Good afternoon, everyone. Thank you for attending the session here. I'm Samik Chatterjee, and I have the pleasure of hosting Teradyne. And from Teradyne, Greg Smith, who's the President and CEO. Thank you, Greg, for coming to the conference. Thank you to the audience as well for attending. Greg, I'm starting off most of my conversations with companies asked, and since we have the pleasure of hosting CEOs, CFOs here, you have regular conversations with your customers. It's useful to us to hear what you're thinking about the macro.
Mm-hmm.
There's a lot of investor concern about what the second half looks like in terms of how much of whether there's a big economic slowdown or not. I really want to get your thoughts based on your conversations with customers. Where do you think the macro is headed? How are you planning to run the business? Particularly given maybe the different scenarios that you see sort of happening in the second half of the year.
Sure. I think the situation is a lot more calm, but there's still a lot of uncertainty. When we were in the very first part of the year, like February, March, there was a rapid deterioration of customer confidence. In that period of time, we heard from customers that wanted to either reschedule deliveries or delay placing orders. Since sort of the middle of March, it's kind of stabilized in terms of that, it hasn't gotten any worse. There's still a great deal of uncertainty about what will happen in the second half of the year, but there isn't the rapid spinning that's happened. I think everybody is hopeful that things will calm down and we'll avoid a general recession, you know, like the tariffs leading to a demand destruction and then a recession.
but our customers haven't gained enough confidence that that is the scenario that will play out, that they've come roaring back, you know. So there's still, caution around placing orders, and we still are, like we have an unusual amount of uncertainty around the second half, you know, that we have, almost an equal number of upside opportunities and downside risks against the second half of the year. The thing that's unique about this year is that that range is far larger than it would normally be. So you asked sort of what are you doing to repair, to prepare. Teradyne's over 60 years old, right? So, this is not our first rodeo. we have experienced very cyclical business conditions in the semiconductor space. really, it was the norm until sort of post-2010. Things have been far calmer, but, we, we designed our business for, uncertain market conditions.
We have a highly variabilized business model. Our, you know, every single employee inside of Teradyne has a variable component of their compensation that's tied either to their growth or to profitability or both. If things turn down, we have an automatic mechanism that allows us to curtail some OpEx. We also are an outsourced business model. As business goes down, we're not carrying a lot of fixed overhead associated with manufacturing capacity. We have kinda designed ourselves so that our OpEx is moderated when we face more challenging business circumstances. The other thing that we have learned over a period of time is that we run a very conservative balance sheet, you know. So many peers run with more debt than we do.
What we've seen is that, because we want to be able to continue our investments, especially in research and development, through business conditions, we want to make sure that we run with adequate reserves and without an overwhelming amount of debt. Right now, Teradyne is very low debt, and we maintain a healthy cash position. If things get worse, we have an ability to operate and not change our strategy, not be forced to change our strategy due to current circumstances. That is sort of the downside protection, but I think we are definitely in a position to execute well, and we have a balanced upside and downside for the rest of the year. I guess if the macro gets bad, we are well positioned to do well through that. If the macro does not get bad, we are positioned for growth.
Okay. Fair. Maybe taking another sort of layer of that, there's been sort of very, very frequent changes on the tariff front.
Mm-hmm.
One, can you just outline, do you have any direct exposure on that front?
Mm-hmm.
Secondarily, tariffs have moved around, even including this week. Do you see customers pulling forward demand or pushing demand based on sort of these week-to-week, month-to-month changes?
Yeah, yeah.
Are you seeing customers make decisions that frequently and change those frequently? How is this sort of unpredictability of the tariff environment playing out in customer orders?
Yeah. Let me take those in backwards order. Let's talk about, like, pull-ins, and then after that, we can talk about the sort of the direct impact of tariffs on Teradyne. There is no, like, no significant pull-ins related to tariff changes that we've seen, that we haven't seen customers say, "Hurry up and get us capacity. We need to beat the tariff, the 90-day window." That has not been a factor. In any quarter, we will always have some push-outs, some pull-ins, and we have had some pull-ins, but none of that is related to tariff policy. It has to do with our customers meeting their customers' demands. Now, in terms of direct effect on Teradyne, what we talked about in our last earnings call is, the way that we see it.
Tariffs have a pretty minor impact on Teradyne in two ways. Depending on the quarter, between 10%-15% of our revenue is shipments to the United States. If those are shipments of, like, a semiconductor tester to an American customer, the tariffs are their responsibility. They are picking that up. Most of the tariff burden would be falling on those customers. There is a portion of the business that we do in the U.S. where we do final assembly and configuration in the U.S. based on stuff that we have imported from contract manufacturers. We have a minor impact to our COGS for a portion of that 10%-15%.
