for us, you know, since our our 93 k platform

10:00 is so prolific and and being used across the industry, across all of the different compute applications, It really doesn't it doesn't matter where that growth comes from because we're positioned with the 93 k so that, you know, in fact, the configurations are very similar in some cases with our our digital instruments and our power supplies. And so oftentimes, as Sanjit mentioned, our customers will will purchase capacity, and they'll be able to configure those for whatever type of application. So that's a it's a strength that we have right now with that platform is is the the usability across all of those different high performance compute subsegments.

12:12 So I have additional questions. Is that because of the increase drastic increase of the number of chips, which drove the upward revision? Because I think advanced people knew a lot about the what sort of test will be conducted by customers. So I don't think there is no surprise. There is no I don't think there is any surprise in terms of testing session or test time. So I wonder, this is kind of about why the vision is driven by the number of chips. Is my understanding correct? Yes, I think so. I think there are a lot more customers now doing custom ASICs. Most of the hyperscalers, as you know, have some type of a strategy of doing chips in house. Those chips are obviously, in addition to the GPUs that we have seen in the past.

13:02 We've also seen significant increase in the compute market and the server space. So obviously, that's additional. One segment that we have not talked about before that we didn't use to split, networking, we actually also have a significant market share in that space. We just didn't talk about it before. And that space also has grown significantly. So the answer to your question is absolutely yes, that there are a lot more chips now. And luckily or fortunately, we are a major player with dominant market share, as Doug mentioned, in all of those segments.

13:52 So we are benefiting from the increase. Thank you much. That was very clear.

15:13 Thank you very much, Yamato san. Now we'd like to take the question from Mr. Nakamura from Goldman Sachs. Thank you very much. I wanted to ask you about the production capacity expansion plans. You mentioned that you're accelerating the efforts right now, but could you clarify whether you're accelerating your capacity expansion plans even compared to your commentary you made back in April? I think at that time you mentioned for SOC testers, you're expanding into roughly 5,000 units by the end of March twenty seven and then to 10,000 units by either end of twenty eight or early twenty nine. So just wanted to understand what your capacity expansion plans are as of now. Yeah. Thank you, Nakamura san. This is Doug.

16:23 Yeah. As as you mentioned, we had stated a 10,093 k's per year by the end of twenty twenty eight, early twenty nine. And, you know, we're we're pulling that in. Not prepared yet to say how much we're pulling that in with a new date. It's something that we're working on right now internally. What keeps happening is there's more wafer starts, more advanced packaging capacity being put in place. And so every time we think we've got locked in on on where we need to be on on the scaling side, we have to go back and and look again as as more capacity comes online. The the good news is our our our production systems are set up to allow us to to scale like that. And so between our supply chain and and long term agreements we have with component suppliers and our relationship with our contract manufacturer, they can react quickly to our needs.

17:24 In fact, our contract manufacturer is putting a second factory in in Southeast Asia to help support the scaling that we're asking them to do for us. So that's a long answer to a short a really a short question, which is, you know, we'll we'll have to get back to you. We're we're scrubbing the information now, and I expect next quarter to have a more clear date for when we'll bring in that that production expansion. Thank you very much. Just as a follow-up,

19:28 in terms of the 5,000 unit mark, are you approaching that number already? I think you used to say by the end of March twenty seven. Is that is that being put in right now as well? And also, could you also touch upon your capacity expansion plans on the memory testers as well? Thank you. Yeah. For five five k, it's safe to assume that we're we're going to blow right through that level at this point in time.

20:14 So that the the our supply chain people have worked hard enough and have been able to find the capacity for us to be able to deliver at that level. And so we're on to the next step, which is that 10 k level. As far as the memory goes, we're also have expansion plans that are are highly correlated to what we're doing on on the 93 k side. There's not as many testers in the memory side as the 93 k, but there's similar growth rates. And so there's lots of programs within the company right now to to expand on the memory side.

21:00 There is additional challenges there because we do more in house production, and so there's things that we're looking at within our supply chain on the memory side to allow us to to keep the pace with the SoC testers. Understood. Thank you very much. Congratulations.

22:39 Next question is from Mr. Mr. Shibano of So I have a question regarding margin.

23:44 On the first quarter, you recorded very high margin. But when I look at full year projection in the remaining nine months, I think you have very conservative outlook for in terms of margin. So second quarter and onward, the second half of this year, I wonder there might be some margin pressure or some product mix, some concerns. Are there any unfavorable factors in terms of margin pressure or product mix? And second issue is actually demand is rather strong. So when it comes to next year, Mayfenti, is it necessary I wonder whether the significant increase of top line might be necessary. If that's the case, do you expect also the margin pressure next year? Or do you think that sort of thing can be absorbable?

