Cohu, Inc. (COHU) Earnings Call Transcript
July 30, 2026
Earnings Call Speaker Segments
Good day, and thank you for standing by. Welcome to Cohu's Second Quarter 2026 Financial Results Conference Call. [Operator Instructions] Please be advised that today's conference is being recorded. I'd now like to hand the conference over to Matt Hutton, Vice President of Strategy and Investor Relations. Please go ahead.
Thank you, operator, and welcome to Cohu's Second. Quarter 2026 Earnings Call. Our agenda begins with Luis Mueller, Cohu's President and CEO, who will provide a business update, followed by a financial review and outlook from Jeff Jones, our Senior Vice President and Chief Financial Officer. Following our prepared remarks, we will open up the call for your questions. If you need a copy of our earnings release, it can be found on our website at cohu.com or by contacting Cohu Investor Relations. A slide presentation accompanying today's call is also available in the Investor Relations section of the website. Replays of this call will be accessible via the same page after the conclusion of the call. During this call, we will be making forward-looking statements that reflect management's current expectations concerning Cohu's future business. These statements are based on the information available to us at this time, but they are subject to rapid and sometimes abrupt changes. We encourage everyone to review the forward-looking statements section of our slide presentation and the earnings release as well as Cohu's filings with the SEC, including the most recently filed Form 10-K and Form 10-Q. Our comments are current as of today, July 30, 2026, and Cohu does not assume any obligation to update these statements for events occurring after the call. Additionally, we will discuss certain non-GAAP financial measures during this call. Please refer to our earnings release and slide presentation for reconciliation to the most comparable GAAP measures. Now I'd like to turn the call over to Luis Muller, Cohu's President and CEO. Luis?
Good afternoon, and thank you for joining Cohu's Second Quarter 2026 Earnings Call. We delivered a strong quarter with sales of $149 million, up 38% year-over-year and recurring revenue of approximately 53% of total. These results reflect solid execution across the company and continued customer adoption of our solutions. Our Q2 results show progress in areas where we have focused investments, advanced thermal test handlers for AI processors, HBM inspection, flexible ATE platforms for power and connectivity devices and software analytics. Cohu is benefiting from durable demand drivers in AI infrastructure and edge computing, where customers are investing to address increasing power levels, production yield and increased factory productivity. Estimated semiconductor test sterilization improved sequentially to 80% at the end of the second quarter, typically a turning point for test Capex by our core IDM customer base. The strongest segments were computing and industrial with test sterilization in the low 80s, followed by automotive and mobile in the high 70s. Bookings generally follow utilization trends. And in the second quarter, computing led with 46% of total system orders, representing an impressive 150% increase year-over-year, driven by eclipse growth in high-performance computing. With utilization above 80%, industrial was the next largest growth area with orders up 87% year-over-year. The balance included consumer up 29% year-over-year, mobile essentially flat year-over-year and automotive down 24% year-over-year in a segment that continues to struggle in this recovery cycle. Let me now review the quarter by product line, starting with the areas where we see the strongest customer traction. Starting with our test handlers. Building on the momentum from Q1, we continue to expand our position in high-performance computing through the adoption of our Eclipse handler, enabled by advanced active thermal control for extreme power and next-generation Jet MAX devices used in data centers. The Eclipse delivers a configurable thermal handler that can be used across multiple device generations, helping customers reduce capital risk, extend the value of their installed base and support faster production ramps. Our high-power thermal control technology remains a key differentiator, improving task quality and first pass yield. During Q2, high-performance computing customers also expanded adoption of Cohu's PACE prescriptive analytics software, reinforcing our strategy to improve equipment efficiency and customer value. Separately, we are increasing our presence in infrastructure at OSATs in Southeast Asia to support fabless and hyperscaler programs as they move from qualification to our production ramp. Overall, this momentum reinforces an expanding high-performance computing customer pipeline that we now estimate at approximately $850 million annually. This includes about $190 million in qualified annual opportunity across 4 customers, $250 million in active qualification across 5 customers and approximately $445 million in early-stage engagement across 10 additional customers. Based on this progress, we are raising our fiscal 2026 high-performance computing revenue estimate to between $100 million and $110 million. To support this demand, we are working with supply partners to increase capacity, and we're expanding our internal manufacturing in Malaysia. We expect this expansion to double output by year-end and support another step-up in capacity by mid-2027. Now turning to our inspection and metrology. During Q2, we shipped additional final inspection systems for HBM 3, HBM 4 and HBM 4E devices to a U.S.