We also do all of our engineering in the U.S., but most of the material that we do the engineering work on has to be imported from contract manufacturers and suppliers. We also have a modest impact to operating expense. To put it into context, our expectation is EPS in a $0.41-$0.64 range for Q2. In that guide, there is about two cents of impact between that COGS effect and the OpEx. You know, it is a small but measurable impact to the overall results. As the year goes on, assuming that there is some consistency around the tariff policy, we expect to be able to increasingly mitigate more and more of that.
Okay. Got it. Maybe just to follow up there, based where you left off, given that there's some amount of capacity coming from the contract manufacturers imported into the U.S., and given the current tariff policies, are you looking to actively change the supply footprint relative to your contract manufacturers to get it to regions like Mexico where you probably get an exemption from that? Or would you generally sort of take a longer-term view and say, "The tariff policy, let's let it play out," and then we sort of actively work to mitigate some of that capacity?
I'd like to put it in a little bit broader context that, we're always working to try to increase the resilience of our supply chain. And oftentimes that means that we will be investing in production capacity, you know, to establish parallel production capacity for the same product line.
Okay.
For test, we feel like we have a very resilient supply chain and that, and the amount of business that we transact in the U.S. is not large enough that we would consider significant production, like supply chain changes to try to build more of that in the U.S. to avoid tariffs. For robotics, we are, it's a smaller business. It's only 13% of our revenue, but we are trying to increase the resilience of that, of the supply chain in robotics.
Got it.
We are actively looking to try and figure out as our, as that business grows, we'll need to add production capacity. We're trying to figure out where that production capacity should be. You know, if you think, like for overall for Teradyne, it's 10%-15% of our revenue is in the U.S. For robotics, it's about a third. It's a much more logical thing for us to start developing production capability in the U.S., but we don't have firm plans or dates for it. It's just something that we're looking at, and tariffs are only one reason that we would.
Okay. Before we go into any of the end markets, I mean, one of the things that's come up a lot through this tariff discussion is the need for more automation in the U.S. manufacturing.
Mm-hmm.
How are you thinking about the opportunity for your robotics segment to participate in that, particularly if capacity has to actually move?
Mm-hmm.
To the U.S., there has to be a significant amount of automation to take off the cost of labor disadvantage.
We have a lot of inbound around this topic, you know. Companies that are looking at reshoring are reaching out to us, and frankly, they're reaching out to our competitors and our ecosystem partners as well around automating stuff. The overall sales funnel for robotics is kind of 12-24 months. The process of developing a factory, like putting in a factory, building it up, getting it running, putting all the automation in, is also kind of a multi-year thing. We think that it's a potential demand tailwind over the midterm, but I, it's not like, "Oh, I need to move production to the U.S." All of a sudden, Q3 robotics is gonna be very, very high. There's gonna be a lag associated with it.
Got it. Okay. Great. Moving into the end markets, and particularly in relation to your decision to withhold the full year guidance at this time.
Mm-hmm.
Maybe just in the context of that, walk us through the visibility you have across the different end markets.
Mm-hmm.
What are customers willing to share with you in terms of their pipeline? How long of visibility are they willing to give you?
Yeah. So, you know, at the point in time when we had our analyst day in early March, we had enough information about the first half of the year that we were able to moderate our people's expectations for Q2. At the same time, we did not have a lot of information about the second half of the year, but we communicated the idea that, you know, instead of having this, like a really good 15% growth year, that it was more like 5%-10% growth. That was the picture that we had in the mid-March timeframe.
By the time we got to our earnings call at the end of Q2, we had, and at the time we said that we did that full year guide more around the idea of like, "Yeah, we know this is more than a Q2 problem, but we don't know the exact magnitude." You know, that, so we tried to make a judgment about that second half of the year. By the time we got to the earnings call in May, the thing that we were sure of is that we had this very wide uncertainty range. Just let me walk sort of segment by segment. In the automotive and industrial space, we had conversations with customers where they were actively asking us to push out forecasted demand for them.
We have an idea in our mind, but if there is a snapback in automotive, you know, like if tariffs are not going to impede demand and there is a snapback and people need capacity, there is upside potential there. If there is a recession from where we are, there is downside from there. In mobile, we were expecting a very modest 2025 anyway. We think that the big year for mobile is really going to be 2026 with the, really the introduction of edge AI, empowered smartphones, things like that.
