Teradyne, Inc. (TER) Earnings Call Transcript & Summary

Okay. We'd like to get started. Good afternoon, everyone. My name is Toshiya Hari. I cover the semiconductor and semiconductor capital equipment space at Goldman. Very excited, very happy to have Tyler Warren from Teradyne. He's the CFO of the Semiconductor Test business. First of all, Tyler, thank you for coming.

Yes. Thanks for having me. Appreciate it.

Really appreciate your time. I think many of us are relatively new to you. So I was going to kick off by asking for your background. You've been at Teradyne for about 18 years. But maybe you could spend a couple of minutes talking about your background, your key responsibilities as the CFO of the Semiconductor Test business?

Yes, absolutely. So I'm Tyler Warren. I've been at Teradyne for about 18 years. I started there in 2006. I went to school at Bentley in Waltham and started at Teradyne right out of school. So I've done a number of different jobs at Teradyne. One of the things that's kept me there is the ability to do multiple things and in multiple different areas of the business. So I was in finance for a while. The role I had prior to this, I was the General Manager of our System-level Test business, which is an important part of the semiconductor test landscape. And then for the past 4 or so years, I've been the CFO of the Semiconductor Test business at Teradyne. And I'd say there's a lot of responsibilities. There's -- you got to do compliance, you got to do all that type of stuff as the CFO but the main thing is making sure that we have a robust business model, a business model that can support our customers, that we have the right application of resources to our R&D pipeline, that we're keeping up with our customers' road map so we have the right applications and support to grow our customers -- grow with our customers. And so making sure that we do all that, but also have attractive financial returns. Those investors in the room, everyone wants to have growth both in the top line and in the bottom line. But to do that, you have to have a robust business model that keeps up with our customers' requirements. And so it's that balancing act of managing the finances, but also managing the ability for us to invest where we used to for top line and bottom line growth.

Okay. Great. I appreciate that. I do want to spend a bunch of time on the individual markets within Semi Test, but since you are the CFO, and since you talked a little bit about the R&D pipeline, I did want to ask you on strategic focus, what you guys are spending your time on, what you're spending your R&D dollars on. When we look at the total company, to Teradyne, I think the OpEx to sales ratio in '24 is probably going to be the highest since 2013. We're aware of all the strategic investments going on the robotics side but how are you spending your time your R&D resources in Semi Test?

Yes, great question. And I think to kind of couch that, what I'd say is we're really excited about the market. And it's the market growth that we're investing in to try to capture. And if you backed up to kind of 2018, 2019, the total ATE market was somewhere around $3.9-ish billion. And today, in 2024, it's around $5.1 billion. So there's been this kind of growth in the market of $1 billion to $1.1 billion. And so our focus is on making sure that we're attacking the areas of the market that are growing. So there's investments we're making across the broad portfolio of our product line, both in R&D and in applications and support to make sure that where there's growth in the market in places like compute what we call VIPs, vertically integrated producers in memory, in automotive, that we have a product road map that's attractive to our customers, that we have the sales and support to win that business on the ground. And that ultimately, in this market share is really sticky. And so getting designed in early and riding that wave as the market grows, it's hugely important to our business model. And so we're spending a little bit more now than we have in the past. But if you look at our long-term model, we expect to be in the 28% to 31% OpEx ratio by the time we get to 2026. And we're investing to be able to facilitate the top line growth to get there.

The mobility TAM within Semi Test, or SOC test is down, I believe, more than 50% peak to trough, $2 billion-ish in according to your guidance, your forecast roughly $900 million this year. There's a lot of excitement around AI, in particular, with the S24 by Samsung earlier in the year, you had Apple come out with announcements earlier in the week. I'm curious how you guys are thinking about the mobility test TAM next 1 to 2 years, say, as AI becomes a bigger driver in the marketplace?

