Advantest Corp Ord

Good afternoon. First, I'll present a summary of the financial results for this period, as shown on slide 3 of the presentation. We are pleased to report that our Q3 results exceeded our internal projections set in October, delivering a new record for quarterly sales. Thanks to demand pull-in for SoC and memory test systems serving AI-related applications, the temporary digestion we had anticipated did not occur, resulting in stronger than expected top-line performance.

The results of the third quarter exceeded our expectations at the time of October, and we announced the highest sales in the past as a quarter. Due to the demand for AI-related SOCs and memory test systems, we shifted from the adjusted area we had anticipated and landed at a level that exceeded our expectations.

Q3 operating income and net income also outperformed projections, supported by product mix improvement and only moderate increases in SG&A. given this solid performance through the first nine months period and solid visibility into q4 we are raising our full year fy 2025 guidance as follows sales of 1.07 7 trillion yen operating income of 454 billion yen net income of 328.5 billion yen and eps of 452.34 yen

We will correct the information by saying that the business performance progress of the 9-month period up to the 3rd quarter and the 4th quarter are both good, and that the business performance forecast for the 2nd quarter of 2025 is 1.7 billion yen, the operating profit is 4540 billion yen, the current profit is 3285 billion yen, and the basic profit per share is 452.34 yen. Looking ahead, with AI-related demand remaining strong, we anticipate sustained growth in the tester market in calendar year 2026.

We continue to advance our capacity expansion initiatives to meet robust demand in our executing strategic priorities outlined in our midterm plan to evolve into an integrated test solutions provider. Through growth investments, particularly in strengthening our R&D capabilities, we are positioning the company for continued long-term leadership.

Next, I will walk you through the business environment surrounding our company and the outlook for the tester market as shown on slide 12. Before providing a tester market size update, I would like to first address our view of the business environment.

While uncertainty lingers due to ongoing geopolitical risks and the possible risk of sharp exchange rate fluctuations, AI applications continue to be the main growth driver of the semiconductor market amid a rapidly and dynamically evolving AI market.

First, I would like to talk about how we see the business environment. Although there are still some unclear factors such as the continuation of intellectual risk and the rapid risk of change in exchange rates, AI-oriented applications continue to be the main growth drivers of the semiconductor market, based on the rapid and dynamic evolution of the AI market.

For calendar year 2026, we estimate a SOC tester market size of $8.5 to $9.5 billion, representing year-on-year growth of approximately 30% at the midpoint. Growth is expected to be fueled primarily by continued momentum in AI applications, boosting unit volumes, and greater device complexity, both of which drive stronger SOC tester demand.

The SOC tester market size in the year of 2026 is estimated to be 85 billion to 95 billion dollars, and it is expected to grow by about 30% in the previous year at the midpoint. This growth is mainly due to the continuous expansion of applications for AI, and the expansion of semiconductor production quantity and complexity of semiconductors, leading to the expansion of the SOC tester demand.

For memory testers, we estimate its calendar year 2026 market size to be $2.2 to $2.7 billion, which represents year-on-year growth of approximately 20% at the midpoint. We expect continued solid investment appetite for testers across a broad range of memory devices.

Thank you.

Thank you very much, Doc. I would like to start the question and answer session.

Mr. Yoshida of CLSA Securities Japan, go ahead, please.

This is Yoshida from CLSA. Conversations on the very strong results and also the guidance. Could you please provide a bit more detail of the breakdown of the SoC tester market, Tom? For example, for the exposure to AI area, comparing to 2025 and 26, can we expect all the growth is just from AI towards 26? And also, is this AI demand can be broken down to the volume contribution and also the intensity, the test hour contribution? And additionally, I'd also like to ask you about your market share assumptions in 2025 and 2026. Thank you.

Congratulations. I'd like to ask you about your market share assumptions in 2025 and 2026. I'd like to ask you about your market share assumptions in 2025 and 2026. I'd like to ask you about your market share assumptions in 2025 and 2026. Also, regarding AI, I would like you to tell us more specifically about whether the test intensity is the driver or how much the test time is, etc. In addition to that, I would like you to tell us more specifically about this growth driver. Also, I would like you to tell us more about the premise of market share. Thank you very much.