We had a relatively quiet expectation for the year, and our main concern there is a downside risk associated with a general recession, you know, or like, and since, you know, like who knows what the tariff rules are today, whether there's an exemption for that product category or an exemption for a particular place of manufacturer. There's a lot of concern around that, but I think the main danger is whether there's like a crisis in consumer confidence, not something around tariffs. AI and compute though, as far as we can see, that is basically immune to the macro that's going on. The investments are over a longer period of time. People are showing commitment to that. The main uncertainties that we have, we still have uncertainties for the second half of the year, but they don't have to do with tariffs.
When it comes to compute, our revenue for compute is primarily in these vertically integrated producers, these hyperscalers, and that revenue is very, very lumpy. There is uncertainty about whether a big slug of that revenue will be 2025 or bleeding into 2026. That is another, you know, if it pulls in, it is a big upside for 2025. If it pushes out, it is a downside for 2025 with an upside for 2026. In memory, the primary X factor that we do not have a good read on is the transition from HBM3e to HBM4. If there is a rapid switchover towards HBM4, that would mean that the major customers that we have in this space would need to add test capacity. If HBM3e lasts more through the second half of this year, then those capacity adds would be delayed to 2026.
You know, looking at it, we see these ups and downs against every single one of the major segments. It is not a matter that we are more pessimistic about the year. It is just that we felt like we would be overstating our certainty by providing a, you know, like taking the average of those factors would give you a number, but it is not a number that we felt like we could confidently stand behind.
Okay. Okay. Moving to, or focusing more on AI compute.
Mm-hmm.
Beyond sort of the overall demand landscape, which continues to be strong, just talk about what are the other drivers that drive more test equipment demand relative to just the number of chips just going up, which is more of a volume increase. How do you think what the other drivers that increase the need for test equipment, over that longer-term cycle?
So you're talking exclusive outside of AI?
AI compute specifically, the work that you do for the VIPs. Beyond volume, what do you see as the other drivers for testing intensity?
In AI?
In AI.
Okay. The complexity is obviously a big factor, but probably the most important factor is the quality requirements. If you're trying to, if you're testing an AI accelerator and then that AI accelerator is going to be used in training, a latent defect in the device can lead to the waste of millions of dollars of compute time. The hyperscalers, the people who are building the servers, building the server farms, are driving very, very high quality requirements on merchant semiconductor providers and also themselves around these devices. The test intensity, even for a device of a similar comp, like if it's a GPU going into gaming or a similar complexity device going into AI, the test intensity could be 2X. Same thing is true for the AI accelerator. The devices are more complex, but they also have these very high quality requirements.
The other thing is that they have exceptionally high scrap cost. By the time you get to a full-up AI accelerator, you have a substrate, there'll be probably two compute die on it, there'll be six or eight HBM stacks on it, there may be other chiplets in it, all in an expensive package that's designed to deal with the 2 kilowatts of power. If that fails after it's all assembled, you're throwing away like thousands of dollars of in-process material. That is driving a shift left, you know, like they're trying to increase the test coverage upstream. That is an upside for the amount of business at, when you're doing wafer level test in the memories and in the SoCs.
The fact that it's complicated, that the yield loss is expensive and the quality requirements are high are the primary drivers for why that part of the market is growing as much as it is, both HBM memory and the compute space.
Got it. All right. You've talked about winning 50% of the sockets. When I focus specifically on the AI compute side, what is the difference when you think about those wins? What is the differentiation Teradyne brings relative to its closest competitor there?
Mm-hmm.
Do you see necessarily, you've talked at, even at the investor day about technology cycles and how that's offered Teradyne a lot of opportunity. Do you see that market share potentially inflecting higher based on the technology differentiation you're creating?
The advantage that Teradyne has is that we have a differentiated product. We have a better product. The disadvantage that we have is that our competitor has historically much higher share in the merchant part of the compute space. They have great reference customers. We have advantages. The advantages that we have, our tester is easier to design. First, the biggest advantage that our tester has is that it scales to be able to test larger parts or a larger number of the same part. If you can test to it once, you're more likely to be able to do that on a Teradyne tester than the competitor's tester. That gives a significant economic advantage. It's also easier to design the interface hardware. You need fewer layers. You get higher yield for the interface hardware.
Our tester is also very reliable in production. All of those things are these secondary factors that customers look at. It is not enough to dislodge an incumbent, but if you are talking about a jump ball, it gives us an ability to compete very, very well when we are either, you know, competing for the first time or operating in an environment where a customer is doing business with both us and our competitor. You were asking like, does that mean that we could get more share over time? I do not know. I mean, the thing that I will tell you is that business is so lumpy, and it is lumpy in two ways.