-based IDM with a strong forecast into the second half of 2026. HBM is the memory backbone of AI training and inference infrastructure, and we continue to invest to stay ahead of customer road maps into HBM 5 and beyond. We recently qualified Neon at a Taiwan-based OSAT, establishing a strategic foothold in a high-volume outsourced assembly environment where we can pursue additional advanced package, mobile and AI adjacent inspection opportunities. We also released a new vision inspection sensor with short-wave infrared capability that detects inner cracks in complex silicon devices. This gives customers a step change improvement in outgoing quality for advanced packages and further reinforces Neon as a reference platform for silicon inspection applications. Moving to semiconductor test. Demand is increasingly tied to 2 AI-enabling requirements. efficient power delivery and high-speed connectivity across edge devices, vehicles, industrial equipment and connected infrastructure. These systems depend on precise power management to achieve peak performance while controlling energy consumption, heat and operating costs. Customers are increasingly adopting gallium nitride power devices for their efficiency and power density advantages. At the same time, AI-enabled devices require seamless connectivity. The industry is investing in both ground-based and non terrestrial networks, including satellite constellations and high-altitude platforms that extend coverage globally. Together, GaN power and advanced connectivity represent an estimated $340 million annual addressable market opportunity over the midterm. During Q2, we continued to engage with leading power and RF customers on Diamondx configurations for GaN, mobile front-end and advanced connectivity applications. Switching to software analytics. We built on last quarter's momentum and moved from early production wins toward a repeatable pattern, land a first deployment, prove measurable value and expand within that account. This business delivered the first $1 million revenue quarter and orders increased an impressive 140% year-over-year. First, land and expand is working. Last quarter, we noted that a leading high-performance computing chip maker had committed to deploying our predictive maintenance technology across its test handler fleet. We're now progressing toward what is expected to become our largest software deployment to a single customer with expansion planned during the second half of the year. Second, we are deploying Agentic AI where the data lives. Semiconductor manufacturers operate under strict data sovereignty requirements and most cannot send process, yield or equipment data to a public cloud. In Q2, we advanced our on-site AI appliance, which runs modern AI models and autonomous agents entirely inside the customer's network with no data leaving the factory. These agents can conduct investigations across equipment, maintenance and past data more frequently and at a lower cost than manual analysis. These deployments are important because they convert Cohu's installed base into a recurring software revenue opportunity while helping customers improve uptime, yield learning and factory productivity. Moving to our interface solutions. This is a key element of our recurring revenue stream and about 19% of Cohu consolidated revenue in Q2. Our high-speed interface technologies continue to gain traction in silicon photonics test. We booked $500,000 in interface solutions used in optical engine test and are pursuing additional customer engagements tied to emerging requirements for co-packaged optical devices. In parallel, we remain focused on increasing share of our core semiconductor customer base, where new applications and replacement of incumbent technologies create opportunities for additional Cohu content. In summary, Q2 demonstrated progress across the strategic priorities we outlined earlier this year. scaling high-performance computing handler adoption, advancing inspection solutions, expanding Diamondx into power and connectivity, converting software pilots into production deployments and broadening interface solution adoption into optical and advanced semiconductor devices. I want to thank our customers for their partnership, our employees for their execution and our shareholders and supply chain partners for their continued support. With that, I'll turn the call over to Jeff to review our financial results and outlook in more detail. Jeff?