Okay, thank you for the question, Yoshida-san. This is Doug. So first of all, from an SOC perspective, we continue to see the largest subsegment obviously to be in the high performance compute or AI area. Roughly in 25, we see that as around 80% of our SOC. business, and we see that continuing into 2026. So that, as I said in my prepared remarks, will continue to be the main driver. From a test content or test intensity perspective, we continue to see the complexities drive more test insertions and higher test times. So we see that trend continuing.

Thank you for your question, Mr. Yoshida. This is Dag. I would like to talk about the SOC that you just mentioned. In the SOC, HPC and AI are the major factors, and 80% of them are limited. This situation will continue until 2026. Also, if this field becomes a major driver in the future, Thank you.

And then finally, on the market share part of your question, we will release the final calendar year 25 market share information in April. But I can say in broad strokes that we do believe we picked up a fairly substantial market share in SOC. And we see that trend continuing into 26 as well.

Thank you.

Thank you very much, Mr. Yoshida.

Next, Mr. Yamamoto of Mizuho Securities. Yamamoto-san, please.

Thank you for your comments on how to look at it for the next 25 years and for the next 26 years. That's all.

So you mentioned that for SOC business, 80%, around 80% comes from HPC slash AI related business. And I'm wondering if you could break it down further between GPU versus ASIC for 2025. And also if you could share your outlook toward 2026 would be appreciated.

Okay, yeah, thank you, Yamamoto-san. This is Doug again. We don't disclose beyond just the categories with the computing and communication. So that basically, the 80% really is inclusive of the high-performance compute. you know, the AI accelerators plus most typically the application processors. And I think it's, you know, fairly obvious that the majority of that is in the HPC side of things right now. And we see that continuing.

Mr. Yamamoto, thank you very much. I will answer from the bottom. Our company's policy is that we will not open any further breakdowns. We are still opening at the level of computing or communication. As I said, 80% of it is related to HPC AI, but among them are AI accelerators, high-performance computing, HPC, And then as to the GPU ASIC part of your question, I'm actually going to ask Sanjeev to answer that part of the question.

This is Sanjeev Mohan. Thank you for your question. As Doug mentioned, our market share is very strong and it's continuing to grow. In FY25, there was clearly a shift more towards custom ASICs. So obviously, because of our market share and presence, our business with those customers also grew. But having said that, GPU continues to be very, very strong and still the number one driver for the business. As far as the forecast is concerned, we expect that trend to continue. We will see more business with custom ASIC providers but the GPU business should remain very healthy. I think that's to the extent that I would comment at this point.

Thank you very much. Regarding GPU and ASIC, I will answer from Sanjeev Mohan. Thank you for your question. As I mentioned earlier, the market share in the market continues to be very strong. In the 25th year, I think that the shift to custom ASIC has progressed. The reason for this is that the customer's business business in this field was also very popular. However, the GPU trend is also very strong, and I don't think it will change to continue to be the number one driver. From the point of view of prediction, we see that this current trend trend will continue in the future. Custom ASIC is also very strong, and we see that the GPU is also very healthy.

Thank you very much, Mr. Yamamoto. Next, Mr. Wadaki of Morgan Stanley MUFG.

Wadaki-san, go ahead, please.

It's a really great record, and it's been released every time. It's a similar question, but it's been estimated to grow by 30% in the 26th year of market growth. What kind of weak points do you see when you divide this into factors? AI ASIC, Custom ASIC, and GPU test time increase, and number increase. Can you give us an image of what kind of order it will be in terms of these three factors?

Well, you announce strong earnings every quarter, and I'm surprised each time. My question is somewhat similar. So you mentioned that TAM 426 will grow by about 30%, and I just want to understand how reasons can be broken down. Let's say three factors, GPU test time, growth of custom ASIC business, And volume growth, so if you have these three breakdowns, how would you rank the contributions for potential time growth for 26?

Thank you, Oedaki-san. I think the biggest driver is just going to be the additional unit volumes for the custom ASIC. We've said before that the test content between, you know, sort of the typical custom ASIC architecture is very similar to that of the GPU accelerator products. So they also have a lot of test content. But in terms of what's driving the trajectory is really the unit volume and the more adoption of the custom silicon versus traditional GPU.