One is that there are very few end customers, but also the amount that the customers build of a particular device depends not only on whether their device works, but whether their device is significantly better than what NVIDIA built. You know, there are a lot of hyperscalers that have invested in developing chips, but when they benchmark what they designed and built versus what they can buy off the shelf, it's not worth them ramping their own product. Teradyne's been fortunate in having a couple of really good sockets that have ramped hard, and driven significant, not only like we won half and half in terms of the sockets, only a few of the sockets have ramped to high volume. We've been fortunate that a couple of those high volume sockets have been on our tester.
As I look forward, I think a lot of it, there's a lot that's up to chance, you know, whether the design is good, whether the design works, whether it ramps, and who's the, of the 50% that we win versus the 50% that our competitor wins, who hits the jackpot on a part that ramps.
Got it. Maybe talk about the acquisition of Quantify Photonics.
Okay.
What drove that? Strategically, what are you thinking in terms of the opportunity that you get as you sort of combine products with Quantify's technology?
Yeah. Let me talk about this generally in terms of the strategic importance of silicon photonics and co-packaged optics. In a data center, if you waved a wand and magically the copper networking was turned into photonic interconnections, you would get 10x as much bandwidth and a 30% power savings. There is a huge amount of interest in the end market to be able to go to this as the primary interconnection medium for AI compute farms for training and for inference. We see that this is gonna be a growing segment on its own, but we also believe that the company that has the best solution for silicon photonics and co-packaged optics is going to be in a position to gain share for high performance compute and AI. Okay.
We are investing to try and make sure that we establish that leadership position. We were very interested in Quantify Photonics because they are founded on the idea of the direction that silicon photonics is going. Higher lane counts and rapidly increasing data rates across the interfaces, they bring an ability to do higher lane counts at attractive capital cost versus the sort of instrumentation grade equipment that is being used today. Teradyne in 2012 bought LightPoint, and LightPoint essentially pioneered this idea of production optimized test equipment for the wireless space. Quantify is doing the same thing in photonics. It is production optimized equipment for high lane count photonics devices. We think that that capability is critical for us to gain share in the high performance compute space.
Okay. The target market here is again the VIP?
I'm sorry?
The target customer here is again the VIPs that you work with as they move towards more CPO adoption.
I think it's interesting because not only the VIPs, but the merchant silicon, you know, the AMDs and NVIDIAs and Intels of the world, this is an important enough inflection and enough of a technology disruption that it may open doors. We are not looking at this as just reinforcing within VIPs. We think that this could be a shift, like a ground shift in the high performance computing space that if you aren't, if you aren't great at co-packaged optics test, that you are likely to lose share in high performance computing.
Okay. Okay. Great. Memory test. You've typically talked about it as being divided between what's a more competitive high volume wafer testing, and then sort of the more low volume performance testing as well where you want to play. Right? Could you talk us through the various stages and where Teradyne is now primarily engaged with its customers? Where are you engaged with your current customers? Where do you see more opportunity in the memory test segment?
Yeah. The high level model that Teradyne uses to understand the memory market is DRAM and flash memory, and then wafer level testing of the memories and then perf, and then final testing of the memories. In the past, the memory market was quite balanced between flash and DRAM. Typically, if mobile is strong, then flash is strong. If compute, cloud compute is strong, then DRAM is stronger. We are definitely in a period of time where DRAM is the strongest in, it was like 80% of the TAM in 2024. That is the first division. Now, performance test, typically the way the memory market has worked pre-HBM is that the wafer level test was non-differentiated. The test time was dominated by the algorithms that had to be run and the tester did not need to do very much.
It just needed to present those test patterns and collect the results. The final test, when the device was packaged, that was when the performance test took place. It was, you know, it's still high volume, but it is much higher in capital intensity. That is where, and it's also where tester performance is much more critical to the delivery of value. You could get higher throughput or higher yield by testing on a better tester. Teradyne has established a leadership position in final test for both flash memory and also for DRAM. Now with the introduction of HBM, that performance test actually happens at the wafer level after the dies have been stacked up.
Teradyne has made great progress in terms of winning memory wafer test business on, like, when it's performance test for this kind of HBM memory.
Okay. Okay. Got it. Let me take a pause and see if anyone in the audience has a question. Any questions? Okay. Maybe if we continue to autos and industrial, and maybe before we get into the drivers itself, maybe give us some of the use cases of what the exact application looks like in the autos and industrial market for test equipment.
