管理層發言
Good day, and thank you for standing by. Welcome to the onsemi Second Quarter 2026 Earnings Conference Call. Operator instructions were provided. Please be advised that today's conference is being recorded. I would now like to hand the conference over to your speaker today, Parag Agarwal, Vice President of Investor Relations and Corporate Development.
Thank you, Josh. Good afternoon, and thank you for joining onsemi's Second Quarter Results Conference Call. I am joined today by Hassane El-Khoury, our President and CEO; and Thad Trent, our CFO. This call is being webcast in the Investor Relations section of our website at www.onsemi.com. A replay of this webcast, along with our second quarter earnings release, will be available on our website approximately 1 hour following this conference call, and the recorded webcast will be available for approximately 30 days following this conference call. Additional information is posted on the Investor Relations section of our website. Our earnings release and this presentation include certain non-GAAP financial measures. Reconciliation of these non-GAAP financial measures to the most directly comparable GAAP financial measures and a discussion of certain limitations when using non-GAAP financial measures are included in our earnings release, which is posted separately on our website in the Investor Relations section. During the course of this conference call, we'll make projections or other forward-looking statements regarding future events or the future financial performance of the company. We wish to caution that such statements are subject to risks and uncertainties that could cause actual events or results to differ materially from projections. Important factors that can affect our business, including factors that could cause actual results to differ materially from our forward-looking statements, are described in our most recent Form 10-Ks, Form 10-Qs and other filings with the Securities and Exchange Commission and in our earnings release for the second quarter. Our estimate or other forward-looking statements might change, and the company assumes no obligation to update forward-looking statements to reflect actual results, change assumptions or other events that may occur except as required by law. Now let me turn it over to Hassane. Hassane?
Thank you, Parag. Good afternoon, and thank you for joining us on the call. Our second quarter results reflect the progress we have made in reshaping the business and our technology portfolio over the past several years and the strengthening demand environment. As we anticipated, the recovery continued to take shape during the quarter with continued strength in our AI data center business. We also saw multiple indicators of strengthening demand with China BEVs and automotive, for example, and energy infrastructure and medical and industrial already showing over market growth. Supply is tightening in several growth areas, lead times are extending, and we are seeing increases in both orders placed within lead time and customer escalations, all signs of a healthy recovery across the board. In Q2, we delivered $1.6 billion of revenue, non-GAAP gross margin of 39.3% and non-GAAP diluted earnings per share of $0.74, all above the midpoint of our guidance. These results reflect the operating leverage in our model with recovering demand, driving accelerated margin expansion and earnings growth. AI data center remains our fastest-growing market. We now expect AI data center revenue to more than double in 2026, driven by stronger demand, further accelerated by broader customer wins and expanding content across the entire Power Tree. We have expanded our role in the NVIDIA MGX ecosystem to supply advanced power systems designed to support the next generation of AI data centers, broadening the number of platforms where our intelligent power solutions are designed in. In addition, we secured 2 power supply platform wins with Great Wall, a leading provider of power solutions for China's cloud and AI infrastructure market. In parallel, we continue to add content in U.S. hyperscalers AI deployments with design wins supporting AWS power supply and battery backup systems. These wins create additional content opportunities for our differentiated high-voltage power portfolio, including silicon carbide solutions and reflect the value of our higher power efficiency and superior power density in next-generation AI power architectures. In 2026, we expect our silicon carbide revenue in AI data center applications to grow nearly 60% year-over-year. We expect our high-voltage revenue to accelerate as power requirements continue to rise and system architectures evolve towards 800-volt DC distribution, driving greater adoption of our intelligent power solutions from high-voltage infrastructure through low-voltage power delivery. As the only broad-based U.S. power semiconductor supplier with technologies spanning the full AI Power Tree, we are uniquely positioned to support this transition from the grid all the way to the processor. Importantly, this is not a position we earned overnight. It is the result of years of investment in solving complex power challenges, particularly in automotive, where power efficiency, thermal management, reliability and system integration have long been critical design requirements. Our opportunity extends beyond the data center into the power infrastructure required to support AI deployments. In our industrial business, we are increasingly seeing the benefits of the AI halo effect where AI growth is driving demand across the power infrastructure required to support it, including energy storage systems or ESS. We expect our ESS revenue to grow approximately 40% year-over-year in 2026, driven by higher year-over-year growth in North America with microgrid customers. During the quarter, we released our next-generation EliteSiC hybrid