Great question. Maybe just to back up a little bit. Historically, we've been quite strong in the mobility space. And like you said, the market back in 2021 and 2020 was in the $1.8 billion to $2 billion range. And it's been cut more than half from that to $800 million to $900 million this year. And at the same time, the compute market has grown from $600-ish million to almost $1.6 billion this year. And so we've been historically very strong in mobility. That's been great for us for a long period of time. The market shifted in the past couple of years to be much more compute-centric and less mobility-centric. And so some of the investments we talked about are to make sure that we can capture that market share in a growing compute area, while protecting our investments in mobility where we already have high share. But we do expect the mobility market to grow. It's coming off a trough. It's $900 million last year, maybe $800 million in 2024. And there's a couple of things that we think are going to facilitate that growth. One is there's been a reallocation of test capacity from the mobility space over to the compute market. And so that's soaked up a bunch of excess capacity that was in the marketplace in '23 and '24. That's part of the reason why the market was a little bit lower in those years than in past years. There's a lot of capacity put in place in 2020 and 2021 and we've been soaking up that capacity as we got through the last couple of years. So some of that reallocation to the compute space is going to tighten up utilization and mobility, and that will help the market grow again as we get into 2025. Unit growth, units are up this year. If you look around, there's an expectation that units will grow a little bit in 2025. If that happens, that's another catalyst for the mobility market going forward. And then I'd say longer term, maybe not next year, but in the future, one of the big things that we think will help the mobility market is kind of the AI and AI edge capability that's going to get put into smartphones. And when that happens, there'll be both an increase in the test complexity of a mobility chip. So it will take longer to test the same chip tomorrow than it does today. That drives incremental test capacity. And then if there's this application that people want in a phone that they can't get without the newest model, there's an opportunity for us to drive the refresh cycle down. It's been kind of expanding in the last couple of years, how quickly people go and get their next phone. If that shrinks a little bit, that could drive unit growth in the future, too, and that would be a benefit for the mobility market. So over the next couple of years, those are kind of the things we see driving the mobility market moving forward.

Okay. That's great. I guess I have a similar question in mobile, but a little specific to a customer, a customer that used to be very large for Teradyne. I think they peaked at 25%, 30% of revenue at one point. Revenue from that customer is down quite significantly over the past several years. As we look ahead, is it pretty much units complexity, the insertion point of N2, are those the kind of things that we should be watching out, looking out for? Or are there other things at play like maybe what they do in PCs or servers as well?

Yes. No, you're absolutely right. They were much more meaningful revenue contribution in 2020 and 2021. obviously, they're a sub-10% customer for us this year. But I'd tell you, without talking specifically about a single customer, the things I talked about in mobility are the same drivers that we would look for that customer to expand their test capacity needs. So it's really around what's the complexity of VIP, what's the complexity of the chip that's going in the phone. That moves to N2, that should be a jump in transistor count and that should drive incremental test capacity needs. And then unit count, units grow, there's a need for more test capacity. So those are the things that we watch out for.

Okay. Great. You talked a little bit about the compute opportunity, the VIP opportunity. I think on recent calls, you've sized the VIP sort of market, if you will, in 2026 at $400 million to $600 million. What's the makeup of that number of that forecast? Is it mostly custom CPUs and accelerators? And how would you characterize your competitive position within that market? Again, historically, I think, compute, you were under-indexed, but going forward, could we expect higher share?

Yes. So just if I take a step back for a second. The compute market historically has not been a place we have very high share. And a lot of it's born out of the x86 architecture was never a stronghold for us. We have focused much more mobility. That was a very good market for us over the past decade and the compute market was much smaller. It was $400 million to $600 million. That shift that's happening is what we're paying attention to right now. And VIPs, vertically integrated producers, is the place that we think we have the best ability to go gain share. And these are going to be guys that are going to be in-sourcing the actual chip design and then specifying the manufacturing and test equipment out into the test world. So think of these as the guys you've probably heard; Amazon, Google, Microsoft, those are all what we described as vertically integrated producers. They are in-sourcing that silicon. So the makeup is everything you could think of. It's custom GPUs, it's custom CPUs and it's custom AI accelerators. So -- and there's other things in there too, like ADAS chips that we characterized in that part of the market. Going forward, the growth is going to be in all those areas. Again, we think that market may be $200 million today, and we have the ability to win about 50% share in that space. And so our compute share right now is in the low 20% range. So as the VIP market grows from $200 million this year to $400 million to $600 million by 2026, and we're winning 50% this share of that market, our compute share is going to start to grow quite significantly as we get towards 2026. And that's what we're excited about from a growth perspective.