Thank you for your question, Mr. Obadaki. I think the biggest driver is the number of unit volumes. I think the number of custom ASICs is the biggest driver. As you can see, the architecture of custom ASICs is similar to that of GPU accelerators. In that sense, I think there is a lot of test content. Thank you very much, Mr. Wadaki.

Next, Mr. Nakamura of Goldman Sachs Japan.

Go ahead, please.

Thank you very much and congratulations once again on the strong results. I just wanted to also ask about your calendar 26 market outlook for both SOC and memory. What kind of scenarios are you assuming for both the low end and also the high end of your guidance for both SOC and memory? And it looks like the strength will continue, especially on the AI side for a while. So based on the conversations that you're having with your customers, is there any visibility that you can discuss on calendar 27 and beyond?

Congratulations! I'd like you to tell us about the 26-year market outlook and SoC Memories. What kind of premise do you have for the scenario of the upper limit and the lower limit? AI is likely to continue to be a hot topic, but can you tell us what kind of visibility you have over the past 27 years based on your conversations and conversations with customers?

Okay, thank you, Nakamura-san. This is Doug. Starting with the SOC market, on the demand side, it's very strong. And so when we look at the assumptions on the low and high side, A lot of that is driven by the sheer supply side economics. And I would say that from wafer starts to coalesce capacity to memory availability is really going to dictate the low and high range. And so it's a little bit out of the control of our company. And, you know, so long as there's enough wafers and advanced packaging and memory, there could be a scenario at the higher end of that.

on the sorry yeah pause for translators If you look at this situation, it is a high-end situation, but this year it seems that the supply side is the driver.

The demand continues to be strong, and the supply side decides the high-end and low-end. I don't know. I don't know. . . . . .

And then on the memory side, it's a little bit different. It's still going to be supply related in terms of how much memory can be supplied. But in this case, there are new factories, as you know, being built, and those really by all of the DRAM suppliers. And those are coming online later in 27 and into 28. And until that time, there's a lot of just reshuffling of wafer starts, reallocation towards the AI market. And so depending upon how well the memory suppliers are able to optimize those wafer starts toward the high-end areas, that's going to really dictate where we wind up in this range.

Yes. Regarding memory, the situation is a little different now, but I think that the reason for the large supply side is that it is similar to a single-seater. To explain in detail, the D-RAM manufacturer, the D-RAM supplier, is now promoting a new FAB construction plan, an expansion plan, but I think that such an expansion plan will actually stand up and operate. I think it will be around 2027 or 2028, but in the meantime, they have a challenge of shuffling their wayward allocation as a wayward strategy, and how to optimize the allocation. Among them, there is a wayward allocation that is more high-end, for example, for AI. Thank you very much. Is there any visibility into calendar 27 and beyond?

Yeah, I mean, it's too early, but kind of in general thinking, a lot of the customers have, as you know, have talked about being sold out all the way through this year and into next year. So that bodes well for the test capacity that they'll need. And coupling that with increasing complexity year after year, that also bodes well for 27. But a little too early to say at this point, Nakamura-san.

Thank you, Mr. Nakamura. I think it's a little too early to talk about the future of 27 years. However, from a general point of view, what many customers are talking about is that from this year to next year, the situation will continue to be sold out. In that case, we will need more test capacity, and the complexity of devices will increase every year. I don't know how it will be in the future, but I think it's too early to say anything specific.

Thank you very much.

Nakamura-sama, arigatou gozaimashita.

Thank you very much, Mr. Nakamura.

Tsuzukima shite, B of A shoken no Hirakawa-sama, go shitsumou onigai tashimasu.

So next, the Mr. Hirakawa of B of A Securities.

Go ahead, please.

Watashi wa testa no seisan capacity no koto ni tsuite o ukagai shitai tomoimasu. The capacity of testers increased by 60% to 70% compared to the end of March of the 27th. In November of last year, the number of testers increased by 5,000. However, compared to the past 24 years and 26 years, the market share we received today is doubled. Moreover, the market share of NenSource C is 56% in 24 years, which is a rough calculation, but it is probably in the latter half of 60% in 25 years. Thank you very much.