ESS module, delivering industry-leading 99.3% efficiency. We also began sampling our industry-first and the world's highest power density hybrid module platform at 500 kilowatts, which increases power density by 20% compared to our previous platform, delivering growth at accretive gross margins. Growth in AI workloads and increasing grid reliability and resilience requirements are expanding our industrial portfolio into higher-value infrastructure markets with greater semiconductor content. Turning to automotive. We continue to expand our content per vehicle through a growing portfolio of differentiated power, sensing and connectivity solutions. In China, our automotive revenue increased 13% in the first half of '26 over the same period last year against a total vehicle sales number that was down 4%, supported by expanding content per vehicle with customers like Geely, Zeekr and Xiaomi. With our market share gains in China EVs, we now expect silicon carbide revenue in that market to increase between 60% and 70% year-over-year as our market share gains continue and programs ramp across existing platforms and newer vehicle launches. In the U.S., we continue to gain share across EV disruptors with a recent example of our power content on Rivian's R2 platform, where our MOSFETs support power distribution throughout the vehicle's zonal controller architecture, while our silicon carbide solutions are deployed in the onboard charging system. These wins highlight our ability to participate across multiple vehicle domains as EV architectures continue to evolve. The demands of next-generation vehicle architecture around efficiency, power density and reliability increasingly mirror the challenges being addressed in AI infrastructure and energy systems, enabling onsemi to leverage decades of power expertise across multiple growth markets. A growing share of our recent design wins are coming from products introduced over the last 2 years, including our 10BASE-T1S Ethernet offering and our inductive and ultrasonic sensing products, reflecting the increasing contribution of Treo, our analog-mixed-signal platform, at favorable margins. By leveraging common technology building blocks across automotive, industrial and AI infrastructure applications, Treo enables faster innovation cycles and more efficient product development. We remain on track to double the number of products sampling this year, further strengthening our pipeline and positioning us to capture additional content opportunities as vehicle architectures evolve. Over a multiyear period, we expect to outgrow underlying vehicle production through content expansion, technology leadership and shared gains. More broadly, across automotive, industrial and AI data center, the industry is moving toward architectures requiring higher levels of power efficiency, power density and system intelligence, all of which place greater demands on power conversion, delivery and management. This is what we do. As we look ahead, our confidence in the second half is grounded in the momentum we are seeing across our key growth drivers. We now expect AI data center revenue to more than double for the year. At the same time, the AI halo effect continues to create incremental growth opportunities across energy infrastructure, where we expect ESS revenue to grow by approximately 40% this year. In automotive, we continue to gain content and share, particularly in China, where automotive silicon carbide revenue is expected to grow between 60% and 70%. As I wrap up, I want to highlight our announced agreement to acquire Synaptics. Beyond the compelling strategic and financial rationale, we are excited about the opportunities this combination creates for all shareholders. Synaptics' market-leading connected compute capabilities complement our strength in power, sensing and control at accretive gross margins. Our combination would leverage our manufacturing scale, global sales channel and mass market engine to drive growth across our highly complementary portfolio. We expect the transaction to close in mid-2027, subject to customary approvals. Let me now turn the call over to Thad to provide more details on our results and guidance for the third quarter.
Thanks, Hassane. Our second quarter results demonstrate the operating leverage in our model, with revenue up 9% year-over-year and non-GAAP earnings per share growing approximately 4x faster than revenue. Free cash flow nearly quadrupled and non-GAAP gross margin expanded for the fourth consecutive quarter. This reflects the earnings power of our focused portfolio, the benefits of our manufacturing cost actions, and we are entering the second half of the year from a position of strength as demand continues to recover. Our results were above the midpoint of our guidance range as we delivered revenue of $1.6 billion, driven by increasing demand in AI data center. Non-GAAP gross margin expanded 80 basis points sequentially to 39.3%, while non-GAAP earnings per share increased to $0.74. We generated $425 million of free cash flow in Q2 and returned $332 million to shareholders through share repurchases. Year-to-date, we have returned approximately 105% of free cash flow. Our ability to simultaneously invest for growth, expand profitability and return capital reflects the structural improvements in our business and differentiates onsemi today. Turning back to revenue for the quarter. Q2 revenue was $1.6 billion, up 6% sequentially and above normal seasonality even as we completed the final $35 million of planned non-core revenue exits. Across the business, our overall book-to-bill ratio has been significantly above 1 for several quarters and continued to strengthen. Given the accelerated ramp in AI data center demand, we prioritized shipments to AI data center over automotive and industrial. We expect this to normalize as supply rebalances to match demand. Automotive