Thank you, Luis. Before reviewing the second quarter results and providing third quarter guidance, please note that my comments refer to non-GAAP figures. Details about non-GAAP financial measures, including GAAP to non-GAAP reconciliations and other disclosures are included in the earnings release and investor presentation on our website. For Q2 2026, revenue of $149 million exceeded the midpoint of guidance. Recurring revenue driven primarily by consumables represented 53% of total revenue. One industrial customer accounted for more than 10% of total sales during the quarter. Gross margin was 45.5%, above guidance, primarily reflecting a more favorable product mix. Operating expenses were in line with guidance at $52.7 million, reflecting our decision to scale resources to support the rapid increase in high-performance compute opportunities. Net interest income after interest expense and a $600,000 foreign currency loss was approximately $1.7 million. The Q2 tax provision was lower than guidance at $2.7 million due to improved profitability in the U.S. Non-GAAP EPS for the second quarter was $0.26 and adjusted EBITDA was 12%. Moving to the balance sheet. Cash and investments increased by approximately $9 million during Q2 to $498 million and cash from operations was $10 million. No stock repurchases were completed during the quarter. Total debt is $304 million and includes $288 million from the Q4 2025 convertible debt offering. Capital expenditures were approximately $2 million, mainly for manufacturing machinery and equipment, facility improvements and IT equipment. We're targeting total capital expenditures to be about 2% of revenue in 2026, including the capital expansion of our Malaysia test handler manufacturing facility mentioned by Luis. Looking ahead, we expect Q3 revenue to increase 14% sequentially and 35% year-over-year to approximately $170 million, plus or minus $7 million. The increase is driven by demand tied to the ramp in high-performance compute opportunities and continued recovery in our core business segments. We're increasing our full year 2026 revenue outlook for growth over last year to approximately 35%. Q3 gross margin is projected to be approximately 45%. And for full year 2026, we continue to expect gross margin in the mid-40% range. The rapid expansion of high-performance computing opportunities has increased demand across our supply chain and production base, resulting in longer lead times and higher input costs for certain semiconductors and specialty components. We're taking proactive steps when available to secure critical components to minimize impacts on our lead times, profitability and customer pricing. Operating expenses are expected to be about $54 million. We intend to continue investing in resources to capitalize on the growing list of HPC opportunities, and we expect quarterly operating expenses through the balance of the year to remain in the low $50 million range, consistent with our Q3 guidance. In light of expanded resources to support HPC-related growth, our operating model continues to demonstrate solid profitability leverage with approximately 40% of projected sequential revenue growth expected to convert to operating profit. Net interest income in Q3 after interest expense and foreign currency impacts is projected to be approximately $1.6 million at current interest rates. The Q3 tax provision is expected to be about $5.2 million, and diluted shares are projected to be approximately 55 million, including 5.8 million shares attributable to the convertible debt. And of that amount, 2.4 million shares will be fully offset by the capped call but are required for U.S. GAAP diluted EPS calculations. In summary, our 2026 priorities remain focused on supporting the R&D investments and production ramp required to secure multiple design wins in the compute market, including AI data center infrastructure, HBM memory and physical AI applications while progressively increasing EBITDA margin and free cash flow. That concludes our prepared remarks, and now we'll open the call to questions.
[Operator Instructions] Our first question comes from Krish Sankar with TD Cowen.
On the solid results and guidance. Luis, I had 2 questions. The first one, I just want to check, of your pipeline of $850 million, you said you have 4 customers qualified. Are those 3 HPC and 1 HBM customer? And the other 5 customers in qualification, are they all HPC for AI handlers? And when do you expect that to potentially convert into revenues?
Krish, yes, you're correct on all your statements here. We have 3 HPC, 1 HBM on the qualified, which we view about $190 million annual opportunity for revenue. And then we have close to $200 million in the near-term qualification here. So your question on the time line, it straddles over months, to be honest with you. We have one customer that is right on the edge of giving us the green light of being qualified. I think the data all supports it, but we don't have the official yet. We are already planning on shipping a production configuration for the actual production IERceP device, which is the next generation. We typically qualify on an existing generation device, so it can do correlation to what they have with other systems out there. But we're shipping the production configuration here at the end of August to get it, I don't know if the right word is certified, so we can go on with the actual intercept. So we should get a qualification pretty soon within a month, I would say. And then I think the fifth one on the list, we're looking at early next year, systems that we're shipping late August and I think accounting for about a 6-month qualification process, I think would put us sort of mid-Q1 for the last one on this bucket of in qualification.
Got it. Very helpful, Luis. And then as a quick follow-up, you mentioned about you're getting traction in silicon photonics for the optical engine. Can you quantify how much that opportunity to be either this year, next year and the next few years? And is this mainly an insertion 1 or which insertion are you targeting?
