Good day, and thank you for standing by. Welcome to the onsemi third quarter 2026 earnings conference call. At this time, all participants are in a listen-only mode. After the speaker's presentation, there will be a question-and-answer session. To ask a question during the session, you will need to press star one one on your telephone. You will then hear an automated message advising your hand is raised. To withdraw your question, please press star one one again. 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 Corporate Development and Investor Relations. Please go ahead.
Thank you, Daniel. Good afternoon, and thank you for joining onsemi's third 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 on the Investor Relations section of our website at www.onsemi.com. A replay of this webcast, along with our third quarter earnings release, will be available on our website approximately one 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 will make projections or other forward-looking statements regarding the 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-K, Form 10-Qs, and other filings with the Securities and Exchange Commission, and in our earnings release for the third quarter.
Our estimates 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, and thank you all for joining us this afternoon. Our third quarter results reflect continued strength in our business and the momentum we are building across our strategic growth markets, with growth above seasonal trends, continued acceleration in AI data centers, and improving demand in industrial. We see particularly strong demand where power density is becoming the key differentiator from AI data centers and the grid infrastructure supporting them to energy storage and next-generation vehicle architectures. These are strategic markets where we focused our investments and are now seeing that strategy translate into revenue growth, margin expansion, share gains, and new customer wins. AI data center continues to be our fastest-growing market, scaling more rapidly than we anticipated entering the year, with new wins that started to ramp.
In automotive, we continue to gain content and secure new platform wins in every region, while industrial demand continues to improve, led by power infrastructure and energy applications. In Q3, we delivered $1.72 billion of revenue, non-GAAP gross margin of approximately 41.2%, and non-GAAP diluted earnings per share of $0.90, all above the midpoint of our guidance. The combination of stronger demand, improved mix, and disciplined execution is driving meaningful operating leverage in our model, with earnings growth significantly outpacing revenue growth. We remain supply-limited in certain technology segments of our business, and we made the short-term decision in the third quarter and into the fourth quarter to allocate to markets where we see strong end demand, like AI data centers and industrial, which came in much stronger than we originally planned.
In automotive, we are supporting all deliveries for end demand, but we are not supporting inventory replenishment backlog. As we balance supply and demand, we expect 2027 automotive growth to be in line with our long-term target. In AI data center, our revenue covers the complete power tree, from power supply units, or PSU, to low-voltage smart power stages near the processor, where increasing compute is driving greater power density requirements and higher semiconductor content. In the third quarter, our AI data center revenue grew 39% sequentially to $180 million, and we now expect to deliver approximately $620 million of revenue in 2026, an increase of 140% year-over-year. For 2027, we still expect our AI data center revenue to double over 2026 as we continue to ship with new product ramps.
In the third quarter, we began volume shipments into multiple marquee platforms, including Google, with both silicon carbide and silicon MOSFETs for PSUs and 54, 12-volt applications. The transition to 800-volt architectures creates the next content opportunity for onsemi, moving AI data center power directly into our strength. Our comprehensive low-to-high voltage power portfolio spanning silicon carbide and vertical GaN remains a competitive advantage, allowing us to address direct conversion from 800 volts to the 54, 12, and 6-volt rails required across the AI rack. As power systems become more complex, customers increasingly need solutions that go beyond individual components. At last month's OCP Summit, we showcased our 800 to 6-volt solution using our differentiated vertical GaN technology alongside our gate driver and MOSFETs to deliver industry-leading efficiency and world-leading power density for one of the most challenging conversions in the AI rack.
The move to 800V high-voltage DC also accelerates the transition from mechanical to solid-state power disconnects for fast turn-off from solid-state transformers and circuit breakers to hot swap applications. Our silicon carbide JFET portfolio delivers superior thermal performance and is already designed in and contributing to our data center revenue. At Investor Day, we talked about the investments we are making to build differentiated capabilities ahead of the inflection points in our markets, and we continue to do that as power density becomes increasingly critical across AI data centers, automotive, and industrial. One example is EPP, our Embedded Power Platform introduced at Investor Day. EPP uses the silicon wafer itself as the package to solve electrical, thermal, and mechanical constraints simultaneously.