Okay. And in terms of market share, are you at 50-ish today? And then -- or are you sort of aspiring to be at 50% over the next couple of years?

So it's a noisy market. And so we track it in 2 separate ways. One is like who's buying, what do people actually spend on test equipment. And for that, we're about 50% this year. And you can see in our numbers, like compute revenue for us in the first half of 2024 was greater than all of the compute revenue we shipped in 2023. So we've seen a significant step-up in compute revenue and a little bit of share in 2024 already. So that's what we're seeing.

Okay. Okay. Got it. Shifting to automotive, industrial, the broad-based semiconductor customers. That market also is down, I believe, from $700 million at its peak to $500 million this year. Industrial is down, I think, about 50% from $600 million to $300 million. Many of your customers on recent earnings calls have spoken to signs of stabilization, maybe an uptick exiting the year going into next year. What's your -- maybe that's the wrong question. What's your near and medium-term outlook in both industrial and automotive?

Yes. So we think both are really good markets. The industrial market and the automotive market are both places where it's historically been quite strong. So we have 40% to 50% share in the industrial market, same in the automotive market. But they're both quite weak right now. The industrial side of it is more tied to macros, PMIs and things of that nature, down across the board in the U.S. and in Europe. So I think we'll watch for as the kind of broad manufacturing environment continues to recover into next year, we should see some growth in that market, and we're well indexed to the industrial market. I think the automotive market is a much more interesting story, I'd say. There's a bunch of tailwinds in automotive that we're really excited about, and we have really good share in the automotive space. So things like the transition from the combustible engines to EVs, like the semiconductor content that goes into an EV car is twice that of what's in a combustible engine car. So as that transition happens, both here and overseas, we're getting a lot of semiconductor growth in the space that should drive a bunch of test capacity going forward. So that's one area that we're excited about. ADAS is another. The ADAS volumes are not high yet, and there's not a lot of like fully self-driving cars obviously out there right now. But the test intensity of ADAS and the fact that it has to adhere to automotive quality requirements, drives a huge test intensity to that chip. And so we've seen revenue for ADAS chips this year, and we expect that to be a meaningful growth driver for the automotive market over the next couple of years. And there's things around that, that also make a difference, too, like battery management systems. So the device revenue for battery management systems is growing at 30% a year. It's off a small base, but it's growing at 30%. So that's a meaningful contribution as we get towards 2025 and 2026. So you put all those things together, we've got an automotive market that we're really excited about from a growth perspective. We've got a share position that we think we can defend that 50%-plus going forward. So it will be -- it's a big part of our earnings model as we move into 2026 and how we're going to grow into that model in a place where we think we can be really successful.

Okay. And Tyler, obviously, you've gone through many cycles historically. Just a quick follow-up on automotive and industrial. Typically, the timing at which your business, the semi test business recovers. I presume there's a lag vis-a-vis when your customers are seeing their business pick up. Is that the right way to think about the relative timing of an inflection? Does that makes sense?

Yes. It's more how they're planning for their capacity, right? I mean if they're going to see an uptick in units they have to look at their planning cycle and they have to say, "Okay, what's my test capacity look like either in-house or at the OSAT, what's the complexity of that going to be, and so there going to be test time growth," all those components go into their planning cycle and then they would migrate that planning cycle to us. And our lead times are usually like the 1 to 2 quarter range. So we'll likely get visibility into that 1 to 2 quarters out. And so we're planning our manufacturing capabilities to be able to respond to an uptick both in automotive and industrial when it happens. But for the kind of next quarter, 1.5 quarters, it's pretty weak right now. On the industrial side -- I mean on the auto side, it's weak. There's a lot of capacity put in place in the last couple of years. And a lot of supply chain-driven inefficiencies that drove a bunch of inventory into the channel. So people have put in place a lot of capacity because of that. And so we're kind of moving through that soaking up that idle capacity right now. And then as the market recovers, we'll see kind of our orders in the kind of 1- to 2-quarter lag.

Okay. Okay. That's helpful. On the memory side, HBM has been a really big driver for you guys and for the broader market. I think the TAM is expected to increase nearly 5x from $100 million last year to $500 million this year. As you look forward, what are your expectations for the next few years? Your customers have plans to transition from HBM3 to 3E, 4 and so on and so forth. Would it be fair to assume that test intensity grows as your customers make those transitions?