24:31 Yes. Thank you, Shivan or Sani. This is Doug. Yeah. On on the margin question, I think, you know, the company is hitting on on little cylinders right now. That's the ends and the product mix was favorable. I think we've had favorable exchange rates as well. And then as as our volumes go higher and and we've been able to to really deploy our platform strategy, it's really yielding big economies of scale for us as well. The headwinds right now, obviously, are some inflationary things on on our bill of materials. Of course, at the top of the list is is memory, and the memory pricing is is a a headwind for us as well as everybody.

25:20 And so that's that's why I think, you know, we have to be conservative because we're we are subject to the the pricing environment on the memory, and and our tester do use a lot, a lot, a lot of DRAM. And so that's the biggest reason, I think, that we're still, you know, not ready to project a consistent level of margins like we have this quarter for the rest of the year. Also, I I think, yeah, I would just note that, you know, we had some reversal of inventory obsolescence that was was was helpful to our margins in this particular quarter, and that's not something we would necessarily see as a consistent quarter to quarter basis. So going into next year, still, I think we're gonna be facing some significant pricing pressure on on memory, and we are working on certain areas where we can pass some of that along.

26:22 It's really case by case, application by application, customer by customer. That's a very sensitive topic, but I think there's gonna be some something we can do there. But I hopefully, that answers the majority of the you had a lot of questions in your question. Thank you very much.

28:36 Thank you, Shibano san. So my name is Wadaki from SPI Securities, and I'd like to ask Doug Sun about the recent AI developments.

29:40 What's going on in the world of AI right now? I think you know a lot because you're related to the leading edge AI situation of the world. I know that your AI testers have geared up really another notch, bet you're up to speed. So what's going on there? Why is AI testers becoming so busy? Is it because of the agentic AI? Is it physical AI? Or is it kinda like the HAI coming from Stargate? Or I mean, there's various trends. But what really do you think is driving this big trend of AI? Yeah. Thank you, Uraki san. You know, I I see mostly the agentic AI right now driving a significant structural change in the business. I I know that we as a company are deploying the use of agents in a lot of what we're doing for our development and on behalf of our our customers.

30:34 And as I as I talk with other executives in the industry, they're doing the same thing. And the the level of efficiency gains that the our companies are getting through the use of AgenTik AI is is really staggering. And I I think a lot of people on the outside of kind of businesses and companies don't always realize what's going on in in the back offices and the development, r and d areas of of the companies. And that's where the big bang for the buck is happening, I would say, is is is through the use of the agentic level, where it can call upon all kinds of different software tools and on its own, in a headless fashion, develop solutions.

31:22 And so for me, that's that's what's really happening. I I don't think that we're we're at the physical AI yet. It's coming. But I think there's there's gonna be a wave of physical AI applications at the edge that are coming. We're seeing some of those. But I I'd still believe that has got a little bit of time before it becomes mainstream. When it does, the volumes will be, you know, extremely high and will make a lot of the data center pale in comparison. So really, agentic level now and and into the future physical So just to follow-up on a question of

33:41 this trend here. So regarding AI, there's just billions and billions of dollars of investment made in this whole area. But do you think these surprising concepts like AGI and ASCI will really come true in the future? What do you believe, Daksan? Oh, wow. That's really putting me on the spot with Daki san. Yeah. I I think, ultimately, that's where we will get to, but I I I tend to think that's out pretty far in time. There there's so many use cases just in front of us to utilize AI to to do really wonderful things. And that's that's where I personally, I hope that, you know, the world will start focusing on all the the beneficial things of AI and and not get caught up in in a lot of the misuses of it.

34:29 And so that's you know, Advantest has an obligation and the position we're at now to make sure that we're pushing the applications that are doing very, very good things, whether it be medical research or agricultural things or or even you know, I was talking about the the the fires and the use of drones for for putting out fires and and seeing what they're going through in Europe right now even as well as in in The United States. Those are the kind of things I think that the world needs to focus on rather than get caught up in the, you know, singularity concepts and and things of that nature because I think those are are out there. They're probably gonna come eventually, but it might be far after you and I are around.

35:15 So so I gave you a philosophical answer to a philosophical question with Akisan. You.

36:47 Next question is from mister Yoshida of CLX Securities. Thank you. Congratulations on the strong results. My question is similar to the prior ones, but I would also like to clarify about, SoC testers. I think there is a structural SOC tester demand growth driven by CPUs for infra AI, which is shifting from Intel to fabulous designed ones, creating additional third party tester time, taking shares from Intel's in house testers. So you have just answered a kind of vague comment about the impact by device types to the SOC test that time growth, but, I guess CPU's contribution is the biggest one, for the upper revision of this year time, outlook. And this impact will get bigger towards the next year considering the increase of in a fabless device CPUs going forward. Do you agree with this view? Thank you.