We are pleased with the customer interest in the platform since the announcement, reinforcing our conviction that power density is becoming an increasingly important differentiator across automotive, AI data center, and industrial applications. EPP is already sampling to alpha customers across these markets and is manufactured using our 12-inch silicon process at East Fishkill. Today, we are also excited to announce the acquisition of Ferric, which further extends our capabilities into integrated voltage regulators for IVR and puts us in the leadership position for power delivery to the next-generation AI processors. As power requirements of AI processors continue to increase, IVR incorporates the complete power regulator, including controller, power switches, capacitors, and magnetics, in one thin package and places it closer to the GPU, reducing power losses and solving the power density problem at the core.
Ferric offers the most compelling IVR solution in the industry with its industry-leading efficiency and transient response performance in an exceptionally compact footprint with best-in-class Z-height. We expect Ferric to increase our market share in the Vcore stage of the AI data center power tree. While this is a multi-year opportunity for onsemi well beyond 2030, we are raising our 2030 AI data center revenue target by $500 million to greater than $3 billion. Moving on to our industrial business, the AI halo effect is driving momentum beyond the data center itself. We expanded hybrid EliteSiC and IGBT module wins with multiple leading customers, including Sungrow and Sineng, driven by the 99.3% efficiency platform we introduced last quarter. Also in our industrial business, we see increasing momentum in robotics and physical AI, an area where our capabilities in power and sensing are highly complementary to Synaptics.
Demand from first-person view, or FPV, drones is driving increased shipments of our 2-megapixel and 8-megapixel Hyperlux image sensors, and we are engaged with more than 10 leading drone manufacturers. We continue to diversify across the robotics market and have secured multiple new designs wins with leading U.S. humanoid makers across power, connectivity, and sensing. For sensing, we are advancing our indirect time-of-flight, or iToF, technology to deliver world-class depth resolution of 1 millimeter, up to 10 times better than competing solutions, enabling the precise depth perception critical for robots to interact with the physical world. As physical AI scales, these applications bring together many of our core capabilities across power and sensing, creating a significant new content opportunity for onsemi. Turning to automotive, we continue to win new customer designs across all regions.
As the industry moves toward electrification, higher level of autonomy, and software-defined vehicles, the investments we have made in new products continue to expand our content per vehicle up to $2,000 as architectures evolve. These transitions create multiple vectors of content growth for onsemi, even in a relatively flat SAAR. We continue to gain share in the North American EV market across traction inverter designs and new onboard charger platform wins, in addition to the Treo-based 10BASE-T1S ramp we already started. Across European OEMs, we expanded our design wins across our power portfolio as they also start the transition to 48-volt zone. In China, we continue to expand our leading position in high-end OEMs for their 900-volt electric vehicles, with new 1,200-volt silicon carbide traction inverter wins at NIO for their flagship sedan and Geely ZEEKR's rapidly expanding export model.
As we look ahead, we are increasingly confident in the trajectory of the business. AI data center continues to scale rapidly while improving demand in automotive and industrial, content gains, and new platform wins create multiple vectors of growth. All these trends are converging around the areas where onsemi is differentiated in intelligent power and sensing, and our disciplined execution positions us to translate that growth into stronger earnings and cash generation. I want to thank our employees around the world for their continued focus and execution. Their work is what enables us to deliver for our customers and create shareholder value. Now let me turn it over to Thad to provide more detail on our results and guidance for the fourth quarter.
Thanks, Hassane. We believe the recovery is gaining momentum, and we are in a fundamentally stronger position than in prior quarters. We have more content, a broader market opportunity, and substantially greater earnings power. With much of the heavy lifting and investments behind us, we are positioned to convert revenue growth into earnings and free cash flow growth. In the third quarter, we continue to see this benefit in our financial results. Q3 revenue was $1.72 billion, above the midpoint of our guidance and up 7% sequentially. Non-GAAP gross margin expanded by 190 basis points quarter over quarter to 41.2%, and non-GAAP earnings per share was $0.90, growing four times faster than revenue growth year over year. During our investor day, we discussed having consistently invested ahead of major market inflection points, building capabilities organically and supplementing them through strategic M&A to accelerate time to market.