Yes. And I think we're already seeing it. So the HBM market from a test perspective didn't exist before 2023. It wasn't anything. So in 2023, there's about $100 million of test capacity that got put in place. Like you said, that grew 5x, we think, in 2024 to about $500 million. And part of the reason is that test intensity. The test intensity of an HBM chip is somewhere around 5 to 10x out of a normal DRAM chip. And so that's why you see such huge market growth in this space. And I think that will continue having that outsized test intensity factors as we move forward compared to like normal DRAM chips. So our position right now in HBM is something we're really excited about. We've got share at a large supplier on the single die part of the test. So there's 2 test insertions in HBM. There's single die, which is more at the wafer level, then there's stacked die, which is more of a performance test. Right now in 2024, the market is kind of 1/3 single die, 2/3 stacked die. So we've got 2/3 of the market that we don't play in right now. We're only in single die test. And so as we look forward to 2025, we've got some big design wins that we're working on to get into the performance test side of it. And we think once we're successful in those, that will move our share back up in 2025 and in a growing market. And we've got kind of a couple of key product differentiations that are going to enable that. One is we've got better throughput for our tester. We can test more devices in parallel and faster than our competitors. So that provides a cost of test advantage that we can harness. And then the second is in memory, you have to be able to test that speed. And so as the speed that the device operates at moves up from HBM3 to HBM3E to HBM4, unless your tester can keep up with that speed increase, you'd have to go buy a new tester. And the tester that we have for the performance test design in is good for both HBM3E and HBM4. So that's a significant competitive advantage when we go try to win these design wins. And so we think we'll be successful in that area, and we'll grow share in both single die and in performance test in 2025 in a growing HBM market.

Okay. So just to clarify, the performance test, you guys sort of making that entry into the market, that's a '25 dynamic?

Yes. I think that will be '25 dynamic. We're in discussions now, we're going to design in now, but the time we get qualified and moved through that, it's likely a '25 dynamic.

Okay. Got it. On the NAND side, where historically you've been really, really competitive and from a test perspective, your customers are coming out of a fairly deep extended downturn. Pricing is starting to improve. Customer profitability is improving as well. What sort of the sentiment out there? Are you starting to see signs of a potential recovery in NAND test?

Yes. So NAND or flash has historically been the stronghold for us in memory. It's back up a long time. We got into the memory business in 2009 by buying a company called Nextest, and they were a flash NAND test house. And so at that time, we had about maybe 4% share of the memory market. We've grown that, both in NAND and in DRAM, to about 40% share in the memory market. So significant share growth over the past 10 to 15 years in the memory market. But as you've all read, the NAND market is down, ASPs are down. It's a difficult pricing pressure out there. And so this year, the NAND market represents about 20% of the total memory market and DRAM is 80%. Usually, that's 50-50. And we have a very strong position in NAND. We're about 80% share in the NAND test market. The thing that's held it up this year is actually, despite the fact that units are challenged and bit growth is not as high as it has been, its technology transition. So things like UFS 4.0 have driven wave of engineering capacity and the need to buy new test capacity to test that new interface. So that's kind of what's held up the NAND test market in the first half of this year. We think that capacity kind of played out in the end of Q2. So now we're in a little bit of a low until that transitions into production test capacity needs in 2025. So we think the NAND market will grow starting next year, but it's from a depressed base.

Got it. Okay. Looking back, I believe the semiconductor test market has marginally undergrown the WFE market over time through cycle. Given everything that we've just discussed on the SoC side and the memory side going forward is the operating assumption inside Teradyne is that Semi Test grows in line with WFE, above, below, do you not look at it that way? How do you guys think about that?

We try to look at it from a multiple number of angles, and one for sure is what's the WFE market or Wafer Front End market doing and how does test compare to that. And in the past couple of years, we have undergrown the test market. So the intensity of test, the WFE has shrank the past couple of years. And there can be a variety of reasons for that. They could be putting in capacity for leading-edge nodes that don't come online. So you've got the capacity going in WFE, but it's not outputting wafers yet. And then there's a lot of capacity going into China, and China is building up their own -- a new set of capabilities for that. So I think and our planning is going forward, at least for the next year or 2, that will flip flop and will outpace the WFE market from an intensity standpoint a little bit. But it can be quite noisy. If you do a correlation back, it's not always perfect one to one.