revenue was $781 million in the second quarter, down 2% quarter-over-quarter and grew 7% year-over-year. Consistent with previous years, the sequential decline was primarily driven by specific customer seasonality in Europe, offset by strength in China. Year-to-date, automotive revenue increased approximately 6% compared to 2025. Industrial revenue was $423 million, up 1% sequentially and 4% year-over-year, driven by strength in our focus areas of energy infrastructure, medical and factory automation, partially offset by declines in traditional parts of the industrial market. Total revenue for the other category in the second quarter was $400 million, up 34% sequentially, anchored by stronger-than-expected AI data center demand as well as growth in other end markets. Our AI data center business continued to grow in Q2, and 2026 is now on pace to more than double compared to 2025. Looking at the second quarter split between the business units. Revenue for the Power Solutions Group, or PSG, was $829 million, an increase of 13% quarter-over-quarter and 19% year-over-year. Revenue for the Analog and Mixed Signal Group, or AMG, was $546 million, an increase of 1% quarter-over-quarter and 2% decrease year-over-year. Revenue for the Intelligent Sensing Group, or ISG, was $229 million, a 3% decrease quarter-over-quarter and a 7% increase over the same quarter last year. Moving to gross margin. GAAP gross margin was 38.4% and non-GAAP gross margin was 39.3%, an increase of 80 basis points sequentially. We benefited from improved manufacturing performance and favorable mix. Utilization increased to 83% from 77% as we continue to quickly ramp production to support the increasing backlog for future quarters. In Q3, we expect utilization to be flat to up. We are seeing incremental increases in input costs, primarily in raw materials and external manufacturing. We are implementing a second round of price increases to offset these costs and expect to see the benefit over the next several quarters. The announced divestitures of our Mountain Top and Philippines manufacturing facilities further advances our Fab Right strategy of exiting subscale legacy operations to improve our cost structure. We expect approximately $35 million of annualized savings with the initial benefit starting in 2027 and the full savings realized in 2028. This represents approximately 50 basis points of the 200 basis points of gross margin improvement from our planned Fab Right initiatives. GAAP operating expenses were $358 million, including $41 million in restructuring expenses and non-GAAP operating expenses were $297 million. GAAP operating margin for the quarter was 16.1% and non-GAAP operating margin was 20.8%. Our GAAP tax rate was 16% and non-GAAP tax rate was 15%. GAAP earnings per share was $0.56. Non-GAAP earnings per share was $0.74, a 16% increase over the prior quarter. GAAP share count was 404 million shares and non-GAAP share count was 397 million shares. Turning to the balance sheet. Cash and short-term investments was approximately $3.9 billion with total liquidity of $5.4 billion, including $1.5 billion undrawn on our revolver. Cash from operations was $460 million and free cash flow was $425 million. We achieved record free cash flow margin on an LTM basis of 24%. Capital expenditures were $34 million or 2.1% of revenue. Inventory declined 9 days to 192 days and flat on a dollar basis. As expected, we continue to drain our strategic inventory, which is down 8 days. Our base inventory declined 1 day to 125 days, reflecting a healthy level of inventory supporting our expected revenue growth. Distribution inventory declined to 10.1 weeks from 10.8 weeks in Q1, with sell-through outpacing sell-in and our mass market revenue increased 20% sequentially. Looking forward, let me provide the key elements of our non-GAAP guidance for the third quarter. As a reminder, today's press release contains a table detailing our GAAP and non-GAAP guidance. We anticipate Q3 revenue will be in the range of $1.65 billion to $1.75 billion. Our non-GAAP gross margin is expected to be between 40% and 42%, which includes share-based compensation of $8 million. This represents a significant step-up function increase as we realize the benefit of increasing utilization since the start of the year. Given the improving demand outlook, we expect sequential gross margin expansion throughout the year. Non-GAAP operating expenses are expected to be between $303 million and $318 million, which includes share-based compensation of $33 million. Operating expenses are expected to increase at a slower pace than revenue, supporting continued operating leverage as we grow into our model. We anticipate our non-GAAP other income to be a net benefit of $18 million with our interest income exceeding interest expense. We expect our non-GAAP tax rate to be approximately 15% and our non-GAAP share count is expected to be approximately 395 million shares. This results in non-GAAP earnings per share in the range of $0.81 to $0.93. At the midpoint, EPS growth would outpace revenue by nearly 3x. We expect capital expenditures in the range of $40 million to $50 million, and we now expect capital expenditure for the year to be below 5% of revenue. In closing, the actions we have taken to reshape the company have created a structurally stronger and more efficient business, and we believe we are still in the early stages of realizing the full benefit of our model. We are entering the second half of the year from a position of strength with a stronger demand environment, a more focused portfolio and additional manufacturing efficiencies still ahead of us, we remain confident in our ability to drive sustainable value for our shareholders. We look forward to sharing more with you during our Analyst Day in New York on September 16. With that, I'll turn the call back over to Josh to open it up for questions.