Yes. Today, we are shipping interface solutions, not handlers, not full handlers yet. We're shipping interface solutions for insertion 3, which is the optical engine test. And that's what we commented here in prepared remarks that we booked in the second quarter. I think it was a $0.5 million order for interface. And this is a continuation of a business we started in Q1. We are shipping or planning to ship a qualification unit for insertion 3 by the end of the year with a handler. We're also demonstrating an insertion 4 configuration with a handler to certain customers that I don't really have a specific time line for shipment yet. I don't know if it's going to be Q4 or Q1 next year. A little tricky now getting systems to go on qualification, given the production orders that we're satisfying. I haven't quantified -- we haven't quantified yet the total CPO revenue in '27, '28 because we view it as part of the evolution of HPC. So it's embedded in the $850 million pipeline at the moment.
Our next question comes from Brian Chin with Stifel.
Nice results and outlook. Maybe the first question, looking at this -- the multistage pipeline graphic you have in the slides, I think last cut, it was aggregating to $750 million. Now it's $850 million. Can you maybe break down what that $100 million increase is? And I think part of this maybe is that recurring portion. Can you also maybe explain and break down what you mean by recurring?
Okay. Brian. Yes, two separate things. On the totality of the pipeline increase to $100 million -- by $100 million, it's really getting better visibility in the forecast. And I think we added a couple of customers on the engagement phase as well. But we certainly have better visibility now on the customers that are in the unqualified or qualified, I should say, the qualified portion of the pipeline. And the numbers are bigger, what they're giving us for next year than what we had originally estimated. To your question on recurring, there are really 3 components to recurring. One of them is device application kits. You -- these device life cycles are typically 18 months in production, something else launches and you got to do a new device kit for that handler. The other component is thermal heads. You don't necessarily change the entire thermo head. But as devices grow in size, so does the required thermo head coverage over the die. It could be multi-dies. It could be actually even multi-skyline heights on dies. So that thermal head touch down on the die has to evolve with the product evolution. So that's an upgrade element of the system. It could also include a thermo head itself if the upgrade includes higher following. And the third element is basically the maintenance of the equipment. There are spares and consumables in the equipment. Part of that today, I guess, a fourth element that is novelty here is the software sale. As I mentioned last quarter, we sold -- I think it was about $330,000 a year subscription software into an HPC customer in conjunction with system orders. And that has a lifetime through the product life cycle, I think we estimated a few million dollars life cycle value -- lifetime value of that software subscription. So that will be sort of the fourth component that I forgot to mention, but it's part of recurring.
Great. That's super helpful. Also, in terms of the full year revenue guide increase going from 25% to 35% growth, I think it's something like $45 million on the math there. You only increase your HPC forecast by maybe, call it, $10 million to $15 million. And so I guess the majority residual there is all the kind of like the core business, historical core business for Cohu. Can you maybe expand upon kind of what you're seeing there in terms of improvement? Industrial obviously is taking up...
Yes, that's right, Brian. Seeing industrial pick up, we're seeing the utilization rate overall pick up and part of the growth there in the second half or actually full year is the growth in recurring revenue. So that's grown nicely. I think it's been a CAGR of 5% here over the last 6 quarters or so. So it's really a nice increase. But you're right, it's the core business that's coming back led by industrial.
Maybe if I could sneak one last thing. Just from a supply standpoint, Jeff and Luis, I know that the in-qualification bucket is not banded within 2027 or a year interval even. But what are you sort of targeting to be able to get capacity to in 12 months' time or whatever horizon in terms of the Malaysia expansion and kind of what that -- given what the business looks like in terms of the following in?
