Advantest Corporation (6857) Earnings Call Transcript & Summary

I don’t have the excerpt itself in your message—only the placeholder “<<foreign_lang>>” appears where the speaker turn should be. Please paste the exact excerpt you want rewritten, and I’ll return only the rewritten version with every fact, number, name, and product term preserved and no 5-word verbatim runs from the source.

Just as a side note, our yearly output capacity goals, naturally, aren’t the same thing as what we actually sell.

<<foreign_lang>>

We’ll keep pushing ahead with the strat—strategic transformation of our supply chain to support possible growth down the road. At Advantest, we see our scale as an edge over competitors.

<<inaudible>>

That brings my prepared comments to a close. Thank you.

We're going to start the question-and-answer portion now.

Okay. First question comes from CLSA Securities Japan — Yoshida-san, go ahead, please.

Yeah, this is Yoshida from CLSA. Congrats on the strong earnings results. Um, regarding the guidance, I get that there are some uncertainties, but given how strong your SoC tester sales progress was in the March quarter, I figured your tester market outlook might be revised up even more. Uh, so could you give us more detail on the background behind this new SoC tester market forecast, like your, uh, market share assumption? And if there have been any changes in inquiries recently that aren't yet reflected in this market outlook, such as demands from GPUs and custom ASICs, could you also, uh, give us some color on those? Thank you.

<<foreign_lang>>

Alright. Yes, thank you for asking, Yoshida-san. I believe we're forecasting fairly solid year-over-year expansion between 2025 and 2026. As reflected in the TAM figures we shared, we keep making headway when it comes to gaining share. In certain instances, constraints arise not on the tester side but simply from how much capacity is available in foundry and packaging. All of that is already accounted for within our TAM projections. Much of it comes down to outside forces. Looking ahead, as we keep growing within the industry, that's exactly why we're continuing to grow our production capacity so we can match the capacity increases happening in the market. I hope that answers it.

<<foreign_lang>>

Are there any shifts in the near term regarding the production focus of your AI-related customers that haven't yet been captured in this market forecast?

<<foreign_lang>>

This is Sanjeev Mohan. We don't really observe the changes as such, but what we do observe is the forecast becoming firmer. Obviously, back in January when we did this, we were estimating — and applying our judgment based on what we were seeing at the time. But now, visibility has clearly improved, and we believe it is firmer than before.

<<foreign_lang>>

Thanks a lot.

<<foreign_lang>>

Thanks a lot, Yoshida-san.

<<foreign_lang>>

Moving on now to Mr. Yoshioka of Nomura Securities. Please go ahead.

<<foreign_lang>>

<<foreign_lang>>

<<inaudible>>

Apologies for the Japanese. I have a question about your silicon photonics business. During your presentation, in the highlights, you mentioned that you want that very large order for the high-volume production tester area for silicon photonics. Looking ahead to silicon photonics, what kind of timeline would you anticipate in this area, and what business opportunities do you see there? I'd also like to know the current status as things stand. What is your present market share, and what sort of level are you expecting, if I may put it that way? Competitors in this area are also very strong. Please let us know how you can have a catch-up plan or try to, if I may say, innovate the market share. I'd like to understand the potential of the silicon photonics market.

Yes. Yes. All right, thank you, Yoshioka-san. Well, to begin with, there's no catching up we need to do, since in our view we're the leader in this category. That said, pinning down the market size remains hard. This is a very new business area. Quite a few uncertainties exist around the various test insertions. Test times, as well as yields, are still being figured out. So this year really marks the first one in which some volume production will actually be put in place. Then, as far as trajectory goes, we expect next year to more than double. And following that, at least another doubling. So our view of the market size is very bullish, yet it remains very hard to determine just how large it will end up being. I think, you know, we'll be doing some work over the course of the year, and my hope is that we can return later this year with a more accurate report on the TAM size. It's just a bit too early to say.

<<foreign_lang>>

One final point I’d like to make: part of what makes us a leader comes from choosing the right collaborators. These test cell solutions for silicon photonics are fairly advanced, so we have been quite intentional about whom we picked as partners, and that has proved a genuine advantage for our solutions.

<<inaudible>>

<<inaudible>>

Thanks a lot. That was quite clear. I really appreciate the very detailed explanation.

<<foreign_lang>>

Much appreciated, Mr. Yoshioka.

<<foreign_lang>>

Moving on, we'd now like to take the next question, and it comes from Nakamura at Goldman Sachs Japan.

Once again, congratulations on the impressive results. I'd like to ask about the capacity outlook for your SoC testers. I believe you said before that the plan involved growing that to 5,000 units by the close of fiscal March 27. Now you're indicating that capacity expansion is being sped up to 10,000 units. So I'd like to see when exactly you intend to grow capacity to that 10,000 unit mark? And also, you've roughly kept your guidance for calendar 26 SoC tester. Yet you are expanding your capacity plans. So I'd like to see what reason sits behind this.