And Tyler, you mentioned this a little bit just on China. For your business, I believe local Chinese customers account for 5% and minus?

Yes, a little less than 5%.

A little less than 5%, which is very different from what the WFE guys are seeing today, 30%, 40%, 40% plus. Is that a function of, again, maybe it's timing, right? Maybe you're about to see a massive inflection in China? Or the investments going into the front end, it's a pull in, it's not necessarily tight output? Maybe it's more competition for you guys in China relative to the Applied/Lam scalers of the world. How do you guys process the differences there?

Yes, the China market is interesting. So first of all, it's a big test market. It's an important part of the business in a place where we have a very strong team. On the memory side, we have very good share in memory in China, share that kind of mirrors our global share, somewhere around 40%-ish. The big challenge for us on the SoC side is government regulations. A big chunk of the market is Huawei and HiSilicon. And back before government regulations, they were a 10% customer for us. So we have the ability to win share there. We have very good coverage there but we can't deal with them anymore. And so they're 30% and any given year, 30% to 50% of the SoC market is Huawei or HiSilicon. So the challenge for us and the reason that we have such low index to China is mostly a government regulation issue. We can't deal with some customers over there because they're on FDPR regulations or the entity list. Where we can, we think we have a very good team, and we're able to win share. There's automotive opportunities out there. There's custom ASIC opportunities out there, all of which we're competing for and have good progress. So there's nothing different about the China market where we can compete. It's just that there's parts of it, big chunks of it that we're not allowed to deal with.

Okay. And those customers are using local Semi Test providers and solutions?

Well, I do a test for a competitor, yes.

Okay. Interesting. You announced a strategic partnership with Technoprobe toward the end of last year. It was a fairly sizable investment for you guys. Just remind us what the rationale was, I guess, why did you feel as though you had to take an equity stake in the company? You could -- you guys were partners prior to that. So what catalyzed the deal?

Yes, it's really to facilitate strategic partnerships. And so just to remind everybody, we took a 10% equity stake in a company called Technoprobe. And Technoprobe makes the pins that go into the device interface, and that's the board that connects our tester to the chip handler. So basically it's the signal router between the tester and the chip and it routes signal back and forth and controls the electrical flow. So the reason that we thought that's a great idea is because there's -- they're the leader in that technology, and we think we can unlock horsepower in our tester that we couldn't otherwise unlock. And so our tester versus our competitors' tester, we have higher horsepower and more throughput. So if we can unlock that throughput, then we can have a competitive advantage in the marketplace. And so the combination of us and them working together on some of these strategic partnerships from a technology standpoint helps us better unlock that tester throughput that can win share in the market. And that's the investment thesis.

Okay. Got it. I'm going to pause here and see if we have any questions from the audience. If you do, please raise your hand. We've got mics going around.

The question is for the logic testing, right? On one hand, we see the non-die testing requirements getting higher given that chip they design so the wafer test requirement is supposed to be higher. On the other hand, again, the chips they design is getting more and more so the final test requirement is also growing higher. So can you tell us what kind of like testing dollar content increase in both area? And what kind of growth can we expect from those trends?

Yes. So the question is about chiplets and what the test intends to the chiplets are?

Yes, for both wafer test.

Yes. So what we're really talking about is advanced packaging, and advanced packaging has been around for a long time. When you put multiple chips and assemble them onto a die, what you get is a little bit of test intensity increase. You might get like a 10% bump because now you have to test each of the individual chips by themselves and then you also have to test the chip as a package. And so it's not, I would say, a huge driver. It's certainly an incremental benefit to us as more companies go towards a chiplet design that's going to allow them to get higher performance in their chip, but it's not a huge benefit.

Squeezing one more question. For the SLTs. Traditionally, SLT is not like performance hungry, i.e., that in general, the testing requirement is a lot lower for SLT. But recently, we do see the HPC or accelerated SLT demand is increasing quite a bit. Do you see SLT will be more performance-driven or still a cost-driven business? Are you targeting in this market?