分析師問答
Our first question comes from Vivek Arya with Bank of America Securities.
Hassane, I was hoping you would give us some more color on the automotive environment. Sales were down slightly sequentially, maybe some of it was because of exits. But how should we think about in the context of the 6% sequential you're guiding, how you expect your automotive business to do in Q3? And then what is kind of the typical seasonality in Q4? And the reason I want to ask about autos is that does the pricing lever apply to autos also? Because I imagine you are exposed to large OEMs and Tier 1s, so do you get the same benefit of pricing even when dealing with the automotive customers? So just commentary on demand and pricing and sequentials would be helpful.
Yes. There's a lot to unpack here, Vivek. Let me take it and if I miss any, just prompt me again. Overall, the quarter came in exactly as we expected. We believe we're shipping to true demand now that the inventory is behind us. That’s what we see as true demand. Regionally, automotive behaves very, very differently. I talked about some of the strength with share gains in North America. China is doing very well for us. All came in roughly where we expected. Q2, if you look historically in 2024 and 2025, is seasonally a down quarter. The numbers for the last few years have been about an 11% decline and a 4% decline. So Q2 is seasonality driven, given a couple of key customers in Europe specifically. Outside of that, we see the demand environment as stable. We expect to maintain our content gains with new products we are introducing, whether it’s the Treo starting to ramp in some zonal architectures already or silicon carbide, where we are expecting stellar growth coming out of China, again based on the competitive nature of the share gains I have discussed on these calls. Overall, we see automotive as very stable with a good outlook. As far as pricing is concerned, pricing for us is not really market-driven because the cost increases we are seeing are material in nature — substrates, gold — and many of those costs apply regardless of market. Therefore, the pricing actions that Thad and I discussed last quarter are offsetting those costs across all of our markets. We do take some targeted price increases where we see strength or tight allocation, but in general the cost offsets are applied across the board, including automotive.
Got it. For my follow-up, maybe one for Thad on gross margins. So Thad, the gross margins went up about 80 basis points, although I think utilization you mentioned went up almost 800 basis points. So is it that a bulk of that benefit you get to see in Q3? What kind of utilization should we assume for Q3? And what is the effect of mix and pricing? And I ask these questions because if I go back to the levels of revenue we saw for on towards the end of '24, it was in this $1.7-ish billion quarterly range, and your gross margins were already in the mid-40s. So I imagine that, that was a different time versus what we have. So just help us think through the effect of utilization mix and pricing for your current trajectory of gross margins.
Yes, Vivek, for the Q2 gross margin, you actually need to go back and look at the Q4 utilization because remember, there's about 2 quarter for utilization to hit the P&L. So in Q4, utilization actually dropped 6 percentage points from Q3 to Q4. So that would have been a headwind to Q2. So we actually offset that headwind with favorable mix primarily, and we saw that coming because our guide was 39%. We came in at 39.3%, so better than we expected. But we saw that favorable mix that we guided to. There really wasn't much of a margin impact at all from pricing because of the input costs that Hassane just walked through. So you can think about anything we did in pricing in the second quarter really didn't give a bump to the gross margin. The utilization increases now here in Q2 that we see going up will obviously impact us in the future quarters, right? So you can think about Q2 hitting Q4. So that's why we're very confident that you'll see additional gross margin expansion through the rest of this year as utilization has been consistently going up since Q4. So as the market improves, as the utilization continues to go up as we catch up supply to demand just because demand has moved so quickly, we'll be able to see that, that benefit coming through the P&L for the remainder of the year.
Our next question comes from Timothy Arcuri with UBS.
Thad, I just wanted to follow up on that comment on gross, or sorry, on utilization tailwind to gross margin. So is the right rubric still to think about 30 basis points per point of utilization? So, all else equal between now and Q4, the increase in utilization should drive margin up by about 150 to 200 basis points. Is that the right way to think about it?
Yes. The math is 25 basis points to 30 basis points of gross margin improvement for every point of utilization. So as we look into Q4, yes, you should expect a margin increase, all things being equal, that's assuming a consistent mix, those types of things. Now again, we had a richer mix here in Q2. So just keep that in mind. We saw that coming. But yes, we will see additional margin expansion in Q4.
Our next question comes from Joseph Quatrochi with Wells Fargo.
You talked about like prioritizing data center demand over auto industrial in the quarter. Was that any sort of impact to revenue? Or how do we think about that dynamic?