Yes, Brian, frankly, that's a key question because as part of the 35% projected growth in fiscal '26 this year, I don't think we have much more room to grow on the HPC side this year. We are expanding capacity between end of Q2 and end of the year. The plan is to increase output by about 50% over the next 6 months, and that's for the HPC handlers specifically. Between now and the middle of next year, the intent is to increase output by a little more than 100%. So double essentially the output or a little more than double between now and July -- June, July next year. And we do see a path to triple that output between now and the end of next year if the market takes us there on the HPC side. So that's essentially -- we are evolving that HPC production pipeline in Malaysia and our factory in Malaysia in line with the expansion of the business in this customer pipeline that yields $850 million here. That's the idea. It's a bit easier to do it in our own site. We are expanding the factory in Malaysia. We already started the fitting out the production floor. We're looking at a new construction of a building, essentially office building at this point. I think it will be suffice so that we can clear up production space on the current facility to be ready probably in Q1 of next year. It's quite a bit more challenging is exercising the supply chain. That's what's taking most of the attention right now so that we can get suppliers and more suppliers to support the expansion plan that we have in plan and been presenting here.
Our next question comes from Kevin Garrigan with Jefferies.
On the great results. Your new customers and engagement, can you just talk a little bit more about how those opportunities develop? Are these customers that are using competing platforms and are looking to switch? And how much additional opportunity do you see beyond the current pipeline that you have? So you mentioned $850 million. I mean is it pretty much sky is the limit at this point?
Well, there is always a limit. There's a finite number of customers out there. We're not really engaged with all of them yet. And I think we're quite honestly, fairly busy here, Kevin, with the sort of the 20 -- sorry, the 19, 20 customers that we have on the list. There's a lot to do here. There's a lot of projects, a lot of applications and qualifications in work. Will we add more? Sure. As these customers flow down this pipeline and it starts to get wider at the bottom, we'll start adding a few more at the top. But you can imagine who the names are, right? I'm not going to rattle them on the call. I shouldn't, but they're essentially the fabless and hyperscalers that are developing or have developed their own semiconductor GPUs or custom ASIC devices, network processors, in the variety of names there, including cancer processors and whatnot that they have their own names for their custom ASIC devices, right? So those are the constituents on this customer pipeline.
Got it. Okay. Yes, that makes a ton of sense. And then can you just talk about the -- you mentioned higher input costs. Any specific components that you can kind of call out? And are these components something that you expect to be a potential headwind for getting systems out the door at some point?
Kevin, it's Jeff. At the moment, it's mainly memory. Memory is sort of leading in the higher costs and longer lead times. And so we've taken advantage of advanced purchases and looked out over multiple quarters and made buys based on the quantities that we need for that time frame. So it's not an issue for Q3 and our guidance has taken into consideration all of the risks and potential constraints. So at the moment, we're working through it. Like Louis said, though, it is probably the biggest challenge at the moment is ramping supply chain.
Our next question comes from Craig Ellis with B. Riley Securities.
Nice job on the execution, guys. I'll start with some things that are just near term. You mentioned that we're looking for third quarter growth of 14% quarter-on-quarter with HPC and some of the traditional businesses contributing to growth. Can you provide a little bit more detail on the relative contribution of each as we look at this quarter's growth?
Yes. Craig, it's roughly about 50% HPC driven and 50% core business. So about $10 million out of each.
Got it. And then as we look ahead to the fourth quarter, remind us what you would think the seasonality would be of 4Q? And then as we look ahead, are there any particular items we should be aware of as we think about more one-off things that could be impacting the business beyond the third quarter?
To answer the first part of your question, we have stated that we now see revenue increasing about 35% year-over-year. So that puts us in a range of about $610 million to $615 million for the year. And to get there, that would basically be Q4 sort of flattish to Q3...
Yes. As far as seasonality, Craig, with utilization now broadly hovering at 80%, right? In a couple of markets here at 82%, a couple of markets at 77%, 78%. We're right at that threshold that if we see a seasonality pull back, I think it would quickly accelerate again in Q1. But we're not really sure exactly how that's going to play in Q4. So at the moment, we're viewing this core business kind of staying flattish going into Q4. As I mentioned before, I think we're kind of maxed out on output on HPC side in Q4 as well. We're still building that capacity through the end of this year. So that's the positive news. We did get a we did get here in early Q3 a single customer order for $26 million again for our Eclipse systems for the HPC market. And that's largely going to ship in Q4 as well. So that Eclipse output capacity is filling up quickly here in the fourth quarter already.
Yes. So you've got really good visibility on the fourth quarter. And going back to the comment on capacity and being pretty high with output relative to capacity, Louis, what are the levers that you have that can give you some wiggle room in the first half of next year before you get the big 100% increase around midyear to the extent that you do have any?
