Yahoo Finance

Gregory Smith

Thanks, Traci, and thank you, all, for joining us today. I'll start off by summarizing our fourth quarter and full year 2024 results and provide some context for our initial view of 2025. Then I'll provide context around our updated midterm earnings model. I'll describe the trends we expect to drive the market and Teradyne's strategy to drive highly leveraged earnings growth through the midterm. Sanjay will then go into greater detail on all of these topics.
Our fourth quarter came at the high end of our guidance range, as trends we noted previously continued through the end of the year. Cloud AI has been the dominant driver of our semiconductor test business. And we have seen some short-term improvement in the mobile space, driven by supply chain shifts in our customer base.
In industrial and automotive, our fourth quarter benefited from customer-specific equipment purchases. Strength in our test business more than offset the continuing weakness in the industrial automation market, which impacted our robotics business. In 2024, after two years of semiconductor test market declines, our SOC and memory test revenue grew 17% year over year, excluding DIS.
AI was the dominant driver of our growth, specifically AI accelerator ASICs, networking, and HBM DRAM. We have previously described a class of customers called VIPs or vertically integrated producers. We use this term because these customers develop custom silicon to provide differentiation in their end products, whether they are phones, cars, or cloud AI computing.
In the first half of 2024, we saw VIP strength for edge AI in automotive. In the second half, strength was driven by Cloud AI compute VIP customers. Our goal in 2024 was to achieve 50% market share in computing VIPs, and we believe that we achieved that goal. This is particularly notable because much of the VIP test demand in 2024 came in the form of upgrades to systems left underutilized by the weak mobile market.
If this demand had come in the form of system sales, our 2024 VIP revenue would have been more than double what we recognized in the year. At the company level, we grew 5% in 2024. If one excludes the divestiture of the DIS business, our total revenue growth was 8%. We grew earnings per share by 10% year over year and generated over $470 million in free cash flow.
Our full-year financial results reflect an inflection in our business, both in terms of semiconductor test cyclical recovery, but more importantly, a successful pivot to diversify our customer base and reduce customer concentration. In 2020 and 2021 timeframe, our business was dominated by mobile with high customer concentration in that market.
Back then, we were highly exposed to mobile in SOC, memory, and wireless test. Now in 2024, the compute end market was a larger component of our revenue than mobile, as our SOC business in the compute market grew more than 3.5 times the prior year. We have been investing to capitalize on the secular shift towards VIP ASICs. And that yielded roughly 50% share in what we believe was around a $300 million TAM in 2024.
We have seen growth driven by our historical strength in the networking space. And we see opportunities in system-level test for AI compute. The pivot we have executed in SOC over the past couple of years is remarkable. In 2023, 11% of our SOC product revenue was in computing, and 51% was in auto and industrial. In 2024, 34% was in compute and 34% was in auto and industrial, a balance that underpins our longer-term model.
Recent advancements in AI inference, which appear to reduce the cost and time to develop AI applications, may be a catalyst to accelerate edge AI development. We think this could directly benefit the markets where we have historical strength, mobile and automotive.
It's early days, but we believe that lower-cost, lower-power. and faster time-to-market AI solutions can drive complexity growth and increased unit demand at the edge, which are key inputs for improving demand for test equipment.
Looking forward to 2025, we expect the SOC TAM to continue to grow roughly 7% year over year. While some of this growth is driven by AI compute, we expect a modest recovery in mobile, automotive, and industrial in the back half of the year. We believe that we are positioned to gain share in the low single digits in SOC test.
Now shifting gears to memory. In 2024, our memory business grew to over $500 million, up 30% year over year. Strength in the market and our growth was fueled by AI compute demand for HBM DRAM. In the second half of 2024, we were qualified for HBM performance test at a major memory supplier. Our higher throughput and forward compatibility created competitive differentiation, enabling us to capture significant share of the HBM performance test market in the second half of 2024.
We expect the HBM device end market to be strong through 2025. However, from a test equipment perspective, we are expecting the market to soften, as customers absorb capacity with higher productivity tools. We expect the HBM TAM to recover in 2026. As a result, we expect the entire memory test market to be flattish in 2025, although we do expect to gain share in the low single-digit range.
Beyond AI compute, we believe that there are other segments in the semiconductor test market that offer the opportunity for accelerating long-term growth. One of these areas is power semiconductors. These devices will continue to grow long term with the crossover to EVs and the demand for more efficient power generation, storage, and distribution.
We are announcing a strategic partnership with Infineon, the market leader in power semiconductors, to acquire their internal tester development team in Regensburg, Germany. This group will enable us to accelerate our road map in power semiconductor space, specifically, in areas like silicon carbide and gallium nitride at the scale needed to serve the automotive and renewables market.
While the semi test business was strong in 2024, Teradyne's other product test businesses, which include our system test and wireless test operating segments, continued to be impacted by weak end market conditions.
Within our product test businesses, we saw some programs push out from 2024 into 2025, but scored key program wins that we expect to drive healthy growth in 2025. We expect our wireless test business to return to growth in 2025 after securing 74 out of 80 tracked Wi-Fi 7 design win opportunities in 2024.
Turning to robotics, the industrial automation market continued to be weak in Q4. We typically see strong fourth-quarter seasonality, as customers place quick turn orders in the back half of the quarter. Visibility is inherently low in this high-turns business.
In Q4 of 2024, this seasonality was far more muted than in prior years. And we ended the year down slightly for UR and roughly flat from year. This underperformed our expectations but outperformed our industrial automation peer group. Despite the headwinds, there were highlights for robotics.
The UR channel transformation continues to progress with the OEM channel delivering 20% growth and the MiR large accounts also delivering 24% growth year over year in 2024. In the fourth quarter, as part of our multifaceted partnership with NVIDIA, UR launched its AI accelerator. Late in the fourth quarter, MiR's new flagship product, the AI-enabled MiR 1200 Pallet Jack, began shipping to customers.
And most recently, Teradyne Robotics announced a strategic partnership with Analog Devices to develop and deploy robots, AI, and software to support ADI's automation initiative. In 2024, we combined UR and MiR operations into a unified robotics operations group.
Now in Q1 of 2025, we are consolidating our go-to-market functions at the robotics level to enable our best partners to sell the full UR and MiR product line, and to serve our customers better with a single customer service organization. This restructuring increases our efficiency and reduces our robotics breakeven revenue from $440 million in 2024 to $365 million in 2025.
Looking ahead to the next four years, we are very optimistic. A year ago, there were questions as to whether VIPs would matter. And if they did, could we win their business. At that time, we thought the compute VIP market would be $100 million to $200 million opportunity in 2024, growing to $400 million to $600 million in the 2026 timeframe.
Our latest estimate is that the compute VIP market was $300 million in 2024. And the compute VIP market will be centered around $600 million in 2026 and could approach $800 million in 2028. We believe that Cloud AI will continue to drive share gains for us in SOC and memory. By the later years of this midterm, as AI moves to the edge for mobile enabled by process technology like 2-nanometer and gate-all-around, we expect robust growth of the mobile TAM.
With the remarkable complexity of AI computing systems and the need for highly reliable performance in the training and use of AI models, we expect growing demand for additional test steps. The addition of system-level test insertions for AI compute both in the cloud and at the edge creates an additional growth vector for Teradyne.
This was a primary consideration in our decision to align the integrated system test unit within semi test. Going forward, we believe that AI will have an outsized impact on the longer-term growth of edge devices, specifically, in mobile and automotive applications. Also, the trends towards electrification, whether pure EV or hybrid, provide considerable growth potential with increasing silicon content per vehicle.
Our investments in this space, including our strategic partnership with Infineon, will help us drive share gains in this highly complex, test-intensive segment of the market. Based on these long-term trends, we expect to see healthy TAM growth in the automotive and mobile segments of the market over the midterm. Our strong market position in these segments will help fuel our revenue growth.
These positive trends underpin our 2028 earnings model. At the midpoint of our model, we expect to grow from $2.8 billion of revenue in 2024 to $5 billion in 2028. We expect EPS to grow from $3.22 per share to $8.25 per share over the same period, implying a 12% to 18% revenue CAGR and a 21% to 31% EPS CAGR over that period, demonstrating considerable operating leverage in our business model.
To sum up, 2024 was a very good year. We have repositioned the company and are seeing the success from our investments in AI, in compute, and in memory. We expect that 2025 will be another good year. We are setting our robotics business up on a sustainable path for long-term growth. And our test business will grow, driven by continued strength in share gains in VIPs, tightening capacity utilization, and the return of higher demand in mobile, industrial, and automotive.
With that, I'll turn the call over to Sanjay. Sanjay?