AI data center is a clear example. As architectures evolve, we are continuing to strengthen our capabilities across the power tree to position onsemi to capture more of that opportunity. Today, we announced the acquisition of Ferric for $600 million, plus up to $175 million in earnouts tied to the achievement of revenue milestones. As Hassane mentioned, Ferric puts us in a leadership position in power delivery for next-generation AI processors. We expect Ferric to be accretive to our growth and gross margin targets, with EPS accretion at scale and no meaningful near-term financial impact. This is a multi-year opportunity, raising our AI data center revenue target by 20% to more than $3 billion in 2030, and continues to accelerate thereafter. We expect the transaction to close in the fourth quarter, subject to customary closing conditions. Additional details on the acquisition are available on our investor relations website.
Turning to revenue for the third quarter, Q3 revenue was $1.72 billion, up 7% sequentially and 11% year-over-year. Revenue was $22 million above the midpoint of our guidance, with growth in industrial and AI data center. Automotive revenue was $777 million, down 1% sequentially and year-over-year. The sequential decrease is primarily due to the reallocation of supply to the AI data center and industrial end markets. We expect automotive revenue to decrease high single-digit percentage points in the fourth quarter, reflecting continued supply reallocation as well as timing and phasing of customer deliveries. Q1 is expected to return to normalized levels as supply comes online. Industrial revenue was $467 million, up 10% sequentially and 9% year-over-year. We saw strength in energy infrastructure, factory automation, and the broader industrial market.
We expect industrial revenue to increase low double-digit percentage points in the fourth quarter. AI data center revenue was $180 million, up 39% sequentially and 175% year-over-year. Demand continues to accelerate, and we now expect our AI data center revenue to reach approximately $620 million in 2026, increasing mid-teens percentage points in the fourth quarter. In 2027, we expect this to double off a higher base. Please note, beginning this quarter, we are enhancing our disclosures by separately reporting AI data center revenue as a fourth market segment to provide greater transparency into this important growth driver for onsemi. You will see this in our supplemental materials on our investor relations website and in our 10-Q.
Revenue for the other end market segment, now excluding AI data center, was $299 million, up 11% sequentially and 10% year-over-year due to normal seasonality in the wireless segment. We expect our other revenue segment to increase high-teens percentage points in the fourth quarter. Looking at the split between business units, revenue for the Power Solutions Group, or PSG, was $890 million, an increase of 7% quarter-over-quarter and 21% year-over-year. Revenue for the Analog and Mixed-Signal Group, or AMG, was $597 million, an increase of 9% quarter-over-quarter and 2% year-over-year. Revenue for the Intelligent Sensing Group, or ISG, was $235 million, an increase of 3% quarter-over-quarter and 2% year-over-year.
Moving to gross margin, GAAP gross margin was 41.4%, and non-GAAP gross margin expanded for the fifth consecutive quarter to 41.2% due to utilization improvements earlier in the year and favorable mix to new products. Utilization remained flat at 83%. We expect similar levels in Q4, with utilization increasing as we move into 2027 in response to improving demand signals. Our fab-right strategy remains an important contributor to the long-term margin opportunity. In Q3, we closed the previously announced divestiture of a backend facility in the Philippines, and we remain on track to close our divestiture of the Mountaintop fab exiting 2027. We are operating with a more focused and efficient footprint, creating additional opportunity for gross margin expansion over time. GAAP operating expenses were $335 million. Non-GAAP operating expenses were approximately $310 million. GAAP operating income was $378 million, or 22% of revenue.