Well, I don't know if it's wiggle room, Craig. It's really a lot of hard work. from the supply chain side and operations side with the expansion of the factory in Malaca. We're also doing a small expansion in the Philippines because that's where we build the thermal heads. So I wouldn't necessarily call it wiggle room, but I think we're on track right now to -- like I said, to really double our output between the quarter just finished and beginning of next year -- end of this year, beginning of next year. So really looking forward to being able to deliver a $200 million to $250 million incremental HPC or $200 million to $250 million HPC revenue next year or more. I mean it depends. I think we have some wiggle room is more into next year where we could potentially triple the output as the market takes us there.
Got it. Yes. Okay. So the step-up is exiting this year to next year, not a year from now.
Our next question comes from David Duley with Steelhead Securities.
I'm sorry to kind of continue along the HPC questions. When I look at your funnel chart this quarter of qualified customers, you have 4 for $150 million and last quarter, it was 3 for $100 million. So there was one customer kind of adding to the qualified segment that added up to about $50 million. I'm kind of wondering when you look at the 5 other customers that are in Qual are how should we think about the mix of those customers? Are they all -- so I just take 5 and divide by the average there? Or how should we think about how each customer adds to the qualified SAM?
Yes. Dave, so we do have -- no, there's a bit of a range here. We have customers that we view as $30 -- low $30 million annual opportunity. We have a couple of customers that are likely to be individually $60 million annual opportunity, and I'm looking at a table here. And that's about the range, actually. It's a sort of $30 million to $60 million on a per individual customer base.
You just gave us -- gave me one of my other questions, which is the way your capacity expansion is unfolding, you will be able to double the revenue stream of your Eclipse high-performance computing segment in 2027?
Yes, into early 2027. And then from there, I think tying a little bit with Craig Ellis' question, we do have some wiggle room to expand further from there, and we'll see how this funnel develops and then we'll drive that expansion in 2027. But for now, you just hang your hat on, we'll be able to double the output we just finished in Q2 of this year by the end of this year, meaning into early 2027. Now remember, that's more than where we started in 2026. So the reality is if we're delivering $100 million to $110 million this year, we should have the capacity to do more than $200 million, probably close to $250 million by the beginning of next year, all things being linear throughout the year. And from there, we can expand more in '27 to exit at a higher rate in 2027 again.
Okay. And as far as -- just remind us who the key competition is for some of these slots and -- or are a lot of these brand new that you know where it's a jump ball? Or is there someone that is kind of the incumbent with a lot of these customers?
It's pretty much a single competitor, so to speak. I mean you can claim there's a second one -- a second competitor out there. But I would say there's primarily a single competitor, which has been the forever supplier at the Tesla contractors. It's Han Precision from Taiwan has been the primary competitor in this space. And as power levels are increasing in these more complex processors, right, all sorts of classification of processors. The management of power dissipation is becoming much more prevalent and driving a much stronger interest by, like I said, the fabless and the hyperscalers into finding a solution to the problem, which Cohu' Thermo Technology is one big company here in the Bay Area said Cohu Thermal Technology is sort of the best thermal in a market period in the story. Let's figure a way to make this happen.
Okay. Well, that's great to hear. Now just switching gears, final question for me is when you look at your core business, kind of -- obviously, you're guiding flattish. That seasonality is being overcome by the return of -- the cyclical business, so to speak. And I think when I listened to the big OSAT in Taiwan's conference call last night, they were basically they upticked their growth rate for wire bonding core assembly business from 13% this year to 20% for calendar '26, and they expect that growth rate to continue. So I'm kind of wondering -- and their utilization rates are also in the 80%, 85% and the quote was we can't keep up with purchasing equipment. So I'm wondering what your customer -- your core customer behavior is. There's a list of 8 or 10 of these guys, and they're not all -- some of them are automotively exposed. And so I think probably that those guys aren't inflecting yet. But could you talk about the customers that are inflecting and what their behavior is? Are they coming in and asking for big orders and big slots and whatnot?