Sanjay Mehta

Thank you, Greg. Good morning, everyone. Today, I'll cover our Q4 and full year 2024 financial summary. We will provide our Q1 outlook, some planning guidance for the full year 2025 and discuss our updated earnings model and capital allocation plan. Now to Q4.
Fourth quarter sales were $753 million, with non-GAAP EPS of $0.95, both at the high end of our guidance range. semi test revenue, which now includes our Integrated System Test business, or IST, comprised of product lines for system-level test and HDD test was $561 million.
Within semi test, SOC revenue was $429 million with memory shipments of $112 million and IST shipments, $19 million. The other product test businesses comprised of defense and aerospace, production board test and wireless task contributed $94 million. Robotics revenue was $98 million, was up 11% sequentially with muted seasonality due to ongoing weak industrial spending.
As Greg noted, we had softer-than-expected performance in the Robotics business tied to typical turns business that did not materialize. Non-GAAP gross margin was 59.4%, just below our guidance range due to robotics. Non-GAAP operating expenses were $284 million in Q4, higher than our guide. A majority of the increase was tied to accelerated engineering spend in semi test. Non-GAAP operating profit rate was 22%.
Some other financial facts. The tax rate, excluding discrete items for the quarter was 7.6% on a non-GAAP basis and lower than planned because of product mix shift to semi test. [GAAP] tax rate was 8.7% in Q4, excluding discrete items. We repurchased $144 million of shares in the quarter as we opportunistically accelerated our share buybacks. Dividends were $19 million, and we had one 10% customer in the quarter.
Turning to the full-year results. Our revenue was [$2.8 billion]. Samsung was the only customer greater than 10% of our revenue for the year. Gross margin for the year was 58.6%. OpEx was $1.08 billion and operating profit was 20.4%.
Non-GAAP EPS was $3.22. We generated $474 million in free cash flow in 2024. We returned $275 million or 58% of free cash flow to our shareholders through share repurchases and dividends. We ended the year with $724 million of cash and marketable securities. Our tax rate for the full year, excluding discrete items, was 12.6% on a non-GAAP basis and 12.5% on a GAAP basis.
Business unit revenues for 2024 were as follows: semi test revenue for the year, including IST, was $2.124 billion with SOC revenue contributing $1.537 billion, memory, $502 million and IST, $85 million. Excluding the impact of our DIS divestiture, our SOC and memory revenue was 17% year over year. SOC growth in the year was driven by AI compute, specifically custom ASICs for VIPs and networking. Our memory sales were up 30% year over year, driven primarily by AI compute demand for HBM DRAM. IT revenue declined 39% year over year, primarily due to underutilized test capacity in HDD.
Turning to our other product test businesses. the System Test group, which is combined defense and aerospace and Production Board Test had revenue of $201 million in 2024, flattish in 2023. Wireless Test revenue was $130 million, down from 2023 due to slower ramp of Wi-Fi 7. The combined revenue of the two operating segments in 2024 was $331 million, down 4% year over year. Now to robotics.
Robotics revenue in 2024 was $365 million, with UR contributing $293 million near $72 million. Considerably lower than expected volumes in the fourth quarter drove profitability well below our expectations, the group had 13% non-GAAP operating loss in both Q4 and the full year.
As Greg mentioned, we are restructuring the robotics business to create a single point of contact for customers and partners across UR and MiR sales, marketing and service organizations to improve customer experience. The result of these actions will help drive top line growth in 2025 and improve our efficiency. These actions will enable our robotics business to continue to outperform industrial automation market.
Now to our outlook for Q1. Since our October call, our semi test outlook has remained strong. However, robotics forecast remains seasonably soft. Q1 sales are expected to be between $660 and $700 million with non-GAAP EPS in the range of $0.58 to $0.68 on 163 million diluted shares. The first quarter guidance excludes the amortization of acquired intangibles and restructuring charges.
First quarter gross margins are expected to be in the range of 58.5% to 59.5%. OpEx is expected to be roughly flat with Q4 and run at approximately 41.5% to 42.5% of first quarter sales. The non-GAAP operating profit rate at the midpoint of our first quarter guidance is 17%. As Greg noted, we believe the semiconductor SOC test TAM will see healthy growth in 2025. And driven by a second half broad-based recovery.
We expect the SOC TAM to be between $4.7 billion and $5.1 billion or $4.9 billion at the midpoint. For a more detailed view of our end market expectations for SOC, please refer to the table in our earnings deck. We are forecasting the memory TAM to be between $1.3 billion and $1.5 billion. Recall that within this, HBM has grown from around $100 million in 2023 to over $500 million in 2024. Our memory TAM forecast for 2025 is roughly flat year over year with the HBM tester market going through a period of digestion.