Yes. I mean, obviously, when we have some technologies specifically in power, where we have constraint, we did see some orders come in, what I would say, within lead time, and we had to make priority calls. So we did shift some, not just automotive specifically, but we did prioritize AI data center that took away from our other businesses in the short term as manufacturing really catches up to the updated signals of demand. So we made those calls within the quarter. You saw that strength in AI data center. We expect that AI data center, of course, to continue strength during the year. So it's not just a temporary Q2, meaning we got the gain. We got the designs. We will continue to ramp and then our manufacturing output will catch up to the, call it, automotive and industrial. So overall, we made those calls within the quarter. That's the long-term beneficial for the company, and we didn't have really a customer impact in the short term, but it is something that we're catching up to here in the third and fourth quarter.
And then just maybe as a follow-up, any sort of help in just thinking about the subsegments of the business looking into 3Q on the guidance front?
Yes. As you look forward, we expect auto to be up low single digits, industrial to be relatively flat, and we expect other, which has our AI data center to be up high teens. Those are all percentagewise.
Our next question comes from Quinn Bolton with Needham & Company.
I guess maybe just following up on that sort of guide. Other was up $100 million sequentially. It looks like it will be up several tens of millions. But the third quarter guide, that certainly implies a pretty healthy AI data center business. I know you guys are saying it's going to more than double, but is it going to be significantly higher than the previous $500 million target you talked about last quarter? I mean how much better than $500 million do you think you can do? Certainly, it seems like if all of the strength in other in Q2, Q3 is coming from AI data center, you've got a data center business with $500 million.
I'll let you run the numbers. The fact that we moved from not disclosing to doubling and now to more than doubling highlights the momentum we're gaining, which is not surprising. We've always said the AI data center is arriving where our technology is most competitive, and we're going to win on the strength of our technology baseline. We're not breaking it out further at this point, but we are seeing sustained momentum and strength, not just a quarterly spike. Looking ahead, the transition to 800-volt DC will drive more of our content. We're excited about our position; our investments have delivered to the market today and will continue to do so.
A follow-up question. In the prepared comments you said your North American silicon carbide business for data center PSUs would be up 60% to 70%. I would have thought silicon carbide was one of your lead products in the data center business, which looks likely to be more than double year on year. So was that comment specific to North America, or is there a reason silicon carbide may be growing slower than the rest of the data center business? Alternatively, what's driving the growth in data centers if it's not silicon carbide-based?
Yes. I've always said our growth in data center is literally from the wall to the core. We do not have a single technology that isn't outsized across the power tree, which gives us a very good distribution from what I call the high voltage because it's not only silicon carbide MOSFETs, we have silicon carbide JFETs, and we have silicon as well all the way to, I'd say, the SPS, which is right at the XPU. Our data center revenue is, I would say, very equal across all of them. So the growth that we are seeing in AI data center is not necessarily anchored on the high voltage. We are gaining share and ramping close to the core as well. But all of it is growing. And those are, by the way, all consistent with what we've said as far as wins that we've talked about over the last couple of quarters are now starting to ramp and contribute revenue. Silicon carbide is specifically part of this. Now, silicon carbide, obviously, we've made the investment thesis on automotive that's still winning in China, in both AI data center and automotive specifically. My commentary is more that we're seeing that growth now in North America, in PSUs and so on where high voltage comes in. But it is not only a silicon carbide story for us in AI data center. It's really across the board.
Our next question comes from Christopher Rolland with Susquehanna.
Yes. In regard to the AI opportunity, and I know you've hit a lot of this, but it was a considerable increase in the slide deck for the AI DC TAM. I think you went from $12 billion to almost $50 billion in that number. You also talked about an expanded role in NVIDIA MGX as well as a hyperscale opportunity, I believe, for battery backup at AWS. I guess, first of all, can you talk about that increase in the TAM and what from a product basis is expanding that almost fourfold.
Yes, a couple of things. The number we're anchoring on is 2030. The projection for the increase is really correlated to what I would call a gigawatt scale of capacity installed by 2030. If you take the installed base of power projected for 2030 and backtrack to convert that into the power installation needed at a data center or rack level, and then apply our content per rack, that's where the increase comes from. So point one is simply the volume that is proportional to the gigawatt increase we've modeled. This is based on market data; we see the correlation with that data and are using it to inform our content assumptions. Point two is broader than just the AI data center. Our TAM has increased because we've targeted, invested in, and introduced products we did not have when we last updated this a few years ago at Analyst Day. Those investments are turning into opportunities as the markets for those products develop. It's especially pronounced in AI data centers given the content growth I mentioned — for example, 10x on high voltage — but our TAM is growing across the board. Think about vertical GaN, additional opportunities with Treo, and silicon carbide JFETs that are finding applications in both data centers and industrial use cases such as solid-state circuit breakers and solid-state power disconnects. Many of these opportunities did not exist a few years ago. When you add them all up, that represents the opportunity in front of us, and it's the result of investments we've already made in products we’re delivering over the next five years. Stay tuned, and we’d love to see you at Analyst Day.