38:26 Yeah. This is Sanjeev. So I wouldn't discount the size of custom ASIC market. I think it's definitely what you said is definitely correct, that CPU and server market is growing rapidly. And obviously, we are benefiting from that greatly. But I would say that the custom ASIC market and also the networking market, they're also growing at similar paces. So again, repeating my comment, I think we're fortunate that we have dominant position in all of these segments. So to some degree, we don't really care because, as Doug mentioned, the configurations, the customers, the systems are very, very similar to each other.

39:22 But if you really want to divide it up, yes, compute is growing very fast, but let's not discount custom ASIC. I think that market is definitely understated, and it's very large for us. Let let me add, Yoshida san. This is Doug. On the CPU side, it you know, if it's x a six or ARM based or if it's fabless or if it's IDM,

40:52 we are positioned every single place with our 93 k system. And so we have full coverage of all of the the CPU accounts globally. So we feel very, very good about our position there and and the market share swapping that might happen between these companies again is is not really relevant since we we have a de facto standard in the 93 k across the entire CPU industry. Okay. Just a quick follow-up.

42:04 You also please share your views on the current outlook of your market share at the SoC and memory tester for this year and if you have next year as well? Thank you so much. Yeah. So we we will the numbers we we share next year, April next year. But I can speak at least qualitatively on on some trends. On the SOC side, you know, we reported for 2025 that we had 66% share. And I think it's it's safe to assume that we're continuing a trajectory upward on that.

42:53 We can see, you know, a path towards 70 plus on that in the future. And, you know, within the HPC market, we're far higher than than that that level. So we see the gap really widening in the SOC market, and and so that's some qualitative looking at the at the SOC market. On the memory side, it's a it's a busier area. There's more competition in the memory side. There's some areas such as NAND where we have not been a traditional incumbent, and and NAND is is growing, as you may know. There are some transitions with device speeds and bandwidths right now also. So there's some insertions that we haven't necessarily captured.

43:49 And so on the memory side, I would say this year may represent a little bit of a downturn for the company, but we see that more than recovering as we go into the next year where our road maps are are well aligned with devices. And so we we are very, very confident in the future around our our memory market share. But you'll you'll see a little bit of a dip potentially in memory while we still grow on the SOC side. And and for us, the SOC market is is, you know, is such a larger market than the memory. So when we have growth in the SOC, that's generally an aggregate positive for the company.

46:22 So Yoshiro san, I just want to add one thing on the memory side. I think when we look at our future, and the reason we are confident is we are going to be releasing a couple of key products. And what's really amazing is that we already have customer commitments for those products. And I think those products are really game changers. And what's also interesting is that the memory market, we expect also to grow significantly next few years as the new fabs come online. So yes, it's absolutely correct that we have some challenges this year due to the transition to new products.

47:15 But I think the feedback from customers says that we are very well positioned for the future. And any share loss that we will have this year, we are very confident that we will recover that loss over the next few years. Thank you. I'm looking for Thank

48:54 So the next question is from Mr. Yoshioka from Nomura Securities. So I want to look at this page, Page seven, and ask you to give us more color on the by region SoC testers sales. You did very well in Q1, which were the main drivers?

50:06 If you can just explain by region as much as you could. But some of my interest would be regarding Taiwan and China. For Taiwan, how about if we take SoC testers alone? Is that also growing as well, pushing the demand? And how about China where they're nationally very much focused on developing their own semiconductors? Is your business channel also growing? Okay. Yeah. I think well, I think for for our ship to region, the reason that Taiwan is so large is because that's where, most of The US, fabless SOC, business is landing. And so, you know, that's that's why that has consistently been the largest part of our business. So the again, this is where we're shipping to, but the business is really being driven out of of US fabless companies.

51:00 There is some memory in in Taiwan as well. I we don't necessarily disclose the the breakout by region between our our SOC and and memory. The same goes for China. China is about 19% of our total sales, and that is both SOC and memory. It's it's it's pretty consistent with our overall, you know, ratio of SOC to memory in China. So this past quarter, I think we were eighty eighty, 85% SOC 15% on on memory. So you can apply that to most of both China as well as overall distribution.

51:52 But Taiwan for sure is is majority SOC. Thank you very much. Okay. Great. If you

53:28 can just give us more color on how you see that 19% China sales proportion will change going forward. Think it will expand? You know, it's been pretty consistent around 20%. There there are new factories being built in in China, a lot of them. I think those will likely come online in 2028 time frame. So there is potential, for, I guess, in in the '28, maybe, '29 where we'll see some expansion of the China businesses associated with, that that scaling of of factories. But until then, I I I would guess, just estimate that we would stay in that 20% or or lower of overall sales.