In both the SOC and memory semi test markets, we expect to gain low-digit share in 2025. In robotics, we are currently operating in a difficult low visibility industrial spending environment. The business is still driven by terms. When we look at our plans for 2025 we see SAM expansion and channel growth initiatives expected to yield approximately 10% revenue growth in current market conditions. Of course, there is a wide range around this growth expectation.
Few points to assist you in the modeling 2025 for the enterprise. In Q2, we expect 5% to 10% sequential growth from Q1's midpoint. We expect first half revenue to be approximately 43% to 44% of full year revenue. Now to gross margins. We expect full year gross margins to be 59% to 60%.
We expect second half gross margins to slightly improve from current levels tied to higher revenue expected in the second half of the year. Regarding OpEx for the full year. We expect full year 2025 OpEx to increase 8% to 10% year over year, which is a reduction from our low teens view in October. The key changes were restructuring to capture synergies between UR and MiR and robotics and the acceleration of semi test projects in Q4. Interest and other line is forecasted at $1 million of income per quarter while we have cash driving the yield.
We also have items like FX gains and losses included in this line in our P&L. Our GAAP tax rate is forecasted to be 15.25% and 15% non-GAAP in 2025, excluding discrete items. Turning to capital allocation. Our strategy remains consistent as we take a balanced approach to maintain cash reserves that enable us to run the business and have dry powder for M&A. For reference, from 2015 to 2024, we've returned over $4.6 billion to shareholders through share repurchases and dividends.
which is 93% of free cash flow. 2025, we plan on executing up to $400 million of share buybacks along with our current level of dividends. Moving to our midterm earnings model. As we do each January, we've updated our model. We share this model with investors to provide insight into how we look at the markets we serve, our competitive positioning and ultimately, the growth and earnings power of the company.
A few points for context. We're rolling forward our midterm model to 2028, which replaces our prior 2026 midterm model. That said, we believe we are tracking with our prior 2026 model in terms of ranges of revenue and earnings.
Over the midterm, we expect test revenue to grow at a 12% to 17% CAGR off of our 2024 results driven by continued strength in AI compute related demand and recovery with long-term growth in broader end markets, including auto, industrial and mobile. Our mobile assumption is for recovery, but we're not assuming a return to the prior peak 2021.
In robotics, we're expecting the industrial markets to begin to recover with AI expanding the SAM and persistent labor shortages. We expect these dynamics to drive a top line of 18% to 24% CAGR off of 2024 with modest growth in 2025, which we expect to accelerate over the midterm. Going forward, the robotics operating model will deliver increasing operating leverage through the midterm, ending towards the high end of our target 5% to 15% operating profit range for this business.
Our updated midterm model is expected to drive 2028 revenue to $4.5 billion to $5.5 billion and non-GAAP EPS between $7 and $9.50. As Greg mentioned, this implies a 15% CAGR from 2024 to 2028, and a 27% EPS CAGR at the midpoint, demonstrating the operating leverage of our test and robotics businesses.
Gross margin is expected to be between 59% and 60%. OpEx as a percentage of sales between 28% to 31%, yielding a non-GAAP operating margin of 28% to 32%, coming up 2024 was a good year overall, driven by strength in semi test. Excluding the DIS divestiture, our overall company revenues grew 8% year over year, and our SOC and memory combined grew 17% year over year, helping to achieve a 10% increase in our EPS to $3.22.
We are making strategic investments to drive competitive advantage in the semi test business, and we are leveraging logical synergies between UR and MiR, to drive long-term sustainable growth in robotics. We enter 2025 feeling good about the year and our line of sight to our midterm model.
With that, I'll turn the call back to the operator for questions. Operator?

Question and Answer Session

Operator

(Operator Instructions) CJ Muse, Cantor Fitzgerald.

CJ Muse

I guess first question, was hoping you could spend a little bit of time expanding on your outlook for low single-digit share growth in semi test. Can you kind of walk through the moving parts in terms of what's driving that?
And then considering that you're expecting a bit of a recovery in your noncompute businesses in the second half. Can you speak to what we could see potentially in terms of upside to that low single-digit number if that were to come in incrementally better?