Yes. We're seeing the earlier inflection on the industrial space, as we commented here. And yes, indeed, the ones that have inflected so far are coming in for the sort of additional volume that we've seen in the past where they're ordering somewhere between 10 and 20 systems in 1 -- so that's -- it's sort of coming back to that original pattern that we're familiar with from the past, predominantly with the industrial-based customers. And I think you can see from the earnings release, which ones are kind of spearheading the return to business. Congratulations on nice results.
Our next question comes from Denis Pyatchanin with Needham & Company.
So I have a question about the HPC raise. So I think previously, it was about $90 million expected for calendar '26. Now I think it's about $105 million, so $15 million incrementally higher. Maybe you can tell us is more of the upside coming from Eclipse handlers or the Non HPM inspection systems?
Denis, yes, you're right. If you pick up the midpoint of the ranges we gave before and now, it is really a $15 million increase. Note that the new range is also tighter. We originally had an $80 million to $100 million range, and now we're calling $100 million to $110 million. This whole increase is on the Eclipse HPC side, entirely there.
Got it. And then I have a question related to some of these challenges with the components specifically related to memory. So do you think you'll be able to pass on some of these costs to your customers within the next, say, 3 or 6 months? Or will you basically have to kind of eat that into your gross margin?
Denis, we have just started conversations with customers. So I would say stay tuned on that.
Our next question comes from Quinn Fredrickson with Baird.
Just on the cyclical piece, specifically on automotive, I think you mentioned in your prepared remarks that orders were soft there. It sounds like one of the few areas that was the case. So what's your visibility of the timing of a turn in that business?
Yes, Quinn, it's a good question. I think that is a market that has been a little bit more sporadic. We had -- if I'm not mistaken here, we had a bit of a bump in the last 2 quarters in the automotive. And then the last quarter, it kind of came back down a bit again. So I think it's been bouncing around. I would expect, frankly, that automotive will not be at 80% up until probably late Q1 or Q2 of next year. That would be my expectation. But like I said, it's been the one that's lagging a bit across the end markets on the core business side.
That's helpful. And then on OpEx, can you discuss just your ability to pursue the full $850 million HPC pipeline? Would you be able to pursue that full pipeline at this third quarter level of about $54 million of OpEx? Or would there be additional investment you'd have to make?
That is the plan and the forecast at the moment is to stay at this, and we think it's a bit elevated from our prior model, but to continue to invest and have OpEx remain pretty constant at this level, about $54 million. So we think that's a good level that provides the resources necessary to capitalize on these opportunities.
And just to help clarify as well, Quinn, for you and for others. When we talk about the $850 million, that's sort of an annual spend, right? That's what we see these customers spending annually on this class of equipment, which is largely eclipse for HPC. So if we were to capture the totality of this opportunity now immediately, we would see an $850 million revenue stream next year. That's not the case. We're qualifying over time, and we'll see how this evolves. The market is also changing and growing, but that's just to clarify that it's not $850 million over multiple years, it's $850 million spend per year for the customers that we're talking about here.
Our next question comes from Vedvati Shrotre with Evercore ISI.
The first one I have is, so with Agent AI, we are seeing the CPU to GPU ratios changing, right? We're seeing a higher CPU ratio versus GPU. How does this sort of play for you in the HPC opportunity? Like where do you see -- how do you see yourself participating in this kind of shift?
This is Luis. First of all, you're correct. We're seeing a much strong -- a very strong demand on the CPU side Traditionally, I would say the CPUs would have been at slightly lower power levels than the GPU, but that's actually changing. And CPU power in test, I should say, is really approaching the GPU levels. How does it change? I don't know that it really changed. The whole intention and purpose of our product configuration is to be flexible and be able to straddle across applications without having to change the capital equipment, but changing the configuration instead perhaps upgrading thermo heads for different applications. So we do a pretty good job right now straddling both CPU and GPU and reusability of the equipment. So it doesn't quite matter to us where the market and that ratio goes because that's one of the fundamental value propositions of our Equipped system. You can do both the thermal management at the higher power levels, but you can also use the same equipment at straddling down to lower power levels and across different applications.
Understand. So how about the penetration like there are 3 across the x86 and the ARM ecosystem, like how are you thinking about the pipeline and the penetration at the customers at the CPU suppliers?