I will certainly be there, Hassane. The other big change here, I think, was the EV TAM that went from like $26 billion to $63 billion. Very interested as to what's driving that. Obviously, the silicon carbide opportunity. I think maybe there was a zonal architecture commentary as well. But what else is almost doubling that EV TAM for you guys?
So you can take a look at it. As you get more penetration of EVs and more acceleration of EV adoption, remember we’ve talked about two things in electrification. One is EVs becoming a higher percent of SAAR. And number two is the acceleration driven by the financial side of silicon carbide and the advancements we’ve made, meaning silicon carbide is becoming a higher percent of content within EVs compared with standard silicon. We’ve also won designs in both China and North America for hybrids, where we’ve always thought hybrids would remain with silicon or IGBT. Given our competitive position and what we’ve introduced, and the extended-range hybrids that OEMs are looking for, that’s content we already have products for. So that’s part of our increased TAM. You mentioned it in addition to a lot of the Treo opportunities we’ve had with 10BASE-T1S ultrasonic sensing and many onboard charging opportunities, one of which I highlighted, all of which are coming into our technology domain. We took a fresh view of where the market is headed given the trends and the technologies we have. You can think of it as a bottoms-up TAM adjustment we’ve done as part of our work for Analyst Day and the strategic work we have been performing to refine and deploy our strategy.
Our next question comes from Blayne Curtis with Jefferies.
I actually want to go back, I think, a prior question. The growth in other of $100 million, I just want to make sure that was a correct statement that data center was the primary driver there before you were giving out percent of revenue. I don't know if you can just dial us in a little bit better. I want to make sure that, that is the source of the growth.
The AI data center was the primary driver, but other markets were also up. So it wasn't solely AI data centers; however, as a percentage, that bucket showed the highest growth.
Got you. Thad, I wanted to ask: you mentioned lead times were extending. Could you give us some context for what that means? And when you mentioned prioritization, I'm curious about utilization remaining flat. Is that just a mix issue that prevented improvement in deliverables? I'm trying to understand; please walk us through the flat utilization.
Yes. Let me start with the lead time. So lead times stretched out from about 27 weeks to somewhere around 32 weeks on average. That's across the entire portfolio. We have some that are shorter, some that are longer, obviously. But we did see lead time stretching out. As Hassane said, we're seeing a lot of orders inside of lead time, a lot of escalations kind of points to, I guess, kind of a healthy environment, demand environment, which kind of gives us that indication of the future. And I mentioned the book-to-bill continues to improve. So we're getting better visibility even into '27, and we've got, in some situations, some customers ordering out into '28, trying to lock up supply. In terms of the utilization, look, we've had to increase utilization very sharply to reflect the increase in demand that we've seen, sharp, right? So if you think about a cycle time, a fab cycle time all the way through from beginning a wafer to a finished goods, it can be 4 to 6 months. And so you've kind of got to get that ramped and then you can catch up with demand. So that's where we expect that happening. Now if the demand continues to increase beyond what we're seeing, we'll continue to take that utilization up. But right now, we believe we can catch up, and that's why utilization doesn't have to go up at least in the next quarter.
Our next question comes from Tom O'Malley with Barclays.
I just wanted to complete the deep dive on gross margins. I know you've gotten a bunch here, but the last remaining component there is depreciation. If you look at depreciation in the quarter, I think it was at a low for over the past 2 years plus. I just want to make sure nothing changed there. And then as you look forward, obviously, you're going to see this utilization step up. But anything unexpected from a depreciation side or anything expected that we can look at September, December? I know that like you've kind of trended in this mid-600 basis points range. Anything changing there just because I know you are doing a lot of these exits that should naturally come down, I would imagine. Anything to help there?
Yes. No. Yes, Tom, so we did that, we took capacity offline about 12%, right? And so that's starting to hit the depreciation as that capacity is coming offline. I would say the depreciation that you're seeing now is really kind of steady state at this point, right? And if you look at our CapEx, it's maintenance CapEx. I said that's going to be below our 5% target for this year. I think our target over multiple years is in that mid-single-digit percentage range, but we're keeping it tight this year, and it's primarily maintenance. So no, no change to depreciation from what you're seeing as a steady state right now.
And then just one if I could sneak it in on the disti side, I saw it stepped up a little bit. Traditionally, you see disti kind of aligned with China, so I saw that step up as well. Is that the right way to think about it? And is that more auto or industrial related? Any color on that step-up in disti would be helpful.