Gregory Smith

CJ, this is Greg. Yes, so there's a number of things that are driving our outlook for low single-digit growth in share. In the SoC space, we think that we're going to continue to make progress with the compute VIPs, maintain that 50% share that we got this year as that TAM continues to grow. And we're expecting to see some incremental improvement in the mobile space, driven primarily by -- there's been sort of steady increases in complexity that has been filling up the big pool of underutilized testers.
And so we would expect that, that sort of continuing march of complexity will actually turn into incremental growth in the TAM in that space in this year. And then when it comes to auto and industrial, right now, we think that there's a short-term slowdown in that space. But the increase of automotive content, the increase of semiconductor content in automotive is kind of proceeding with each model year.
So even at similar or depressed end unit sales that we expect that to be modest growth in the automotive space. The other thing that I'll say is that the electronics content in hybrids is nearly as high as the electronic content in pure EVs and the trend in the market towards more hybrid vehicles is something that is certainly impacting the plans for pure EVs, but it's impacting the auto semiconductors a little bit less.
Now the last is in industrial. In industrial, one of the things that is actually helping to drive the market there is an indirect pass-through from the AI compute space. The AI compute is incredibly power hungry. And it needs really, really good power for all of these GPU fueled servers. And that has been a driver for some of our power and electronics customers to -- with things like point-of-load converters, that's actually helping to sustain that market even when the traditional industrial market is not quite as strong.
Now if you move over to memory, the gradual recovery in mobile is going to help in parts of the market. So incrementally, year-on-year, that part of the market is going to be a little bit stronger, the HBM part of the market is going to be a little bit weaker because of that digestion factor. But the appetite for DRAM in AI servers is incredible. And that's driving both LPDDR and DDR memories. So we're expecting the non-HBM part of the DRAM market to sustain a little bit better than the HBM part.
So if you look at that whole thing and you look at the share gains or the new test insertions we've won in HBM in 2024, plus our strong position in final test for DRAM, especially LPDDR DRAM and flash, we're expecting that to accrue to low single digits in memory as well.

CJ Muse

Very helpful. And then, I guess, as a follow-up question, on the robotics side, you're restructuring that business once again. Obviously, I think you had multiple kind of paths that you could pursue on that front? And I guess, what gives you the confidence that this is the right path? And what kind of timeframe are you giving yourself for points of success given the struggles that Teradyne has had within this business for many years?

Gregory Smith

Yes. So I think one of the things that we're trying to do is to tease apart what part of this is Teradyne struggles and what parts of this are market struggles. So if you look at the performance of our Robotics unit, it certainly is coming in under what our expectations were, but it is outperforming the companies that we compare ourselves in the industrial automation space.
So we actually think we've made some significant progress in a very, very weak market. Having said that, there's definitely areas that we thought that we could make changes that would help improve One of the big things that was a positive in 2024 was the pace of new product introductions.
And we think that the R&D groups associated with UR and MiR, are really clicking. And the releases that we have that are adding AI content to our product or enabling our partners to do AI-based robotics are very important positive. The positive that we've seen in terms of building an OEM channel for UR is something that we think is a positive, and we need to preserve.
And we've also made significant progress in terms of building large account competency in the MiR organization. So if you look at the restructuring that we're doing, we're really -- we are not -- we're making very modest changes in the product creation part of the business.
We had already made the changes that we wanted to make in the operations part of our business. And so now we're focusing on the commercial side of the business. And there, we're trying to really do think two things. The first is we want to help our partners be more successful. And the way we can help our partners be more successful is by giving them a broader product range that they can sell.
And that, we think, is going to really help out. It also drives natural efficiencies of our largest partners between UR and MiR, actually sell both UR and MiR already. And so we're going to be able to get a natural improvement in efficiency by covering those partners in a single way.
Also, as we approach this large account success that we're seeing with MiR, we definitely see that service is a very important component of success. And so by having a single service organization, we believe that's going to allow us to expand the success that we've had with large accounts and MiR to the UR product line as well.
So the main thing we're trying to do with this restructuring is to sell more robots. The next most important thing that we're trying to do is to set up our breakeven so that we have the ability to overperform, that we have the ability to deliver better operating margin against what is a growth rate that is always going to be impacted by the end market conditions.

Operator

Mehdi Hosseini, SIG.

Mehdi Hosseini

From my end is focusing on semi. I wanted to understand this partnership with Infineon. It's interesting that you're doing this at the bottom of the power semi. I'm not sure if silicon carbide could get any worse. But what's in it for Infineon? And how should I think about partnership benefits with Teradyne and benefits to Infineon I do have another follow-up.

Gregory Smith

Sure. So yes, it is definitely an adverse times in the discrete semiconductor market, especially in like Wi-Fi bandgap, silicon carbide, gallium nitride. The way we look at that is this is sort of a classic hype cycle situation, where people saw the promise of these technologies in terms of delivering higher efficiency power conversion for both cars and for alternative energy applications.
And there was -- I think people got ahead of themselves in terms of how quickly the crossover would happen between in internal combustion and EVs. So definitely, there was a lot of enthusiasm over on this market, and we think that, that enthusiasm has waned.
So actually, we kind of think this is a good time for us to make this kind of a move. The advantage for Teradyne here is as we look to this long-term path from here to 2028, we think that the discrete semiconductor market is going to -- the test market is going to be increasing in a pretty healthy way. We also believe that all of the other semiconductor content associated with that battery management system whose other conversion products, isolation products is also going to increase.
So the advantage for Teradyne by making this partnership with Infineon is that we're going to be able to accelerate our road map to be able to cover these new types of devices and the higher powers that they're going to be trying to address. The advantage to Infineon is inside of this group, they had a number of technologies and a number of specialty test equipment that they were using to support Infineon's market-leading business.
By moving those capabilities into a commercial ATE company like Teradyne, now those products are addressing the entire market rather than just the Infineon market. So that group has a broader horizon than before. Paradigm has an opportunity to accelerate our road map and Infineon gets to focus on designing and building chips, which is kind of the point of their company.

Mehdi Hosseini

We should assume there is a payment here? Are you actually acquiring this asset for a purchase price?