Non-GAAP operating income was $400 million, or 23% of revenue. GAAP net income was $314 million, and non-GAAP income was approximately $354 million. Our GAAP tax rate was 21%, and our non-GAAP tax rate was 15%. GAAP share count was approximately 397 million shares, resulting in GAAP earnings per share of $0.79. Non-GAAP share count was approximately 392 million shares, resulting in non-GAAP earnings per share of $0.90. Turning to the balance sheet, cash and short-term investments were approximately $4.4 billion at quarter end. Total liquidity was $5.9 billion, including $1.5 billion undrawn on a revolver. Cash from operations was $516 million, and free cash flow was $489 million, or 28% of revenue. Capital expenditures were $28 million, or 1.6% of revenue, and we expect full-year capital expenditures to be approximately 2% of revenue.
Inventory increased $22 million sequentially, but declined on a days basis by six to 186 days. Strategic inventory continues to drain, declining seven days in Q3, and base inventory increased by one day as we continue to support higher demand. Distribution weeks of inventory declined to 9.9 weeks from 10.1 weeks in Q2, with sell-through continuing to support a healthy channel and mass market growth. We repurchased $20 million of shares in Q3, with activity limited by trading restrictions related to the Synaptics acquisition and renegotiations. Looking forward, let me provide the key elements of our non-GAAP guidance for the fourth quarter. As a reminder, today's press release contains a table detailing our GAAP and non-GAAP guidance. We anticipate Q4 revenue will be in the range of $1.715 billion to $1.815 billion, representing above-seasonal growth of 2.5% sequentially at the midpoint.
We expect non-GAAP gross margin to be between 41.5% and 43.5%, and includes approximately $8 million in stock-based compensation. At the midpoint, this represents 130 basis points of sequential expansion. Non-GAAP operating expenses are expected to be between $303 million and $318 million. This includes approximately $34 million in stock-based compensation. We anticipate non-GAAP other income to be approximately $27 million favorable. We expect our non-GAAP tax rate to remain at 15% and our non-GAAP diluted share count to be approximately 392 million shares. This results in non-GAAP earnings per share in the range of $0.95 to $1.07, and we expect capital expenditures in the range of $40 million to $50 million. The fourth quarter outlook reflects continued demand momentum and another step forward in our model.
Based on our current visibility, we expect this momentum to continue into 2027, with continued revenue growth and gross margin expansion as we scale our differentiated products and realize the benefits of our more efficient operating model. As we look ahead, the opportunity is about more than a recovery in demand. We have built a stronger business with significantly greater earnings and cash flow potential. Our portfolio is more focused, our manufacturing footprint is more efficient, and our path to growth is capital efficient. As we exit the year, our priorities are clear: execute on the demand in front of us, scale our differentiated platforms, and capture the large market opportunities ahead. We believe we are well-positioned to deliver sustained growth and expand margins, creating long-term value for our shareholders. With that, I'll turn the call back over to Daniel to open up the call for questions.
As a reminder, to ask a question, please press star one one on your telephone and wait for your name to be announced. To withdraw your question, please press star one one again. In the interest of time, we ask that you please limit yourself to one question and one follow-up. Please stand by while we compile the Q&A roster. Our first question comes from Vivek Arya with Bank of America Securities. Your line is open.
Thanks for taking my question. Hassane, you have been sort of, I guess, rightly de-emphasizing automotive in Q2 and Q3, and I think even in Q4, with this signaling of, I think, high single digit, from what I heard. But then you suggested that next year, auto would go back to growing per your long-term target, which I think is up whatever, 9% or 10%. I am curious, does underserving that market have any market share implications longer term? Where does the extra supply come from to start regrowing in that market from next year?
Yeah. Just one thing, I want to correct the words you used. We are not de-emphasizing the automotive market. Automotive remains strategic as an end market, and even in our five-year outlook that we highlighted in Investor Day, automotive still is the largest market, and that is not going to change. What you are seeing today, or this quarter and next quarter, what we have hinted at is short-term in nature, and it is supply-driven, not strategically driven. What I mean by that is we are not losing share, because we are working closely with the OEMs, as I mentioned in my prepared remarks. We are supporting all of their end demand needs.
What we are holding back, again, with conversations with them and alignment with them, is their replenishment that they would have liked to see during the quarters. We will resume shipment in Q1 as the supply comes online.