<<inaudible>>

To make the numbers clear: we had said we'd reach at least 5,000 systems per year by the close of this fiscal year, and we're running ahead of that plan. We also stated that a few years down the road we'd be at 7,500 systems per year, and we've pushed hard to pull that forward. What I wanted to make clear today is that we're now taking yet a third step — targeting 10,000 systems per year, with readiness hopefully by the end of 2028, though more likely in 2029. That covers the first part of your question. On the second part: we have to plan this capacity more than a year ahead. We're looking out to 2027 and 2028 and even further. We're quite bullish that the market will keep up its high growth rate. To have the right supply chain and production capacity ready, those decisions have to be made roughly a couple of years ahead of time. That's what's behind these repeated step functions moving higher. Hope that helps.

<<inaudible>>

Thanks. To follow up on that, what figure did you report for SoC historic capacity as of the close of fiscal March 2026? And would it be reasonable to say that by March 2027, it will land somewhere around 5,500 or perhaps 6,000 units, roughly along those lines?

<<foreign_lang>>

No, Nakamura-san, I don't think we ever said publicly how much production capacity we'd add over this past year. You can sort of infer from the buffer what our capability is. Our focus, I'd say, is more on gazing ahead toward the 5,000, 7,500 and 10,000 steps we're working on right now.

I don’t have the excerpt to rewrite—only the placeholder “<<foreign_lang>>” came through, with no actual speaker turn text. Please paste the exact excerpt you want reworded, and I’ll return only the rewritten version, preserving every fact, number, dollar amount, percentage, product name, and person name exactly while avoiding any 5-word verbatim run from the input.

I really appreciate it.

<<inaudible>>

<<foreign_lang>>

Thanks very much, Nakamura-san.

<<inaudible>>

Moving on, we'll take the next question from Mr. Shikanai at JPMorgan Securities Japan. Please go ahead.

Okay. My question concerns replacement demand compared with new demand. Since SoC testers can, in theory, be utilized across multiple generations, how should we view the path — new demand versus new system demand versus replacement or upgrade demand?

<<foreign_lang>>

Okay, this is Sanjeev once more. We anticipate that demand for new systems will keep going, since at present we remain in a growth phase. And we expect this to persist for at least the coming 2 to 3 years. Past that, we think that as these devices become increasingly complex, they will need upgrades. So at that stage there may be an upgrade cycle for the systems, yet we still expect overall use of AI and of tokens to keep growing for several years. So yes, at present the majority of our business consists of new systems. At some future point, 2 to 3 years out, there could be more upgrade business. But right now it is mostly new systems. As for worrying about it, will we ever plateau, given that these systems can be used for a long time, we really have not observed that. New catalysts for the industry and new segments always emerge, and that is not something we are concerned about at this point.

<<foreign_lang>>

I'd just like to make one more brief point, which is that unit volume will keep increasing — you know, as wafer starts and things like CoWoS capacity and new memory fabs come online, unit volume is going to keep rising. And then there's a second piece, which is that test content keeps going up. So because of those two factors, there's always, uh, a need for more capacity. And within that capacity, there are certainly upgrade, uh, opportunities for things like, uh, you know, power supplies and so on. But, uh, on the whole, the trajectory of, uh, unit volume times test content gives, uh, two factors, uh, that drive, um, new capacity being added for test.

<<foreign_lang>>

Got it. Thanks a lot.

<<foreign_lang>>

<<foreign_lang>>

I appreciate that, Ms. Shikanai.

<<foreign_lang>>

Yes, the next question comes from BNP Paribas Securities Asia Limited. Alex Chang, sir, go ahead please.

Yes, thanks for giving me the chance to ask this. Actually, I noticed the company turned in a really solid full Q, because with the SoC tester, it looks like gross was above what the memory tester delivered. Is that why your margin came in better for the full Q? And a follow-up: with the guidance you've given for next year, you're assuming an operating margin that's largely flattish. Based on your TAM projection, the SoC tester is still going to grow faster than the memory tester. Is that a bit conservative when it comes to margin expansion next year? Thank you.

<<inaudible>>

Thanks, Alex. That's a solid question. For Q4, I'd say we had an ideal combination of product mix and licensing, and that's what pushed the gross margin up so high. Obviously, SoC is one of the elements involved. Inside the memory portfolio, we carry several different memory systems, and the ones we shipped during that quarter sat at the upper end of our margin range. On top of that, licensing revenue was strong. It all converged in Q4, and that's not necessarily a signal of what we'll deliver consistently as we move into next year. We're also dealing in the market with some inflation, and, as you know, we use a great deal of memory in our systems too. We're being careful not to overstate certain assumptions, since our bill of materials is being impacted by some of the macroeconomic factors as well as supply constraints specific to the memory area.

<<foreign_lang>>

Thanks. Could I go ahead and ask my follow-up question?