Yes, SLT is going to be a very important part of the test flow going forward. And for context, we have an SLT platform that we developed internally. We brought that to market in 2017 and it's been a significant part of our business for the past 4 or 5 years. And SLT is really -- it's System Level Test, but it's meant to capture the last like 1% of defects or something much smaller than that after you go through ATE. And so where ATE does parametric test and you can only test a couple of devices in parallel. In SLT, you're just doing a functional test. So you're going to boot the chip, you're going to their runs applications. You're going to do a pass/fail test on it. And because of that, there's a lot less instrumentation that goes into it. And so the cost per unit is much less. And so I think as devices get more and more complex, there's only so much coverage you can get in '18, and you're leaving a number of transistors untested and as that goes from -- you've got 5 billion transistors on a die to 30 billion transistors on a die, you're leaving a lot more transistors untested. And that's the reason why SLT is necessary because you have to get that fault coverage to get DPPM down to an acceptable rate. So I think it's going to become a much more meaningful part of the business for us going forward. It's a place where we have a great platform and great customer contact and especially in HPC, where the number of transistors on a die is exploding, it's going to be an important test step for us moving forward and for the market.

Great. I wanted to ask about M&A in Semi Test and how you all think about cash flow return going forward. I think back in the day, 20 years ago, Semi Test used to be sort of the Wild, Wild West and you had a bunch of companies. It's very consolidated now. When you look at Teradyne over the past 5, 6, 7 years, pretty much all your acquisitions have been in sort of the robotics arena. Within Semi Test, how do you all think about potential M&A? Is that a consideration? Or do you feel like at this point, you're in a steady state?

Yes. No, it's absolutely a consideration. And it's not that we think we have a gap in our portfolio, to be adding a place in the market that underappreciated or we can add a technical advantage. It has to be something we bring to the table that is going to make it so that the company is stronger once we acquire them than they are independently. A lot of the reason that the M&A has been in the robotics space in the past 5 to 7 years is that there's a lot of growth in the robotics space. Historically, 5, 7 years ago, there wasn't a lot of growth in the semiconductor space. from a test market perspective. That's changed. That dynamic has changed. We now see a growing ATE market, as I described, it was $4 billion, 3 or 4 years ago, now it's $5 billion. So it's become much more attractive from an M&A perspective to go out and look at opportunities. But at the same time, we're really diligent on how we look at it. It has to have a good financial return, has to have a business model that kind of lines up with our business model. And there has to be some rationale for why it's a strategic fit for us and why we can make that company stronger, have a better return inside of Teradyne than outside of Teradyne. And we balance that with a balanced capital allocation plan. We reduced stock buybacks with the acquisition of the equity stake in Technoprobe. We measured that down as we rebuild our cash back up to kind of our $800 million floor. But certainly, going forward, we plan to have a mix of stock buybacks, dividends and M&A, both in that -- we look both in the test and the robotics space for M&A.

Okay. Great. In the last 2 minutes, I wanted to give you the opportunity to speak to anything that we may have missed or I don't think you spend too much time with investors and analysts, but to the extent you have anything as a collective unit we missed or underappreciate about the Semi Test business or Teradyne overall?

Yes. No, thanks. I think the things like spend 2 minutes talking about is just our growth aspirations in conjunction with our long-term model, right? And so there's 3 things that we're really excited about in the semiconductor market. One is the VIP space, where we've been under-indexed in compute. We think we have the ability in the VIP space to grow share at a 50% clip in that area, and that part of the market grows from $200 million to $400 million and $600 million by 2026. So that's a huge growth vector for us. And the second is automotive. We spent some time on this, but the growth tailwinds for automotive for the semiconductor test market. are large, and we have a very good share position in automotive. So that's a place that we think we can expand our share as well as grow with the market. And the third is memory. The HBM market is going to grow. We're going to gain share in the HBM market. And so we're going to have growing memory TAM and a growing share position. And those 3 things are going to fuel us to hit our long-term earnings model by 2026, which is 12% to 18% test growth for the company. So from 2023, we expect to grow between 12% and 18% on a CAGR basis into 2026, and those are the things that are going to fuel that growth. So I think that's the thing I'd like investors to appreciate is the growth tailwinds that are there and how we're going to prosecute them for success.

Great. Thank you so much for your participation.

Yes. Appreciate it.

Thank you very much and thank you all for coming.