We have always had the supply coming online or increasing supplies in the technologies we have been constrained on, but you saw the results from industrial and AI data center even beat our guidance and our expectations. In the short term, we decided to shift the mix a little bit, but that is not a strategic decision. Our strategy remains the same as we have highlighted in Investor Day. That is why we are reiterating 2027, resuming the growth, which tells you there is no share loss. It is purely a replenishment, not an end demand, and it is short-term in nature.
Got it. For my follow-up, maybe Thad, if you could walk us through the puts and takes of gross margins, Q3 to Q4, volume mix, utilization. Then let's say, conceptually next year, if your auto industrial businesses grow according to the long-term growth rates you set out, and data center can potentially double, how should we think about the potential for gross margin expansion for 2027? Thank you.
Yeah, Vivek, our gross margin bridge doesn't change, right? The biggest driver in the gross margin expansion here in Q4 is utilization improvement. Now, keep in mind, this is utilization improvement from two quarters ago, right? So you'll start to see that. You also start to see a little bit of the pricing impact that we've had, where we've already had the headwind on the input costs hitting us, so we'll start to see some of that pricing impact come in as well. But the math of every point of utilization being 25 to 30 basis points of gross margin improvement still holds. That will be true as we go into 2027 as well. So as I mentioned in my prepared remarks, we expect further growth, we expect further utilization improvement as we go into 2027, which then will continue to give us margin expansion throughout the year.
Obviously, as we have revenue growth, that will help as well. Then the other piece that really starts to drive is the new products. So we get favorable mix, and we're already seeing that here in Q3, as the new products start to ramp at favorable margins. So we'll start to see that as a nice tailwind going in there too. Then, there's a little bit of the fab-right with one of the divestitures that we've announced on the backend facility. We'll see a little bit of that in 2027. We'll see more of it in 2028 from the mountaintop facility, but that's a nice tailwind as we go forward as well.
Thank you.
Thank you. Our next question comes from Blayne Curtis with Jefferies. Your line is open.
Hey, guys. Thanks for letting me ask a question. I just wanted to follow up on the last one. I am assuming auto would have been up in December, and it is down high single digits. I am just trying to understand the level of restocking that might have already happened in the December quarter, and then I am just curious whether you see those trends increasing.
Yeah, I mean, where the trend is increasing. So there will be, like, think about it as there will be a catch-up because it is restocking. It is not really the end demand. The end demand we are supporting, so there is no change to that. So you can assume some seasonality on top of it. But from a replenishment, the demand is not perishable. So, you will see that resuming in the first quarter as supply comes online. So it is purely a, you can think about it as a quarterly timing. We are not guiding Q1, but once we get to Q1, you can see that some of it would have been in Q4 if you normalize it. I have always said in automotive or even in multiple markets that we service, I do not look at the quarter-on-quarter, and I do not read anything into it.
I look at it from a trend perspective and the decisions we are making. So I have zero concerns about the decisions we are making in automotive. They are the right decision for us to make as a company to service the increased demand that we have delivered in industrial and AI data center in Q3 and looking out in Q4. And then we will resume to normalized supply and demand starting here in 2027. And for the rest of 2027, you will see us deliver the long-term target again.
Thanks. Data center business doubling again in 2027 is pretty impressive. Just wondering if you had any broad strokes you could walk through in terms of the drivers of that.
Yeah, I mean, look, it's a lot of the broad strokes as far as across the power tree, which I've been saying for a few quarters that we've been doing design wins across the power tree from the high voltage all the way to the core. Those have started to come online and generating revenue, so we've already transitioned from design-in to design wins and now revenue. The breadth of customers, we've highlighted some customers in the past. Some customers you can kind of draw a line to who they are, and it's across the power tree, meaning directly with the compute vendors when we're sitting on the board, but also with the supply chain when it comes from PSUs and so on, and we've highlighted those throughout the last few calls. All of these are starting to get maturity.
We ramped up a few more in the third quarter here. We hinted at it at Investor Day as well, so that came in. It's really execution, execution, but more importantly, new products that we have been introducing over the last couple of years are starting now to deliver, and that ramp is what's going to continue in 2027, and top of the new stuff that we're going to ramp into revenue in 2027 that we're doing design wins on now. Just to highlight, the doubling in 2027, again, is off of the higher base that we just talked about today. So, 2027 is higher than what we expected even 90 days ago.