Gregory Smith

Yes.

Mehdi Hosseini

Have you disclosed how much it is?

Gregory Smith

No, we're not disclosing the purchase price. It is not at a material level for disclosure.

Mehdi Hosseini

Got it. And very quickly switching to industrial automation. If I just look at what you have guided for Q2 and extrapolate the year-end and look at the margin and OpEx, you're better off with less contribution from industrial automation and the numbers just speak for themselves. So my question to you is, what's the plan B? You have spent several years reorganizing, enhancing the channel partners.
But to me, it seems like you may have to end up partnering with the system integrator or there needs to be more. And again, the question to the management team-- is there a plan B? How much more investment would you be willing to make before saying, okay, maybe you have to part way with industrial automation.

Gregory Smith

Yes. So we're nowhere near the point where we would be making a part ways kind of a decision. But I want to address sort of the first point in your question that we are definitely not better off with less industrial automation. We are better off with more industrial automation, and that has a lot to do with the leverage that's built into the business model for industrial automation. So we've reset our breakeven to this $365 million level in 2024.
But that -- we are -- given current market conditions, we believe that we'll be delivering 10% growth above that level, and we would be delivering positive operating margin from that group. For certain, it's dilutive to the percentage earnings across the whole company, but it is accretive to earnings.
And we believe that this is something that's going to be an important growth vector for us over the long term. So we definitely saw an opportunity to improve customer experience and save money and set us up to be able to put our investments into the things that are driving the highest growth. But we're not really at a plan B stage yet.

Operator

Steve Barger, KeyBanc Capital Markets.

Jacob Moore

This is [Jacob Moore] on for Steve. First one for me is on 2-nanometer data all around. What is your outlook for the timing of that ramp look like over the year? Has that changed at all? And are there any notable differences you'd call out in testers for the new transistor architecture versus current generations?

Gregory Smith

This is Greg. I'll take that one. So right now, we are looking at the leading edge of devices that utilize 2-nanometer occurring in the -- like very, very late in 2025. Just -- and early production in late 2025. The bulk of that is going to be in 2026.
And at least in terms of our line of sight, we see both compute and mobile devices that are going to be taking advantage of that note. In terms of the differences for that, the primary difference is that these devices have significantly more complexity. And with that higher complexity, it is actually -- these devices are higher -- requiring higher peak power from the tester and they're requiring much, much more tester memory. And so the impact of that is there's some -- there's a degree of technical obsolescence that is going to be creating opportunities in the ATE space and also in the system level test space for us. So we're -- we think that this discontinuity in device power is going to be helpful.
It's not going to be like cataclysmic increase in demand because of the transition to two nanometer, but we do think that it's a (inaudible).

Jacob Moore

Got it. That's helpful. Second one for me is on end markets. I think AI and compute are clearly strong drivers, but PC notebook and mobile are still big volume based. And I know you -- I think you said a gradual improvement over the year, but can you help us frame up how you're thinking about total growth in those segments for the year?
And has there been a more meaningful shift in tester conversion to HPC from mobile that could affect that dynamic?

Gregory Smith

Okay. So first off, we think that the second half recovery is going to be pretty balanced. And in the compute space, I would expect that the primary driver is going to continue to be AI cloud related demand. And that accelerators, CPUs and networking for that part of the market. There's probably going to be a nudge upward in terms of client compute.
I would expect that our primary benefit from that increase would actually be in our wireless business unit in the LifePoint Group. When you think about mobile, the primary thing is that there are -- it depends on which analysts you're looking at, but generally, people are a little bit optimistic in terms of units, and there's definitely a march towards greater complexity in the phones.
And then I already spoke about industrial and automotive, where it's more increasing semiconductor content by model year. And we think that there's an inventory digestion and a weak spot that we're in now, but we'll be returning to sort of more normal dynamics by the end of the year.

Operator

Timothy Arcuri, UBS.

Timothy Arcuri

I had 2. So Greg, I wanted to ask about your VIP test TAM forecast. I think you said 626 and going to maybe 828. So my question on that is, if I listen to what Marvell and Broadcom are basically saying about the custom ASIC TAM. One is saying it's going to be $40 billion in 2027.
The other is saying it's going to be $140 billion in 2027, 2028. So if I average those 2, I can pretty easily get to $80 billion plus in that timeframe. So these numbers would imply that the test intensity is like less than 1% for that stuff. And this stuff all has tons of transistors, and it seems like that's a pretty conservative forecast. So I'm wondering what your custom ASIC end market revenue numbers that would underpin this because it seems like your test intensity number is very low based upon what the end customers are talking about relative to the size of the market.

Gregory Smith

Yes. Our model is built more bottoms-up based on like unit quantities device complexity. The one thing that is definitely not factored into our model is any variability in the margin that a Marvell or a Broadcom may be getting on the devices that they're producing. And so that's one area where be a disconnect in our forecast versus theirs. But to sort of answer your question directly, we would definitely be more aligned with the low end of that sort of (inaudible) to $140 billion outlook for 2027 than the high end.
And that's mainly because we're working off of sort of line of sight of what we can -- what we've discerned from these more the actual hyperscalers versus the -- what we are hearing from the chip folks in the mill.

Timothy Arcuri

Okay. Yes, seems low. But -- okay. So then on IA. So PBT has been negative for the past six years.
Cumulatively, it's -- you've lost $150 million. And usually, when companies lose money like this, the markets are growing very fast, and we're chasing growth. So can you just talk like is the plan here to cut costs and try to maximize profitability? Or do you really see something bigger coming that you still want to do this? And I kind of asked this in like the context of do you need outside investments to really scale this business?
Or is it not a dollar investment thing. It's more just that the market hasn't grown which you thought it would? And I guess the question there is like what's going to change to actually make this market grow?