So the mass market actually increased 20% quarter-over-quarter. You can think about a lot of that mass market as industrial. There is some automotive in there. But I wouldn't call anything else mass market, and that's a good leading indicator, right? We've been investing and putting inventory into the channel to satisfy that mass market. We're now seeing that happen with the 20% sequential growth in that revenue.
Our next question comes from Tore Svanberg with Stifel.
I just had a follow-up on some of these capacity/utilization questions, Thad. So obviously, Fab Right strategy makes a lot of sense, but it sounds like you do have some delinquencies. So I'm just curious like why wouldn't you ramp the capacity utilization faster? And if you do get continued upside orders here, how much flexibility do you have with your external partners to keep ramping capacity?
Sorry, you cut out for like a few seconds in the middle, but I think I caught onto how would we reconcile divestitures with increasing demand to capture?
No, I was asking how much flexibility do you have with your external partners to ramp more capacity.
Yes. So external, obviously, everything is constrained, whether it's internal or external. When it comes to short lead time, we believe we have good capacity and good allocation coming from the outside. The issue is not really capacity of what we can support as far as max capacity or max revenue. It is how quickly the demand came in. We can service a lot of demand from die bank, but at some point, like Thad said, when we start a wafer today for an order that we got within the quarter that we didn't anticipate, you have a short-term allocation we have to deal with. That's both internal and external, even if you have the capacity secured is how quickly you are able to get to the capacity. So we feel comfortable about external supply. We're not fighting for a lot of the advanced nodes like on the compute side. Our capacity is well understood. A lot of our capacity is also internal. We still do over 60% on internal manufacturing at onsemi even with the divestitures. So all of these put together put us in a good place, and that's what Thad said. It's a short term where we're going to catch up. If demand continues to accelerate, we'll take utilization up. We're not capped out. It's just how quickly we get there. And in the short term, we just have to make these calls of prioritization until manufacturing output catches up.
Yes. Not all capacity is fungible, right? So we do have some supply constraints on certain lanes. That's where we've just got to catch up with that demand, and then we believe we can service it.
Yes, that makes sense. As a follow-up, Hassane, you talked about power from grid to core and said content per rack could grow from $15,000 to $115,000. How should we think about that journey? Will it be gradual, or will there be a step change, perhaps when 800-volt systems ramp and maybe by 2028 when that market takes off? Any additional color you can add would be really helpful.
Yes. Look, an 800-volt DC rack with the backplane at 800 volts is a step up from the prior rack. You won't see that immediately as a revenue step because there will be a build-out. As these racks are deployed into newer data centers and as older data centers decide whether and how quickly to retrofit, we model that as a gradual but very healthy growth trajectory rather than an immediate step. It will be an aggressive long-term expansion. You can think of the ramp starting around the end of 2027 or the beginning of 2028, give or take a quarter, and then maturing depending on data center build-out. We are already starting to see some platforms designed this way, so the content looks justified in our model. The inclusion of vertical GaN, JFET and similar technologies is already solving density and efficiency problems in these new architectures.
Our next question comes from Jim Schneider With Goldman Sachs.
In light of all the metrics you cited around lead times extending and everything else in terms of expedites, and normally, at this point in time, I would have thought you would see OEM customers' inventory start to expand or customers taking a little bit more risk on their own balance sheet. Is that something you're seeing yet at this point in the cycle or not quite yet in terms of absolute dollars or in terms of days?
No. Obviously I can't speak for our customers unless they publicly talk about it. From our perspective, we are not seeing an inventory build. We are seeing consumption-driven demand, which is why you see it specifically in auto and industrial; we believe we are shipping to end demand. There are technologies where you do see constraints, and we are making a lot of hard decisions, the same things we did during COVID. This quarter we decided to reduce days of inventory in the channel to keep close control of potential inventory build. We are very disciplined about this. We are encouraged by the market signals I mentioned and by the opportunity in AI data centers, but we are maintaining a disciplined, process- and KPI-based approach to manufacturing and resource deployment so we do not get ahead of it. You can see that in our inventory position and the reduction across the board. All metrics show a disciplined approach even against the backdrop of a strong market.
That makes sense. And then just in terms of the input cost increases you talked about, you mentioned increasing price to offset those input costs over the coming quarters. As you look into, for example, 2027, what are your suppliers telling you about further potential input cost increases? And maybe if you could handicap your level of confidence in staying ahead of those in terms of pricing or potentially significantly ahead of those?
