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

Good morning, everybody. Welcome to the Goldman Sachs Communacopia and Technology Conference. My name is Jim Schneider. I'm the semiconductor analyst here at Goldman Sachs. It's my pleasure to welcome Teradyne and CEO, Greg Smith. We're happy to have you today.

Glad to be here.

Maybe just starting off, high-level strategic for a second. Greg, you operate a diverse business today, both high-performance test and measurement at its core. If you think about the exposure across semiconductor test, system-level test, including industrial robotics and other areas, how do you think about Teradyne's business mix over the long run? And what objectives are you trying to drive the business going forward?

So right now, I mean, Teradyne is sort of a tale of 2 businesses that we have these test businesses, and we've been in that business since 1960, right? So it is a mature market, highly penetrated. And the way that, that business really develops is through the introduction of new technology and segment shifts and share gain and loss in that space. And right now, for that test business is both product test and Semi Test, we are in a sort of the beginning of a significant growth period, driven primarily by AI, but it extends across all the different segments of product test and semiconductor test. The robotics business is a much newer thing. We've been in that business for 10 years. We got into it in 2015 with the acquisition of Universal Robots. It's a much earlier-stage business. So we are in a part of the robotics space called advanced robotics, high AI content, more flexible, more adaptable than sort of traditional industrial robotics, but it's an early-stage business. So we're in that primarily to capture the secular growth that we expect to happen in that space as physical AI becomes more of a thing. So it's less predictable. It is a sort of a chaotic early-stage market, but we believe that we have some really strong platform and technology advantages that will set us up for growth over the long term. But if you take a step back, I think you have these 2 very different businesses with a very similar setup over the midterm that both of them are set up for significant growth over the next 4 to 5 years, but one is going to be driven primarily by the end market changing towards the AI space and the other one by it being an early-stage market that's poised for more growth.

If we get back on stage 5 years from now and look back, what do you think is the one thing that investors are going to be most surprised by?

I think people are going to be kind of stunned by how strong Teradyne is in compute. So Teradyne has historically been very, very strong in mobile and automotive in the test space. And we've been historically much, much lower share in the compute part of the market. But over the past couple of years, we've invested significant amounts of energy in R&D, sales and marketing and applications to try to capture new customers in this space. And I think if you look back 5 years from now going backward, people are going to be like, wow, where did that come from? It's like whenever they do the Grammys and they talk about like best new artists, when they interview the best new artists, they're like, I've been seeing in clubs for a decade. That's sort of, I think, we've been doing the hard work to set up that share gain. And I think we're going to be looking back and seeing that it's occurred over that period.

Interesting. Interesting. Over the last couple of months, maybe can you talk about any kind of notable inflection in terms of the demand outlook that you see or more fully the visibility you have into the overall business?

Well, the -- like we have more visibility, but I think the important thing to recognize is that our sort of the normal cycles of the semiconductor test space have really changed. If you look back a couple of years, our business cycle was very much driven by sort of the ebb and flow of consumer demand. So people buy new phones for Christmas, for Lunar New Year. There's this huge peak in demand. And so people could sort of set their clock in terms of when the testers would be put in. Now with the rise of AI and especially cloud AI, that's driven by the capital budgets of these -- of the hyperscalers and these huge companies. And so it is far more related to whatever schedule they set versus anything in terms of the actual calendar. So it's become -- like we have a pretty good idea in terms of which customers we have and the business potential associated with them. We have much less certainty in terms of the timing of that business than we used to.

Interesting. Got it. And maybe you've coined a term called VIP, it's kind of -- maybe talk about what that market consists of what are the subsegments of that demand? And maybe talk about where do you think your share of that market is today and what you hope to achieve in the 2-, 3-year plus range.

So when it comes to -- so the term VIP stands for vertically integrated platform. And it's really around the notion of cloud computing and AI. And we -- a lot of people talk about this as hyperscalers. But we felt like that was a somewhat incomplete term because there's also these huge investments in AI going into vertically integrated platforms and things like automotive. So we wanted to try and broaden that scope a little bit to capture customers that were behaving the same way, which was they wanted to control the entire offering all the way from silicon to either cloud services or products in the end market. So -- and if you look at this space, it didn't exist like 5 years ago. There was essentially nothing. The only player in the space that was offering a complete vertically integrated platform in compute was really Google with the sort of early generations of TPU. But now all of the other cloud service providers are really getting into this space, and some of them are scaling to a significant fraction of the merchant GPU folks. And if we look in 2024, we thought there was about $400 million of testers that were bought for these vertically integrated platform customers. And we think that, that could grow to like $800 million or more by the time we get out to 2028. So it's growing rapidly. It's still not the majority of the whole compute space, but it's certainly a part of that space that's growing fast.