It's -- I mean, I would have to go count to tell you where we are today. I don't know at the top of my head. But I'm going to venture to say well, I'm not going to venture to say. I'm going to say I would have to go count. But it is, again, not that relevant to us. We have probably strongest shipments in Q2 that were maybe by 86 or maybe that's going to be Q3 shipments. I would have to go look, but it straddles across both.
Understand. Okay. And then one last question. So on -- you talked about sort of the input costs and also alluded to your kind of the supply chain shortages. So maybe can you talk about how your lead times have changed in the last 3 months for the Eclipse tools?
Yes. I mean just for clarification, we haven't seen any shortages yet. And again, when we've got the opportunity to make some prebuys, we're doing it, and that's worked for us pretty well. However, these prebuys, particularly on -- mainly on integrated circuits are purchased at an increased cost. And so that there's the higher input cost. Although we have -- we're securing the supply, it's coming at a little bit higher cost.
And as I mentioned before, we're just initiating discussions now with customers about how we pass that on. And Eclipse handler right now is -- I think the lead times are still in check. We're holding well to 13, 14 weeks from receipt of PO. With that said, we have signals that the orders are coming. We have the customer forecast. And so we're getting ahead of it, so to speak, in part of what Jeff just said. So when we get the PO itself, we can respond. Now like I said, earlier, we got -- a couple of weeks ago, we got a $26 million single customer order, right? So as you can imagine, there are tens of systems. We're not going to ship all that quantity in 13 weeks lead time. We got a certain capacity per week, and that quantity is going to fill up multiple weeks of shipments. So 13 weeks to the first system shipment, but straddles across multiple weeks from there and then gearing up for the subsequent order from another customer now that would ship in the latter part of Q4.
Our next question comes from Christian Schwab with Craig-Hallum.
Great quarter and guide. I just have one question. It's been quite some time since we've been operating in our core business with 80% plus utilization and customers ordering to add capacity. Can you remind us historically when capacity is added when utilization rates go above 80% and the demand environment works to be continued. How many quarters or how long does capacity typically get added? You talked about initial orders kind of being 10 to 20, which was in line with historical norms for systems. But how long does that happen?
Yes. We would say about 6 quarters is probably the typical average or call it a cycle. And I've got this table in front of me, and it goes back to '21, and that was a pretty unique time frame in '21 and '22, and that happened -- we happen to be above 80% utilization for 8 quarters or at least 7. So I think 6 is probably the norm.
Great. Fantastic. And then lastly, I know we started a few quarters ago talking a little bit more aggressively or about M&A. But given the fact that the core business and the AI market growth is -- well, we're chasing extremely strong demand. Are you still looking at M&A or we don't want to get distracted with M&A when the core business is so strong?
Yes. This is Matt. Yes, I think you're right, right? The #1 priority is obviously execution. There definitely are opportunities to accelerate in some areas, our growth areas in HPC and in software, and we'll continue to look at build versus buy opportunities there. But yes, I think you nailed it. It's execution then looking at other ways to possibly accelerate some of our growth areas.
That concludes today's question-and-answer session. I'd like to turn the call back to Matt Hutton for closing remarks.
Thanks, operator. Before we sign off, I'd like to note that we will be attending the following investor conferences over the next 3 months: the Needham Virtual Semiconductor Conference on August 19, the Jefferies Semiconductor Conference on August 24 in Chicago and the CEO Summit on October 13 in San Francisco. If you plan on attending any of these conferences, please reach out to your conference contacts or let us know and we'll arrange for a one-on-one meeting. I'm also pleased to announce that Cohu will host an Investor Day on November 10 in New York City, where we will provide a deeper look at our strategy and long-term financial framework. Additional event details will be shared closer to the date. Thank you for joining today's call. We look forward to speaking with you soon.
This concludes today's conference call. Thank you for participating. You may now disconnect.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Cohu, Inc. transcript - plus 251,000+ transcripts from 12,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.
Get an API key View API docs →For developers and AI pipelines
Programmatic access to Cohu, Inc. earnings transcripts and 251,000+ others is available through the
EarningsAPI REST API and the hosted MCP server.
Quarterly plans from $105 - full transcripts, speaker segments, full-text search,
and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.