Yes. We have a very broad range. With some suppliers we have long-term views and provide long-term visibility, including potential upcoming cost increases. The consistent message is that costs are not going down. Because they are not going down, we are making sure our business reflects that, whether through actions now—since many increases will be effective in Q4—or by communicating changes for 2027. We are not expecting any reductions. It is not one size fits all, but we have visibility on where we are landing for 2027 and will gain more clarity as we get closer. We have started these discussions, and nothing is coming down, so I do not expect a softer pricing environment in 2027.
Our next question comes from Joe Moore with Morgan Stanley.
Utilization in the low 80s, how do you think that compares to some of your direct competitors? And if we sort of continue to get into a more supply-constrained environment, do you anticipate that you'd be able to take share given where you're sitting with capacity?
Yes, Joe, look, I don't want to comment about our competitors. I can tell you what we're doing in our business, right? Again, we've taken this utilization up quick in response to the snapback in demand. For us, getting to kind of in that low 90% with 92%, 93% is fully utilized. Once we get there, we start flexing to the outside. So there's a certain amount of our products today that we're manufacturing inside that we can take to the outside. So I don't think we're going to be limited. For us to really start to get to a point where we even think about getting capped out, revenue is still 25%, 30% higher than the run rate today. So we're not worried about that given our flexibility with our fab right and our ability to flex inside and outside.
Look, the focus for our growth and new product introductions, take Treo for example: that investment in capacity is complete. It’s in East Fishkill, and East Fishkill is not fully utilized, so we have the runway to ramp. As Thad said earlier, not everything is fungible, but in the areas where we have growth and have invested, we have runway and will continue to ramp aggressively across all markets.
That's helpful. And then in terms of things getting tighter and starting to see some constraints, are there any specific areas that are more impacted than others? Any hotspots, silicon carbide different than everything else? Just any sense of where supply/demand might be different?
No. If you think about power broadly, it’s not just high voltage; for example, data centers consume power at many levels, not only high voltage. As I mentioned earlier, we’re ramping everything from the wall to the XPU, covering high voltage down to low voltage, and we see constraints in some of those lanes. We have one of the most efficient power products, so demand has been outsized, and we’re producing it in multiple fabs now. We are increasing capacity and flexibility to support it. Lead times have extended, and we are working with customers to support them. We will take share where others cannot, but that’s not the only reason we’re winning. We’re winning because the product is superior. Multi-sourced products tend to have less attractive margins, and that’s not the business we play in—we exited that business. We win where our products provide true differentiation to end customers, and that’s consistent. You see that in silicon carbide. I called out China specifically because we win across the board based on efficiency and product strength. You see that in China, North America, and Europe. That is our focus, our investment, and how margin expansion will continue.
Our next question comes from Vijay Rakesh with Mizuho.
Just a quick question. On the AI data center side, I saw the TAM is $48 billion, a big increase in that opportunity. Is there a way to look at how your content breaks out between site power, power in the rack, and the compute rack as you previously broke out? What is the split of the opportunity set within that?
Yes, I'm going to give you an approximate. And the approximate is if you look at the numbers we've given as far as content per rack, and you think about high voltage, which I would put the first order in the side card going to 50% of the total, $115,000 of content, 50% of it is high voltage, 50% is medium and low voltage. That gives you that split 50-50. Today, on the $15,000 of content, that split between high voltage and then the rest is 70-30, 30 is high voltage, 70 is medium to low. So the increase from a dollar content is on the high-voltage side, but both are going up, one is going up 10x, one is going up like 5x to 6x. So both increasing, but an outsized increase is, of course, in the high voltage given the market trend we are talking about.
Got it. And then, when you look at your auto, industrial, and data center other, is there a way to look at how the order trends are by the different geographies? Like what are you seeing in the U.S. versus Europe versus China, I guess?
Yes. I think if I take it in order, automotive specifically. All 3. Okay. It's very, very different across all three. So I would say, for automotive, I'll rank order: China, U.S., Europe. For industrial, I would say China, U.S., Europe. And AI data center is really the U.S. and some China — I mentioned Great Wall on this call — that's starting for us as well, but primarily in the U.S. In Europe, you have, of course, the AI halo that I call out within our industrial business, not on AI data center, but it is an AI halo that's supporting it. I think that's kind of how you can think about it.
I would now like to turn the call back over to Hassane El-Khoury, President and CEO, for closing remarks.
Thank you all for joining us today. Before we close, I'd like to recognize our employees around the world for their dedication, innovation and execution. The momentum we are seeing across our business, the opportunities ahead of us and our confidence in the future are all made possible by their hard work and commitment. On behalf of the leadership team, thank you for everything you do to serve our customers and move onsemi forward.
Thank you. This concludes the conference. Thank you for your participation. You may now disconnect.