In terms of the market share?

So if you look at -- so we aggregate VIP compute with merchant compute and networking into something that we call the compute space. And our share in compute is really for merchant compute is very, very low. For networking, it's very high. And for VIP compute, it's kind of 50%. If you put it all together, it's sort of in the -- in 2024, it's probably 10% to 15%. In 2025, it will be a bit higher than that, but still in sort of the teens. And looking forward, over time, we expect that we'll be able to continue to increase share from there.

Got it. Now you provided the kind of initial 2026 outlook back at your Investor Day earlier this year. We're getting closer to 2026 now. So maybe give us any kind of incremental update, if you have any, on what your expectations for next year might be.

So to just sort of flip back to what I said, predictability is the thing that's really missing right now. And if you look both from a TAM perspective and also from a revenue perspective, there are a bunch of very high-volume devices that are positioned right on the cusp between 2025 and 2026. So we're not sure whether -- like how that will influence the TAM and revenue in 2025. And we certainly don't know how 2026 will develop. And just to sort of put it into context, coming into 2025, both us and our major competitor have the size of the compute space at about half of what it's going to end up being. So it's a very, very difficult market to predict because you have a small number of devices with really strict quality requirements and a lot of advanced technologies, both in the device itself and also in the packaging technology and the HBM memories and everything else. So there's so many moving parts that there are a lot of X factors against the TAM. The one thing that I'll say is we expect the overall market, both memory and SOC to be stronger in 2026 than it was in 2025, but we can't really put a number on it.

You think about your -- those more traditional markets, SOC and memory test, where do you think your share is at now? And then which one you have more confidence in incremental share gains going forward?

So right now, we are -- I think we are positioned for significant share gain in memory from 2024 to 2025. And then we think that will probably continue at a moderate rate through the rest of the midterm. For SOC, our share has been coming down as the whole market pivoted to AI compute and the large merchant GPU makers really leaned into huge investments, and we didn't have access to that. So we're kind of in the 30% to 40% range against the SOC part of the market. And we believe that we're positioned to grow from that base going out in time, both because of share gain in compute, but also because -- the other elements of SOC, the mobile part of the market and the auto and industrial part of the market are at a pretty low point right now. And so we expect a segment-related recovery in both of those. So we think there's a lot of share tailwinds going into 2026 and beyond.

And then as we head in the back half of this year, how do you think about the sort of step-up in your VIP business going forward? And then maybe as it tracks in the beginning of 2026, talk about the sort of mix between sort of -- in terms of that piece of your total SOC business going forward?

So there's no doubt that AI and networking are kind of the hottest parts of the whole test equipment market right now. And we think that there are a number of high-volume sockets in both the networking space and the accelerator space that are right on the cusp between end of '25 and beginning of '26. So in aggregate, they're pretty impactful. But in terms of exact timing between Q4 and Q1, we really don't have a good picture right now.

Okay. And then on the merchant GPU space, you've talked about sort of the prospects for getting into a very large potential customer there. How is the visibility on that customer count going? And sort of do you expect driving meaningful scale of that customer if you're able to win it next year? Or is that something that's more of like a beyond next year, beyond 2026 opportunity?