Well, so my question has to do with tested production capacity. So I think originally your comment was that toward the end of FY26, the year ending March 27, you would aim for capacity increase of between 60% to 70% versus 2025. And I think that the number, the unit number that you gave out in November last year was 5,000 units. Hearing you speak about TAM today for 2026, which suggests almost like doubling pace, and just doing the math of market share, which could reach high 60% in 2025, I think that 5,000 units won't be sufficient. So what's your current thoughts on capacity? And also, where are you currently in terms of the capacity ramp? Where do you think you'll be at the end of March? Do you think that you'll be running at, you know, full utilization or full capacity?

Thank you, Hirakawa-san. Yeah, very good question. I think we recently held the fireside chat and I mentioned that we were going to target capacity of at least 5,000 systems by the end of our fiscal year 2026. And we're definitely on track for that and actually now are looking at expanding that even further because of some of the demand that we showed in the TAM that you recognized. So we said at least 5000, meaning that that would be. the low minimum area and we have the scalability to go beyond that certainly. We also mentioned that we were going to then take a step towards 7,500 systems within a couple of years and then beyond that to 10,000. So we may be in the position to have to accelerate that build out faster than we had originally planned. But that's just a matter of executing faster than we had anticipated.

Kirakawa-san, thank you for your question. It was a very good question. As I mentioned in the fireside chat last year, our goal is to have at least 5,000 annual production units by the end of 2026. This plan is currently going according to the plan. Currently, we are working on various initiatives aiming for further growth. As you can see from the TAM numbers, demand and demand are rising very strongly. Therefore, 5,000 units are the lowest line, and I think it will be a low-minimum number. I think we need to increase the number of these units with a new scalability. The next value of 5,000 units is 7,500 units over the next two years. Thank you very much. Thank you very much.

Thank you very much. Mr. Hirakawa, thank you very much.

The next question is from Mr. Shimamoto of Okasan Securities. From the TAM that you showed us this time, it seems that sales will increase by about 30% to 40% compared to this period. Is that correct? Is it more likely to outperform than that? I think that the review of the medium-term planning that you showed in the first half of the year will go to the next quarter, which will be much higher than that, but as an element that the situation has changed significantly, what kind of changes have occurred in the last three months? I've been talking about it so far, but I'd like to ask you again, so thank you.

Well, so my question has to do with your outlook for FY26, the period that will end in March 2027. The TAM figure that you disclosed actually suggests that FY26 revenue could grow between 30 to 40% year-in-year. Is my assumption correct? Is there a possibility that you could actually beat this, you know, grow even higher? And you gave an MTP update before, but I think it might be likely that your FY26 could outperform your revised MTP figures. So I want to know what's happened in the last three months. What's changed? What's different?

Okay, thank you, Shimamoto-san. So obviously, we don't do any FY26 guidance until April, but certainly looking at our TAM projections and using assumptions around market share, there is a case where there could be that kind of growth. which actually is in line with many other companies who are projecting into 26 right now.

Mr. Shimamoto, thank you very much. As for the forecast for the year 2026, we will not disclose it until April, but as you can see, we have made some predictions. We have made some predictions based on the market share and other numbers. However, the situation of growth has become quite clear, and I think that not only our company but also other companies will be able to make similar predictions.

And then as for the midterm plan, yeah, certainly we're surpassing lots of the goals we had set for the three years. And we'll be revisiting those numbers as we go into fiscal 26, which will be the final year of our third midterm plan. So please wait for that.

Yes, and as for the medium-term, I think that some of the goals that have been set for the past three years have already been exceeded. In that sense, it may be the last year of the third medium-term management plan of the 26th year of next year, so it may be time to review the numbers again. I would like to have a little more time on this.

And then for the final part about what's really changed the last quarter, I'm going to ask Sanjeev to comment on that.

Yeah, I think that's a really good question. I think, first of all, in the last three months, what we saw was a lot of pull-ins from customers in terms of their devices coming and ramping more quickly than we had thought. the some of the digestion period and the new product launches that we thought would happen early this year they were actually pulled in so i think that was one of the big factors as we look at the next year we talked about this a little bit earlier i think we can really point to the growth in custom asics as being one of the big reasons In 25, there were a lot of new designs that were being launched, and now it seems that in 26, they're going to be moving into high-volume manufacturing, and that is going to directly contribute to our business.