Yeah. So the two primary drivers in the auto and industrial space, one is actually AI. The buildout of data centers, it's, you know, it's estimated that a significant portion of all of the electricity in the U.S. is gonna be consumed by data centers. The amount of power in data centers is gonna go from like 50 gigawatts to 225 between now and 2030. That's an enormous load on the grid. There is an increased investment in the technologies to do power regulation and power supply into these areas. That is driving growth in power semiconductors, especially in, you know, exotic materials. Data center is actually a key driver of the industrial space. Now, the other driver for this space is the slow motion crossover between internal combustion and battery powered vehicles.
If you look at, if you compare the way the automotive market looked in 2022, the forecast in 2022 to the forecast in 2024, what has happened is the growth expectations or the volume expectations for gas powered vehicles has been revised down significantly. The growth projection for battery powered vehicles is basically the same in each of those forecasts. The key to Teradyne is that the semiconductor content in an electric vehicle is at least 2x what it is in an internal combustion engine. The number of battery powered vehicles is increasing every single year. We think that there is this TAM growth tailwind in this space.
It's, you know, obviously it's growth, it's growth cyclical because of the end market conditions, but we think that there's a good market here and we expect that that market is gonna grow at a healthy rate over the midterm. We're in a leadership position. We have like 46% in 2024 of this market. By the time we get out to 2028, we expect we'd be in the mid-50s. The market's gonna grow and we're gonna grow share because of that leadership position.
Okay. Got it. I know current macro headwinds are impacting your outlook here, but what is a typical through the cycle long term growth rate in autos and industrial in your view?
Oh, so if you look at like the growth rate through this midterm, yeah, we would probably expect that overall Teradyne, we're expecting about 15% CAGR growth through this midterm, 2024 out to 2028. Inside of semiconductor test, we expect the growth rate to be in that same neighborhood overall. In auto and industrial, I would expect the growth rate to be slightly lower than that, maybe between 10-15% over that period of time. Still healthy growth, but there are other parts of the business, especially like system level tests that are gonna have a much higher growth rate because it's coming off of a low, a low number here to a much bigger number.
Got it. Got it. To then finish up, maybe we move to mobility for a couple of questions. If in the scenario we do have a pullback in the macro in the second half and the smartphone market, or, responds to that slowdown in the macro with lower volumes, what do you see as offsets? Because you've talked about underutilization of test equipment in that space. As you think about then sort of the drivers in the second half, do you see any other offsets that would help sort of offset some of that volume decline?
I, you know, it's a tough scenario to, to game out what you just said because the thing that would cause a mobile slowdown would be a general economic slowdown. You know, like a, an erosion in customer confidence. The erosion in customer confidence is unlikely to be just mobile focused. It would probably also affect automotive sales. Right? We would probably be tracking to the low scenario in that space as well. I would expect that the factors driving the AI and memory part of the business are largely decoupled from the, from consumer confidence and a recession. I think that would, that would not be correlated with that downturn, but I don't think it would necessarily get stronger because of that, you know, like there's no offsetting effect there.
I think the upside would be more around like if we started to see leading edge around this reshoring or upside in robotics because of, you know, now even that I think, if there is a general economic slowdown, then you'd see a contraction in manufacturing capital investment. Like I don't see a lot of offsets to a recession.
Okay. Okay. And just finishing off, I mean, you did have a strong quarter by the way in mobility. What do you attribute that to? I mean, one of the questions we get often is why is not that an indication of pull forward by the segment? Just maybe clarify that before we wrap up.
Yeah. So, we, you know, we're proud of the results that we delivered in Q1. The primary driver of our upside was actually a relatively large order around a supply chain shift in mobile. You know, it's not that anybody's gonna sell more phones, but where the parts were coming from to go into those phones was coming through a different source and that source needed to add capacity to do it. They were interested in adding that capacity as quickly as they could. We were able to accelerate those deliveries around our operational capability. Our overperformance in Q1 wasn't really an indication of a strong rebound in end demand. It was more around meeting customer desired delivery dates by improving our operational, you know, sort of our operational capacity.
You know, looking into Q2, we, you know, in our earnings call, we guided our Q2 in line with our expectations we set at the analyst day time. And, you know, we expect, you know, we certainly don't think that the strength that we saw in Q2 was like the leading edge of a recovery. It's much more related to the one time factors.
Okay. Got it. I'll wrap it up there, but thank you for coming to the conference. Thank you to the audience as well. Thank you.
Awesome. Thank you.