Yes. So like just to be as clear and careful as I can be. We have not won anything yet. Let me tell you a little bit about the sort of general motivation for this. So the merchant GPU players have generally had a single-source test strategy, and that was to make their operations as efficient as possible. They have a few different SKUs that they're trying to ship against. They're not sure what the volume will be for each one of them. And so by having a single platform, they're able to utilize that fleet much more effectively. That's a really good strategy when you are a minority player in the overall ecosystem when TSMC has much bigger customers and the OSATs have much bigger customers. But now that merchant GPUs are such an important part of the AI accelerator space, they're actually a significant portion of the total demand for test equipment. And a big factor for both of those players is supply chain resilience. So they want to be able to be sure that they'll be able to get capacity of HBM memories, of packaging [indiscernible] of test services of test equipment. And so that is the motivation that they're looking at in terms of qualifying multiple vendors in this space. And so we're attempting to do that. And we think that they have an interest in setting up a competitive environment between us and our competition. And the question is timing. So when you're trying to qualify against these complex parts, it takes a while. And I would expect that since they have a broad range of parts that are on a competitive platform, we would be looking at a gradual increase in share, not like a flip the switch and all of a sudden, it's an even split. But I think there will be revenue next year, but I don't think it's going to be -- it's going to be a slow climb in terms of market share.

Fair. Okay. Maybe switch gears a little bit and talk about your industrial robotics business. And there's a lot of renewed optimism for that business, I think, in the investor community, particularly around a key marquee customer there. When do you think that business becomes meaningful at that customer in terms of revenue? Is that something next year? Or is it beyond that time frame?

So I think just to set a little bit of context, at the same time that we're very excited about this large customer opportunity, the end market macro is terrible for robotics right now. So if you look at our robotics business, our strongest region is Europe. And our strongest segments are automotive and metal and machining. If you like wanted to pick the worst places to be, that's kind of it, right? And so we are in the midst of trying to reposition that into more faster-growing end segments like logistics, pharma, semiconductor and electronics, and that work is ongoing. We're also trying to do work to try and focus on solution providers and very large customers. So even though we are really excited about this large customer opportunity, we've got some hard work to do between now and when that starts to contribute meaningful revenue. So in 2025, we're doing that hard work, and we'll have sort of early revenue to support new product introduction work at that large customer. Next year will be sort of the beginning of volume deployments, but it's not going to be until 2027 that it's a very significant part of the revenue mix.

And given the visibility you have right now, you said, obviously, there's some cyclical weakness across this end market broadly speaking. So how do you think about your ability to grow that robotics business in '26? And then obviously, it sounds like you have more confidence in '27.

Right. So in '25, it's really all about trying to constrain our OpEx and rightsize the business on a path towards positive operating margin. So we're focusing on getting the investment against this new large customer and to focus on getting into these new segments and trying to be as careful as we can about spending everywhere else. So really focused on spend discipline. Into '26, we think that the large customer will ramp. And we also believe that the other work that we're doing in logistics, pharma and electronics and semiconductor will start to pay off. So we expect that 2026 will deliver modest growth, and we expect stronger growth in 2027.

Great. With respect to mobile, I think you previously talked about sort of $800 million TAM as being kind of the right ballpark of.

Yes, for '25. Yes.

Yes. Do you think that's still like a reasonable expectation? Do you see that as a growth TAM going forward? And maybe you could just frame for us your exposure to Apple and sort of the key drivers that you see for that business and that customer going forward?

So I won't comment about any specific customers, but just in terms of mobile, in 2025, we think this is kind of an $800 million-ish TAM overall. And just to sort of set context, that's pretty bad. It used to be kind of more like a $2 billion TAM was kind of a good year for mobile. When we look forward, we think that there are prospects for our TAM recovery in mobile starting in 2026. And that's really driven around 2 factors. One is if you look in both the Android ecosystem and the iOS ecosystem, there is a significant inflection in compute power that's coming to phones in the late 2026 product introductions. And in order to deliver that inflection in compute, there's a flip to latest process technology, so 2-nanometer gate-all-around and also new packaging technologies and memory technologies that allow them to get into thinner form factors with wider memory bandwidth. So what that means in terms of this change in terms of the memory technology is that the test intensity per part is going to go up significantly. So even if the unit volumes in the mobile space stay about the same, the amount of test per device is going to go up a significant amount. So we think that, that's a setup to drive the need for additional capacity in that space in 2026. The question is kind of how much. And that's going to be dependent on the unit growth. So if there's -- like if everybody in the world gets really, really excited about foldable phones in 2026, then that could be a really good tailwind. If non-Chinese phone makers really struggle in China, that could be a headwind. So there's these puts and takes against unit volume that we're really not sure. So we think that mobile will be better, but we don't think it will be back to like the 2020, 2021 boom days.