Thanks, Hassane.
Thank you. Our next question comes from Joe Quatrochi with Wells Fargo. Your line is open.
Yeah, thanks for taking the questions. Maybe just to follow up on that, the $500 million that you are increasing for your 2030 revenue target, is that all from the acquisition or is that also from just the base increasing from 90 days ago?
This is tied to the acquisition.
Okay. Thanks. And then as a follow-up, as we think about just the cash outlay that is kind of in front of you in terms of this acquisition and then also Synaptics, and you are raising some debt for that as well, but how do we think about just the use of free cash flow and capital returns in that vein?
Yeah, Joe, so we will be using cash off the balance sheet from the Ferric acquisition. As we close the Synaptics acquisition, we will be focused on delevering. That will be our top priority. That is obviously after investing in the business as well. But we will focus there, and we will be opportunistic on buying back shares over time. But the intent here is to delever quickly. Operator, you there?
Thank you. Our next question comes from Tom O'Malley with Barclays. Your line is open.
Hey, guys. Thanks for taking the questions. My first is on industrial. It looks like it is well above seasonal in September, and then when you go into December, it looks like some very strong results again. Any sort of vertical that you would call out in industrial that is doing a bit better? Those are some impressive numbers in the back half. Just wanted to start there.
Yeah. I think I would call out the AI halo that we talk about, anything from energy. Think about it from the grid to the AI data center. So everything outside the walls of the AI data center has been very strong. It came in better than we expected. A little bit on the PMIs are improving, so the broad industrial, but if you ask me what is the beat and the outlook tied to, is really the AI halo that we highlighted on Investor Day is coming in as we expected, and in the short term, higher, which is the reason I mentioned servicing that ahead of some other markets.
Yeah, and the broader industrial is about 40% of the total industrial bucket, and we are seeing strength there as well. I think just as you are seeing PMIs improve.
Got you. Thad, second one for you is just, you talked about the automotive revenue just being a bit lighter because of supply relocation, timing, phasing of customer deliveries. But it looks like the utilization is flat. Are you short external components? Because I would expect if you are not able to get more supply, the utilization would continue to move up. You are saying that is more next year. Maybe just describe the dynamic, where are you seeing the constraints?
Yeah. Keep in mind that what you are seeing come out of the line today was utilization from two quarters ago when we were running at 77%. We ran at 83 last quarter. We are running at 83 going into Q4. So this is us catching up to that supply as you get cycle time through the fab and the backend. So that is where we have supply coming online. The supply constraints are in specific lines in certain areas, so it is not across the board, or obviously we would be taking utilization up further. But as we look into next year and we see demand continuing to strengthen across all the end markets, we will be taking utilization up.
Thank you. Our next question comes from Gary Mobley with StoneX. Your line is open.
Hi, guys. Garv Saini here on behalf of Gary Mobley at StoneX. Thanks for taking my question. Just wanted to ask specifically about the utilization lever in the gross margin bridge. About half of the path to that 53% gross margin target you guys set out comes from utilization. Just given the restructuring and also just changes to depreciation, how should we sort of think about the incremental gross margin on each dollar of revenue from here on out?
Well, look, the way you should think about the Q3 results is there's about 450 basis points of underutilization charges. Again, you got to go back to the Q1 utilization of 77% for that. If you take that, getting up to what is kind of peaking out for us, 92%, 93%, somewhere in there, you can do that math on that 25 to 30 basis points, and you get to that 450 basis points. To do the walk for everybody, getting from here to the 53%, you've got about 450 basis points of utilization improvement. You've got 225 basis points of fab-right. I talked about the activities we're taking there.
You got another 400 basis points of mix and the new products kicking in, and I've already kind of talked about we're starting to see that hit the P&L, and we'll start to see a lot more of it next year. The last component is we've been raising pricing, but we haven't seen that necessarily fully hit the P&L, so there's a little bit of tailwind from there. If you start to add all that up, you're getting right around that 53% long-term target.