Gregory Smith

Yes. So I think the x factor here is an extremely long down cycle in industrial automation investment, really from starting in 2022, all the way to now, it's been -- PMIs have been consistently weak. And we've been caught in that. While we've outperformed -- while we've outperformed our peers, it hasn't delivered the growth that it's supposed to.
So one of the things that we -- like our view that may be a little bit idiosyncratic from the other advanced robotics players is that we do believe that growth in this market is going to be a marathon, not a sprint that we are hoping to align our OpEx so that we're able to generate positive operating margin even under weak business conditions, and then incrementally increase that operating margin by making sure that we have leverage between any top line growth and limit the OpEx increase and improve our gross margin as we go along.
So we think that there is significant growth potential in this space that is being masked by the end market macro headwind. We don't have visibility into when those headwinds are going to come to an end. So we definitely were focused on trying to maximize our efficiency during this period to make sure that we are setting ourselves up so that there isn't incremental negative impact on earnings from our Robotics business even under weak business conditions. So that's the primary thing. I think I'd like to pass it over to Sanjay for a couple of comments on this as well.

Sanjay Mehta

Yes. Just really quick, Tim. Just for a point of reference, since 2019, 2019 to 2022 at an operating profit level, we were profitable. I'm not sure where you're kind of $150 million. In the last two years, it's true.
We have lost money. But cumulatively, since 2019 to 2024, we'll have lost an operating profit level of about $23 million. Just for context.

Operator

Krish Sankar, Cowen.

Krish Sankar

I told them first one, Sanjay or Greg, if I look at your color on calendar '25, it looks like your revenues in '25 are going to grow about 15% from 2024 levels. A, is that right? And along the same path, a couple of months ago, you seem a little more confident on maybe high teens growth for the year versus mid-teens right now. So what kind of change on Teradyne's outlook for 2025? And then I have a follow-up.

Sanjay Mehta

Sure. It's Sanjay here. Yes, at the midpoint of our guide, it's 15%, obviously, or the midpoint of our estimation. And it could be plus or minus there. And what's changed over the last 90 days, I'd say, is that robotics, our view of the market, obviously, from Q4 and our forecast or our thinking of our plan for robotics has come down to a growth rate of about 10% year over year.
I will add that what hasn't changed is that our view on semi test actually over the past 90 days has actually strengthened. We saw acceleration into more business into Q4 of '24, but we held our plans roughly the same over a 5-quarter period. It's actually been enhanced, which gives us confidence in the year for semi test. And I'd say the other product test businesses are roughly in line of what we thought about.

Krish Sankar

Got it. Got it. And then as a follow-up, thanks for the color, you kind of spoke about how your SoC compute revenue has been growing from 11% to 34% last year. I'm curious where you think it could end up being this year in the context of your $4.9 billion SOC TAM. If you can just give some color on how much of that is compute, how much is mobile and what percentage SoC revenue you could get from both compute and mobile this year.

Sanjay Mehta

Yes. So I think compute will be one of our largest growers and should increase as a percentage of the SOC revenue that we have. We also see, obviously the TAM growing in compute. So we're tracking to it. We see continued strength in VIP.
As Greg noted, roughly a TAM in 2024 for VIPs of 300 -- of compute VIPs of 300. And in 26, we see that as roughly centered around 600. So as you'd expect, we see growth in 2025, and we're tracking to that growth in compute.
I think from a mobile perspective, we do see it continuing just likely growing along the 2025 level -- or consistent with 2024 similar to 2025. And then when we take a look at auto and industrial, we do see that TAM growing along with our business growing.
So I'd say a little bit more strength in compute and TAM as well as from our shipment perspective, and then auto and industrial growing with mobility flattish to slightly up.

Operator

Toshiya Hari, Goldman Sachs.

Toshiya Hari

My first one is on the VIP business. When you think about growth outlook in '25 and '26. I was hoping you could sort of touch on the breadth of your customer base in that business. Is the growth expected to mostly come from existing customers? Or are you looking to -- or do you expect for a new product or new customers to ramp in '25 and '26 as well within the context of the VIP business and AI compute more specifically?

Gregory Smith

Yes. So it's Greg. The VIP business is pretty concentrated. I mean, there are very few companies that have the scale necessary to pay for a 2-nanometer ASIC accelerator. So this is definitely going to be lumpy.
If you look at our 2025 business we are expecting a robust ramp of the existing customers that we have. And we're expecting between '25 and into '26, that would be adding one or two logos to the customers that we have. But like if you're trying to get an understanding for this is always going to be a situation where it's like very big single sockets that drive a lot of capacity adds, not a broad-based.

Toshiya Hari

Yes. That makes total sense. And then as my follow-up on HBM Test, you talked about digestion in '25 and then the potential recovery in '26, I think. Is that view primarily based on a bottoms-up based on customer forecast? Are you guys making internal assumptions around HBM bit growth, technology evolution?
Is it a combination? I'm just curious how you formed your view on '25 and '26.

Gregory Smith

So we have very good relationships with most of the major memory producers. And so a lot of this is based on our conversations with them and their advanced capital planning. The -- and having said that, there is a fair amount of capacity in place that needs to be filled out that is capable of doing HBM3E.
The thing that is an x factor in our model is the timing of a transition to HBM4. If that transition comes in, in time, then there's upside to the TAM, if that transition delays in time, that would mean that, that TAM would be further weakened.
But right now, we've kind of going with our best view from what we're hearing from our customers in terms of when that transition is going to happen.

Toshiya Hari

And when you say in time, do you mean sort of first half '26? Or -- I just wanted to clarify.

Gregory Smith

The real question is, yes, whether like people are certain that HBM4 is going to be in volume production in early 2026. There is there's certainly a customer desire to try to have more significant volumes of HBM4 in the latter half of 2025. And if the schedules support that, that would pull in the capacity needs for that into '25 from '26.

Operator

Samik Chatterjee, JPMorgan.