Yes. Maybe one go and kind of finish on memory market for a second. DRAM has been the lion's share of the market for quite some time now given -- especially given the HBM inflection. How do you think about the sort of split of DRAM and NAND in the business and the market specifically? And do you expect that mix to shift further in favor of DRAM or maybe a little bit of a resurgence for NAND?

So I -- like inside of Teradyne, we have sort of this constant argument about like should you measure things as a mix, like percentage of DRAM, percentage of flash because like you've got a numerator and a denominator that are both moving, right? And so the way we tend to look at it is, okay, what's the size of the DRAM market? How do we think it will grow? And what's the size of the flash market and how do we think it will grow? So the DRAM market is hot, hot, hot in 2024, it's plateaued in 2025, but we think it's going to grow robustly as you get into HBM4, 4E and 5 and LPDDR6. Those are all going to be significant growth drivers. So the DRAM market is going to grow, but from a very high base. The flash market, in 2026, there's going to be a transition in the mobile flash market to new next-generation protocols, in both Android and the iOS ecosystem. And the demand for solid-state drives for AI applications in the cloud is really going strong. So I think flash is going to grow faster from this low, low base. So if you look at it from a mix perspective, right now, it's probably like 90-10-ish in terms of DRAM to flash. I think more normal would be sort of 80-20, 75-25. So I think we'll be going towards that, but it's not because like DRAM is going to shrink. It's more that both of them are going to grow, but flash is going to grow faster.

NAND can recover from an awful level.

Right. Yes.

Okay. And then if we think about system-level test for a second, what are the moving pieces there? And what's the growth trajectory you expect heading into next year? And how do you think this market is headed over the next few years?

Yes. So inside of our -- so we took our hard disk drive test group and our system-level test group. They've always been the same team of engineers, similar technologies, but it's addressing a couple of different markets. It's addressing the hard disk drive market, the solid-state drive market, the mobile SLT market and the compute SLT market, okay? So those 4 things put together are in what we call the Integrated System Test group, and that's part of our Semi Test division. So Semi Test division does semiconductor test, ATE and the system-level test stuff. So if you look historically, that group has in the like 2020, 2021 timeframe was big, $250 million kind of big. And in 2024, because of the slowdown in the hard disk drive market and the mobile market, it was down in sort of the $85 million range. We think that, that is going to recover to above its prior peak through this midterm. So we think that, that business could easily be 4 or 5x bigger in -- towards the end of this midterm as it is now. So -- and that is going to be spread across all 4 of those buckets. So HDD is something that was big, has come way down, is going to come back both because of TAM recovery because of bit growth rates and also because we think we have the opportunity for share gain in that space. Solid-state drives is a space that we've had very little share, but we think we're positioned to grow our share. Mobile is a case where it is very, very low right now, but we believe that the inflection in complexity is going to drive an increase in the overall TAM, and we think we're positioned for share gain. And then finally, for compute SLT, this is really a new area for us this year. But with AI accelerators driving towards the high-level quality that they need to achieve, we think that, that is positioned to grow sort of 0 to a significant tens of millions of dollars over the next few years. So like it's kind of crazy to talk about 4x-ing that business, but it's going from $250 million prior peak to $400 million, not from the sort of low point that we are right now.

Yes. Maybe a couple of financial questions, if we could. On the gross margin front, you've executed very well. If you think about your kind of long-term horizon of kind of 60% gross margins, -- what are the levers for you to get there? And is that just more volume or more revenue gives you better gross margin, which is kind of obvious statement. But are there any offsetting factors in terms of large customer concentration or other exposures that would kind of offset that in any way?

Well, the thing that I'd like to emphasize is that our gross margin doesn't change much with volume. So when we were in the prior peak in 2021 to sort of the low point in 2023, there's a couple of points of change across that, but it's not like a -- some companies with significant internal operations, absorption would be a really big thing. And you'd see gross margins go to hell when things dip. By the same token, as our business gets stronger, our gross margin doesn't get that much better. There's a little bit of a tailwind, maybe a point or so, but it's not like all of a sudden, you're printing money as the volumes go up. At the same time, we have -- the biggest factor, and it's really kind of -- the thing that I have to keep emphasizing with people is when you see our gross margin move around, don't think it's a trend because it's really related to the sort of the specific configuration of products that are being purchased in any particular quarter. And that mix changes on a constant basis, but it always tends to work out towards this mean of the -- like our model is 59% to 60%. Right now, we're kind of in the 58% to 59% range heading towards that. But I wouldn't expect a significant shift in that at all going forward.