54:58 Great. Thank you. Very clear. Thank you very much, Mr. Yoshioka, for your question. Next question is from Mr. Hanayel of SMBC Nikko Securities. Go ahead, please. So this is Han Ah Yeo from SMMCBC

56:33 Securities. I have a question regarding capacity and sales. I'd like to understand the concept of those two things. You said in your presentation you are accelerating capacity enhancement in your work. And do you think I think you always think about the buffer in the capacity when you try to build up the capacity. This time you have revised your outlook drastically. And do you think that is include the buffer as well? I think the demand has been expanding more than your expectation. So I wonder this kind of revision may include the buffer or not included? And or if you cannot have buffer, you may, can you say, lose your share or you may give the opportunity for your peers and competitors. So I think your upward revision of sales is rather drastic and I think you can achieve the projected sales for SOC with buffer or without buffer. What is your outlook?

57:28 Yeah. Thank you, Hanayashan. The this is a really good question. And and the the short answer is we always have to have buffer. So when we start to impede from the demand side on our our buffer, that triggers us to expand further. And right now, what we look at very closely is what are the supply constraints in in the semiconductor industry. For accelerators, for exam for example, it's in the advanced packaging area. So as, you know, the foundries and OSATs add packaging capacity, we then have to expand our capacity and and not just on a one to one ratio because we have the test content generation to generation increasing.

58:28 So we can't we can't just scale at a unit volume level. We have to scale at a unit volume plus the additional test content generation to generation. And so that's how we model our capacity for our for our testers. And, definitely, we have to have buffer because what happens is there's unknowns in particular around test times. So oftentimes, our customers don't know what the test times and test insertions are gonna look at until they've taped out the device and gone through a characterization cycle. And so there's a lot of unknowns, and we have to be prepared to be able to supply against those unknowns.

59:13 So we always will build in buffer into our production capacity planning and make sure we stay above that kind of danger zone. We we never want to be the critical path for our customers' businesses. Thank much.

1:01:32 You Now we would like to take the last question. And the last question from Ms. Tamura from Morgan Stanley MUFG Research Japan. I'm curious about the visibility Advantest shares with the customers. Like how long will the demand discussion with the customers extending? Historically, it's been, I think, months, but how much visibility that does Ventas have specifically into 2026 and beyond? And what's driving this unprecedented visibility if that, you know, is the case? Yeah. Hi, This is Doug.

1:02:38 Yeah. Historically, we we were never given really much more than a six month forecast. But just because of the scale now in the industry and because of the constraints in advanced packaging or wafer starts, it's it's much easier for us to see out, I would say, about eighteen months. And and so we have conversations with our our close partners for eighteen month visibility so that we can do the planning. And, obviously, that that helps us because we we can scale our our capacity plans when we have that kind of a lead time. Six months is is not enough for us to trigger supply chain scaling. So, you know, the again, the short answer is, in a lot of cases now, we're seeing eighteen months, and that's because the industry has realized that they need to give their supply chain that kind of a heads up.

1:03:40 And we we learned the hard way, you know, in the past that six months can lead to some some constraints, and and we all work together now collectively to to try to get, you know, at least an eighteen month forecast. Thanks for your answer.

1:05:01 Just a quick follow-up on the previous answers. You mentioned that the you expect accelerating the capacity expansion going forward. And you mentioned, like, you're pulling in the 10 ks unit capacity target by 2028. Should we assume that there should be some upside to the 10 ks numbers in 2028, the target numbers. Yeah. Thank you.

1:06:06 I think yeah. I think step by step, but, you know, you're likely to see us look at the next step beyond 10 k. So right now, we're focused with pulling that in in the near term. But in the midterm, I think it's, you know, highly likely that we'll do another step on on the capacity expansion. Just just to give people an idea, though, in in '20 around 2020, we were producing, you know, maybe a 93 k's, and those were smaller configurations. If you if you use that configuration for our 10 k, it's about the equivalent of of 20,000 of the of the 2020 size. So we've done 20 x in, you know, seven years, eight years.

1:06:55 And so I I don't know that there's been that level of expansion for for anything in our industry. So it's really quite an achievement for us to get to that that 10 k, but we have, you know, high confidence that we can take that that next step function. It may be 15. It may be higher, but we'll try to clarify that in the coming quarters. We're just not ready to do that yet. Thank you.

1:08:43 Thank you, Ms. Tamura. So although we have received furthermore questions, we would now like to conclude this session. Thank you all very much for taking time to join our financial briefing despite your busy schedules today.