Samik Chatterjee

I guess for the first one, the seasonality between the first half and the second half that you're indicating this year is definitely more second half skewed than sort of the last couple of years and that brings into sort of the question in terms of you're talking about second half recovery in markets like mobile and then auto industrial.
How are you sort of derisking that second half when you think about upside, downside risk, particularly given your visibility, how should we think about sort of that first half versus second half improvement and sort of upside, downside risk there already? And I have a follow-up.

Sanjay Mehta

Yes. It's Sanjay. So that's right. As we have the visibility into Q1 or the first half, second half, I'd say we're in an environment where we see the upgrades that have occurred that we've talked about to underutilized capacity, we see that as coming to and then some time of these upgrades, really getting to really high levels of utilization. First half-ish, and while there'll still be probably some minor upgrades, we see capacity tightening.
You can see that with the VLSI reports as well as how when we go and count the testers and utilization and our methodology of getting there. So we see a tailwind as utilization tightens -- the second thing is that we do have a pipeline as we think about the customers we're engaging that gives us confidence in the second half.
And in the second half, it's broad-based, I'd say, more focused around compute as well as auto and industrial from a semi test perspective. And then also tied to the end market recovery. And as Greg noted, end market recovering in industrial automation as well.

Samik Chatterjee

Got it. And maybe going back to the [VIB], ASIC TAM and you sort of highlighted, we made some comments on the last question as well in terms of logos that you're expecting. But specifically, I think there is a broader investor concern about your ability to maintain that 50% share that you're talking about.
And when we sort of look into 2025, 2026, is that visibility in terms of maintaining the 50% share more in terms of continuing to work with your existing customers? Or how does the pipeline look in terms of sort of new wins, either in terms of sockets with existing customers or new customers that's driving that confidence to keep that 50% share.

Gregory Smith

Yes, this is Greg. I'll take that one. So we believe that we have the best product for these leading-edge devices. But one of the critical factors in ATE selection is the installed base of tools and training and everything else that you have around the platform that you're using. So the Teradyne's traditional lose share in the compute space has a lot to do with the way the market was shaped for the last 10 years.
The success that we're having in VIPs is that with these new devices and the use of these design service providers like a Broadcom or a Marvell or a Samsung (inaudible) UC that we have excellent relationships with these customers and we're able to demonstrate the differentiation of our product. And we think that people are getting a good result by using our product, and they're choosing our product for that reason.
So from a product differentiation perspective, we think we have the right recipe in order to maintain that share. We think that we have the right customer relationships to be able to do that. And we believe that we have the right relationships with the hyperscalers themselves to try and add the logos that we need. So we're pretty confident that we're going to be able to keep going at this 50% level.

Operator

Vivek Arya, Bank of America.

Vivek Arya

So first one, compute TAM grew 57% last year, but you're expecting it to grow only 5% in '25. I'm curious what is causing that a slowdown? And then versus the, I think, $300 million in VIP last year, what is the VIP TAM for '25?

Gregory Smith

So in terms of compute growth, you can -- I think you can look at this in a little bit of a similar way to the situation in HBM. The end market for this stuff is going to stay very, very hot. The -- but -- the situation in terms of capital equipment or test equipment to support the volumes that are being produced doesn't necessarily follow that in lockstep.
Right now, there are a lot of devices in the space that are still coming up an experience curve and have very, very long test times that could be reduced as parts become more mature. And we believe that, that's going to have -- that's going to limit the sequential growth of the compute space.
So we think the primary opportunity in the compute space is in these new sockets from the VIPs, not in the traditional space where it's sort of a year after introduction and experienced situation. So we think that there's definitely there are balloons and anchors that are going to be affecting this. The balloon is obviously this huge CapEx for data center that's going in the anchor is increasing efficiency of production for the devices.

Vivek Arya

Okay. But then how do we reconcile that with the strong I think almost 30-plus percent half-on-half growth that you are expecting this year driven by compute. If the TAM is not growing, then how can one depend on that to grow much more than what Teradyne's normal seasonality has been half-on-half in the second half.

Sanjay Mehta

Yes. So I think you should think about it as share growth, some of that.

Gregory Smith

Yes. It's share growth. And the other thing that I would say is we're kind of expecting compute to be delivering throughout the year. and we're expecting the automotive, industrial and mobile segments to be more back half loaded. So it's not that we're expecting a big back half for compute, I think.

Sanjay Mehta

Yes. I think compute, we are seeing -- it's a little bit more balanced, but back half is going to be stronger than the first.

Vivek Arya

Okay. But it's not 30% increase, is it?

Sanjay Mehta

It is. Okay. It's going to have a reasonable increase.

Gregory Smith

But I think that the logic behind that is related to this new part introduction that if you're introducing new parts, you need capacity to support that if you are increasing volume on parts that are already in production, then that's affected by the experience curve.

Sanjay Mehta

And then, Vivek, you had a question on the VIP TAM?

Vivek Arya

I think the $300 million, what is it this year in '25?

Sanjay Mehta

Yes. What we said in our prepared remarks that it's $300 million and then growing to but we expect to be in 2026, centered around 600. You should expect that that's going to grow on a trend line. It has a range. But if you think about $400 million to $500 million, maybe bias towards the higher end range for 2025 is how you should think about it.

Operator

Brian Chin, Stifel.

Brian Chin

A few questions here. Sorry if I missed this, but -- and I know there was sort of some movement around of certain revenue streams into different buckets. But what was the size of the SOC test market in 2024? And if no one asked us yet, can you decompose sort of how that $4.9 billion TAM for 2025 breaks down across compute, mobile, et cetera?

Sanjay Mehta

Sure. On '24, I believe we have a slide in the backup, but compute is $2.2 billion. Mobile is 0.8. -- auto and Industrial's 0.9, services, [0.7] and 4.6 SOC TAM in '24. And then if you go to '25, you've got $2.3 billion in compute mobile at $0.9 billion, auto and industrial at $1 billion and service at $0.7 billion to get us to the $4.9 million at the midpoint.