Fair enough. And then relative to OpEx, how do you think about OpEx growth from here from a leverage perspective, is sort of 2x or 1/2 the rate of OpEx growth for every dollar of revenue growth the right way to think about the leverage?

That's -- at the top level, that's the best way to think about it. If you take a step back, one important thing to remember is that a significant part of our OpEx is variabilized. So -- and every permanent employee of Teradyne has a variable component to their pay, whether you're a repair technician or an executive, you have a portion like that's related to our profitability. So as Teradyne becomes larger and more profitable, that OpEx scales with it. So that's a part of this 50% that you're talking about. Now 10% growth on the top line, 5% growth on OpEx. Some of that is just in compensation. The rest of it is around increasing the parts of the business that need to get bigger as we get bigger. And that's really customer-facing sales, service applications. Our R&D spend would be pretty efficient as we get bigger. So we think we can constrain our growth to that half of the top line growth. It's a little bit harder to do when growth is low. If growth is 5% year-on-year, it's hard to constrain growth inside of that envelope. But if you have sort of 10% top line growth, that's sort of where 4% to 6% OpEx growth would be the right rule of thumb.

Then maybe just kind of closing out a little bit on M&A. You recently did the Quantifi Photonics acquisition. Maybe talk a little bit about what that brings to the business overall at Teradyne. But then if you could follow on, talk a little bit about your M&A strategy and whether you see kind of more potential for M&A incremental product areas and so on or areas that are additive to what you already have?

Yes. So Quantifi was a really important strategic acquisition for us. Right now, we are on the cusp of a very significant technology change in networking and in data center architecture. Right now, optical networking is primarily sort of site to site. And within the server rack, it's mostly all copper-based. Over the next few years, there's going to be a significant shift towards optical-based interconnections within the rack, and that is going to be really a disruptive -- like a disruptor in semiconductor test and also in product test. So the big thing there is the number of optical connections, sort of the number of channels is going up rapidly and the data rates that are being applied in this space are also going up rapidly. So when we looked around, we saw that Quantifi had the best technology around this high channel count optical test. And we saw that our semiconductor test equipment was going to need that capability and also our -- we had an opportunity to help expand them faster because of the exposure we have to CMs and ODMs through LitePoint and our production board test business. So we could help them grow faster in what their original target market was, and we could also use them as a technology core for our silicon photonics work inside of Semi Test. So we think that this is going to help us maintain the lead that we have in networking. And we also think that it is going to give us a positive differentiation in compute as optical interconnects get on to AI accelerators. So that's like 2-ish years out from now, but we think that, that's going to be a very important aspect of testing these high-performance devices.

An appetite for [indiscernible] beyond that.

So it's really kind of awesome. If you look back 10 years ago, Teradyne embarked on this advanced robotics business strategy because semiconductor was like especially the back end of semiconductor was very -- it was done. It was a mature business. Packaging technology was advancing relatively slowly. Foundries were on a certain stair step. What's happened now is with the introduction of AI compute, there's this race for compute capacity that is pushing advanced technologies into production very, very fast. So whether it's HBM memory or advanced packaging, [indiscernible] wafer on PCB. So there's all of these new technologies that are coming. And there is a tremendous amount of yield loss across that whole flow, upwards of 2/3 of the money that's going into an AI accelerator is turning out into scrap. So that's a huge economic opportunity. And so we're very focused on trying to find companies that are helping to solve problems in that space and where our M&A strategy can bring them into our overall portfolio, we're really excited to try and do that. And where we see technologies that could accelerate our road map like the Quantifi Photonics and frankly, the deal that we did with Infineon, that helped accelerate our road map for advanced power semiconductors by 18 months. So we're really interested in acquisitions that will help us become more competitive in our core markets and that's where most of our energy is going right now.

Fantastic. I think with that, we're out of time. Greg, thank you very much for being with us today. We appreciate it.

All right. Thank you.

Thank you.