Finally, I would like to ask you about the changes in the past three months. Thank you for your questions. What happened in the past three months? First of all, I would like to say that there have been many changes in the past three months. Regarding various devices, I think that the ramping has progressed at a faster pace than expected. I was thinking about the adjustment phase and the release of new products at the beginning of this year, but I think one of the factors is that all of those things have been pushed forward. As for next year, it may be a little early to tell you at this stage, but basically, it is expected that Custom ASIC will grow. Thank you very much. Thank you very much, Mr. Shimamoto, for your question.

Thank you very much, Mr. Shimamoto, for your question.

Next, Mr. Shibano from Citigroup Global Market Japan. Go ahead, please.

Thank you for having me. I would like to ask you to explain the cause and effect of the overclocking of this SIMO machine. If possible, I think that the third and fourth SIMO machines are overclocked against the forecast three months ago, but as a driver, the forecast of the GPU, whether the GPU is even longer, or whether this SIMO machine, from the third SIMO machine, is already expanding to a custom ASIC tester, I would like to ask you about this area. Thank you.

So I think your second half performance beat your original expectation three months ago, not just third quarter, but fourth quarter will probably beat your prior assumptions. And is it mainly GPU driven, or is it the case that custom ASIC has already contributed to your better than expected performance in Q3 and beyond?

This is Sanjeev again. It's many factors. We were able to expand our production capacity much sooner than we thought. That was one of the big reasons. And we were able to resolve some of the constraints. Beyond that, the business across the board is strong. It's not just one application that I can point to and say this is what led to the second half beat. The whole high-performance compute market, whether it's GPUs or custom ASICs or inferencing devices they're all doing very very well and given our market position we we benefited from from the growth

Thank you very much. This is Sanjeev. I think there are various factors to consider. First of all, I think this is one of the reasons why the growth of our production capacity is progressing faster than expected. This is one of the reasons why supply restrictions have been lifted. Thank you very much, Mr. Shibano.

Mr. Yasui of UBS Securities, please.

My question is about China. And you were talking about very strong demand, deploying demand in December quarter. And the China revenue went up nicely. So you've been talking about very strong demand, customer demand. Does it include China, the customer? And then what is the outlook for the 26th China demand for the SOC? That's my question. Thank you.

はい、私の質問は中国に関する質問です。 第三四半期についてはプルイン、前倒しがたくさんあったというふうなお話だと思います。 第三四半期の地域別の状況を見ますと、 中国の売上が顕著な良い伸びを示したというふうに認識していますが、 Please let us know if China was included in the pull-in and pre-pandemic. Also, please let us know about the 2026 pre-pandemic.

Yeah, thank you for the question. Yeah, our China business remains quite healthy, around 20, 25 percent of the overall revenue. It's a it's an important region for us, both on the SOC side as well as memory side. And without getting into specific segments, I'll just say that it has a very robust Fabless community in China. So there's literally thousands of Fabless companies and we're leveraging our platform strategy on the 93K, which has served that market very, very well and has led to some of the good revenue figures that you're seeing.

Thank you for your question, Mr. Yasue. Business business in China is in a very healthy state. We are currently closing 20% to 25% of our sales. In that sense, the Chinese market is a very important area for us in terms of SOC and memory. We can't talk about SMIC in detail, but we can say that there is a very strong and strong Fabless community in China. As you can see, there are thousands of Fabless companies in China, and we use our platform approach of 93,000 people to deliver various products to our customers. This is one of the reasons why our sales are so high in the Chinese market.

Thank you very much. And a little bit of a follow-up question. The 26th demand includes the AI-guided customer in China. It's going to be expanded into the 26th? That's my last question. Thank you.

フォローアップの質問をさせていただきたいんですけれども、26年の需要についてなんですけれども、御社が見渡している26年の需要の中には、中国のAI関連のビジネスも含まれているかどうか、ヒントいただけませんでしょうか。

I think for that part, I think the most we could say is that there is a variety of device segments in the SOC side in China. So they serve everything from the consumer electronics and mobility and high performance compute. So that's all baked in also into our overall global TAM numbers, but nothing out of the ordinary when it comes to China.

Thank you very much. Yasui-sama,ありがとうございました。 Thank you very much, Mr. Yasui, for your question.

Although we have received additional questions, I would like to now conclude today's session.

Thank you very much for taking your time to join us in our financial briefing despite your very busy schedule today.