Sure. Yep.

Moving to my second question, which concerns the competitive landscape. Your peers, I believe, said earlier that they're aiming for roughly 15%, 20% market share in the GPU market over the midterm. So I was wondering: when it comes to your customers — their purchasing and supplier strategy, and how they choose a particular GPU model — will they bring in a dual supplier, or will it just be, you know, one dedicated supplier for a particular GPU model? And if your peer, you know, becomes more aggressive on pricing, will you feel pressure from that, and will it affect your margin? Thank you.

<<foreign_lang>>

Yes, Alex. I know about these kinds of discussions. Unfortunately, we don't comment on what specific customers do. All I can say is that we remain very comfortable with where we stand in the market with GPU customers and other accelerator customers. We just keep our heads down. We do our work and let the numbers speak for themselves, and that's all I'll say about it.

<<inaudible>>

I'd like to add something too — the pricing you brought up is very important, and it's one variable. We always need to stay very aware of it and remain very competitive, which obviously we've already been doing. As for dual vendor strategies: yes, of course, as the business has grown larger, customers worry about assurance of supply. So we always want to stay ahead of that. There are other reasons as well — technical capabilities matter a great deal when selecting a platform, and obviously we have a strategy to stay ahead of the competition there. Lastly, the dual vendor strategy can benefit us in certain areas too. We should always factor that in as well.

<<foreign_lang>>

Thanks, Alex.

Thanks a lot. And once more, congratulations on those strong results.

<<inaudible>>

Thanks, Alex. <<foreign_lang>>

Our following question comes from David Dai at Bernstein. Please proceed.

Thank you very much. David Dai here, from Bernstein. I appreciate your time in taking my question, and congratulations on a strong set of results. There are two questions I have. The first concerns the opportunities in CPU testing, since where AI data centers are concerned, up to now investors and the industry have concentrated on GPU accelerators. More recently, with agentic AI on the rise, I think both Intel and Arm have each made a big case that CPUs will become a larger contributor incrementally when paired — as we're seeing more CPUs paired with GPUs and accelerators. I'm wondering what kind of impact that will have on Advantest's tester market, and whether that has been factored into your market outlook for this year as well as the future capacity expansion to 10,000 units?

<<inaudible>>

Yes. All right. Thanks, David. That is a genuinely solid question. And yes, you're correct. At this moment, you know, since the shift toward heavy inferencing is underway, CPU usage is going to rise. Through intensive math along with pre- and post-processing, there are cost and power advantages. All of those factors are pushing activity across the data centers — more agentic tools, you know, invoking software that runs on CPU. Yeah, CPUs, LPUs, TPUs — every one of those fits into, you know, an enormous migration happening right now. You know, we're positioned very strongly for that segment of the market. We have traditional customers, longtime customers, who have had us in place for decades. We've been quite deliberate with certain fabless accounts worth noting, where we maintain a large partner ecosystem working on the fabless side. Since the V93000 is really a de facto standard across the foundry and OSAT space, we collaborate very hard with those partners to maximize the value of our systems, you know, in the OSAT and outsource space. Lastly, you know, these devices remain very complex. Once again, leveraging our partners to build the full test solutions is critical for this area as well.

There's been talk, I think, about inferencing requiring orders of magnitude, or even several orders of magnitude, more compute than training. A great deal of that has been taken into account in our 2026 forecast. But we see that continuing through 2027 and beyond, you know. And that is, again, the reason, uh, we were so aggressive when it came to expanding our production capacity. Um, so hopefully that helps.

<<foreign_lang>>

Thanks. That was quite useful. Now, my second question is about demand in China. As everyone can see, China is building, or needs to build, its own ecosystem for accelerators as well as HBM memory capacity. As they do that, they are obviously going to require additional tested capacity too. I just want to confirm that, with Advantest, we can expect Advantest to hold a similarly strong position in China as well. If you have any visibility into the timeline for when that might take off, that would be extremely helpful. Thank you.

<<foreign_lang>>

So yeah, on China — that's a key market for us, and it always has been. We keep taking part in the high-end SoC market there. China is a complicated situation, honestly. Domestic competition is growing up in the country, lots of it. And there's clearly a geopolitical piece too, with decoupling and the regulatory environments. So we're navigating all of that, but we still feel we hold a leadership position when it comes to our technical capabilities, which are needed for the high-end SoC market. So we plan to keep competing as much as we can in China. And you're right — there's a lot of activity there in new device applications, and some pretty exciting things are happening there.

<<foreign_lang>>

Thanks a lot, and congratulations once more.

<<foreign_lang>>

<<foreign_lang>>

Thanks a lot, Mr. David.

<<inaudible>>

Much appreciated. A few more questions remain, yet we need to wrap up today's session at this point. Many thanks to everyone.

<<inaudible>>

We appreciate you making the effort to be with us for this financial briefing even though your day is quite full.