Brian Chin

Got it. That's helpful. And just -- and you sort of touched on this a little bit, but just to underpin that view for the modest test recovery in the second half. Can you give us maybe a better sense of where semiconductor test cell utilization rates are now versus a year ago to kind of how close we are to bridging that gap?

Gregory Smith

Yes. So I mean we've cautioned before in terms of the absolute accuracy of these utilization. So we tend to look at changes from quarter-to-quarter. I will say that utilization has drifted up through 2024, and it inflected more strongly in the end of the year than before. And that's really around the impact of the upgrades that we shipped through the year.
The upgrades that we shipped through the year that turned previously idle testers into testers that are being actively loaded. There was a delay of maybe one quarter from those being shipped to when they are actively loaded by our customers.
But it's probably about a 10% increase in utilization year-on-year like exiting '24 from coming into '24. And right now, we're at the point where we think our business has already shifted in mix more away from upgrades and more towards system sales. So we think that, that's -- we've kind of passed the tipping point there.

Brian Chin

Okay. Maybe if I could sneak in a box question. Would you mind unpacking a little bit what you mean by consolidating the go-to-market? And also in 2025, does your kind of modest growth outlook suggests that you think the benchmark of the broader robotics industry is down something like 10% this year?

Gregory Smith

Yes. So the details of the consolidation is that we've combined the marketing functions, the sales functions and the service functions of UR and MiR together so that we have one unified go-to-market organization. And that is the primary way that we are going to improve our interface to our partners and improve the interface to our largest customers.
There's a lot of overlap in terms of our largest partners and our largest customers where they are customers of both UR and MiR, so it was an obvious place where we could make things simpler for them and also drive some efficiency for us. There's really no change in terms of the primary go-to-market growth vectors that we see for these groups.
And that really is OEMs and for -- and large customers. And so we're continuing to lean into those programs. But now we're doing those across both product lines versus on an isolated basis. Now the -- could you remind me what your second question was in there?

Brian Chin

Yes. It was just that I think you previously said you kind of said growth targets against a benchmark outperforming benchmark for the industry, something I think like 10% to 15% better than the benchmark. And if you're guiding kind of for modest growth this year, does that mean that, that benchmark is down maybe 10% this year?

Gregory Smith

Yes. I think we're -- everyone is trying to figure out exactly what the shape of a recovery in industrial automation would be. I think you're right in thinking that we are -- our intent is to outgrow the traditional automation players by like 20% per year or maybe more like 15% per year.
So I think it's probably more like we're looking at 5% down with us delivering above that. So if a -- if the automation market were flat, I think we would be significantly higher, up around 20% growth. If it's down, around 15% growth. If it's down 5%, then we deliver like 10% growth.

Operator

Shane Brett, Morgan Stanley.

Shane Brett

So I understand the importance of giving investors long-term forecast of the 2028 model, but I just wanted to better understand, what are the kind of more finer details behind the revenue assumptions for that model? And sort of how far your visibility extends? The context of this question is that since 2022, it appears as though you are continuing to set up a little bit of a higher bar for your business.

Sanjay Mehta

Yes. It's Sanjay here. Yes. So as we've said in our prepared remarks, we see test revenues -- we see overall revenues growing 12% to 18%, and that's really broken out from a test perspective of 12% to 17%. And then from robotics, an 18% to 24% level.
I would say, starting with robotics, there are some key drivers that we have, AI-enabled SAM expansion really focused on channel growth through OEMs and large accounts and new product introduction.
And so we believe we are in a sub-5% penetrated market. The market has had some tremendous headwinds, as Greg noted over the past couple of years. However, we see that as we develop these incremental solutions, AI-enabled capabilities, in our products that, that will enable us to help address incremental segment verticals. And so it's with that, we have the confidence, we just need the market to return, and we are predicting that in our model. From a test perspective, I would think about it like this.
We analyze the TAM. We take a look at and maybe, Greg, do you want to add some comments when I'm done. But we analyze the TAM. We look at things in the different segments like compute, how we see the AI compute or high-performance compute go in networking as well as the forecast from external players, but also talking to customers. and what our view is of the custom ASICs.
So we look at it segment by segment, high-level drivers, Greg mentioned, silicon content in cars, growing as well as a fast-moving driver or an increased driver in hybrid as well as EVs. But it's both a top-down and a bottoms-up.

Gregory Smith

Yes. One bit of color in terms of the semi test growth. I think one way that you should think about it is that there was this incredible inflection in 2024 for the memory market. So the memory market, like really bumped up in 2024. So if you're looking at 2024 to 2028 growth rates, we think that the memory market has already sort of built in some of the growth of that longer period.
So we're expecting sort of a lower incremental TAM growth in the memory market going out to 2028 and a more linear model of growth rate for the SOC TAM through that period.

Sanjay Mehta

And maybe I'll add one more point on share. In the compute (inaudible), as we see the market shift to more custom ASICs, we also see share gains, we believe we're going to continue to win kind of one in two sockets. And then as the market recovers in mobile as well as auto and industrial markets that we've had traditionally higher share in as those markets become a larger portion of the TAM relative to 2024 and those markets recover, we should inherently gain share in those markets.
One other thing that I think you probably want to try to consider as you're trying to flesh out a view for 2028 is that we think system-level test is going to be a long-term growth factor. That adding those test insertions is going to be important to achieve the quality levels that our customers are demanding.
And so we believe that, that's going to be one of the things that helps to drive that -- you said it was 12% to 17% growth for the test reps, that there is a chunk of that growth that is coming from that system level test stuff.

Shane Brett

Got it. Thank you very much.

Operator

Thank you. There are no further questions at this time. I'd like to hand the floor back over to Traci Tsuchiguchi for any closing comments.

Samik Chatterjee

Thank you again for joining us this morning. We hope that many of you will be able to join us for our Analyst Day on March 11. Until then, we look forward to speaking with you soon. Bye.

Operator

This concludes today's conference. You may disconnect your lines at this time. Thank you for your participation.