Got it. Okay. Thank you. You guys also talked a bit about just now the new product ramp in 2027. Kind of wanted to double-click on that and the contribution we'll see in the fiscal year for next year. How should we sort of think about the revenue from vertical GaN, EPP, and Treo as these new sort of layers of revenue that really didn't pre-exist before?
Yeah. So, what you are going to start seeing is obviously Treo will continue to ramp. Treo is already contributing to revenue this year. We already talked about vertical GaN contributing revenue in 2027, and then you layer on top of that towards the end, EPP. But those are, other than Treo that started the ramp last year, and obviously we have been more than 2x-ing as you kind of move forward from a smaller base, the other ones will start to ramp and will follow such a trajectory.
So, you can think about the base for 2027 as new products that we have had in 2025 and 2026 that will start contributing meaningfully, while the new products we have introduced, like EPP and vertical GaN, to start in 2027 and continue forward, all leading up to a more mature mix of new products towards that 2030 timeline, depending on auto or industrial.
Got it. Perfect. Thank you.
Thank you. Our next question comes from Chris Caso with Wolfe Research. Your line is open.
Yes, thank you. Good evening. The first question is with regard to the Ferric acquisition. Can you talk to us a little bit about where that acquisition stands with regard to revenue and design wins, and realizing you use that to take up your 2030 target, but as we look into sort of 2027, 2028, what can we expect from that business?
No, look, I'm not going to give a guide to revenue as far as the acquisition or 2027 at this point. I will tell you, during the diligence that we've had, obviously we all know from a market ourselves and all of you know the trajectory that IVR has taken, that it is the solution for higher and higher power of AI processing. That's point number one. We know integration, power density, and the criticality of the Z-height is important, and Ferric has been engaged with all of the hyperscalers from a design and platform architecture perspective. That's from the technology and the capabilities that they bring. We bring the scale, manufacturability, and of course, the breadth of our reach within the hyperscalers, but also the compute platforms. That's the combination that makes this strategic.
From an engagement perspective, they are engaged and in a very good position with all of the hyperscalers, based on our touchpoint and the diligence that we've done with the team. Of course, it brings in a high-caliber, talented team that complements our own internal efforts, and that's the win-win.
Got it. Thank you. As a follow-up, I've got a question on CapEx and capacity as you go into next year. Correct me if I'm wrong on this one, but it sounds like some of the supply constraints you're facing now is just a function of not having started the wafers earlier in the year, and so now you're catching up. I guess if that's the case, where do you stand with regard to capacity as you go into next year? Do you think you have the capacity to be able to meet what you see for the market next year? Then from that, what sort of CapEx should we be expecting as you head into next year?
Yeah, Chris, I think that's a fair characteristic, that this is a catch-up. That's why Hassane talked about this as being temporary. We'll catch up in Q1. So we do have the capacity that we need, based on the outlook that we have today, and even some upside to that outlook. So, I don't think we're going to be anywhere near capping out our capacity as we go into 2027. This year, our CapEx intensity is about 2% of revenue. I expect it to be about 5%, in line with our long-term model next year. That will give us the ability to bring in not only maintenance capital, but a little bit of capacity where we need it on certain lines where we can get higher throughput.
But we believe we are in a really good spot for a market that is going to grow in 2027, and even beyond that, in terms of just the capacity we have online.
Yeah, and the capacity that we talk about that will contribute to catching up in Q1, you can think about this as the capacity is already behind us and the investment is behind us. It will just come online and deliver in, think about it, in 2027. So it is not something that we are looking forward, it is something that we have done, and we are just going to get the benefit, just from the latency perspective.
Got it. Thank you.
Thank you. Our next question comes from Jim Schneider with Goldman Sachs. Your line is open.
Good afternoon. Thanks for taking my question. Hassane, I was wondering if you could maybe characterize the OEM customer behavior in terms of their willingness or interest in restocking. I think you referred in your prepared comments to automotive customers are not supporting restocking, but only supporting increases in end demand. Can you maybe sort of broadly describe what's happening with those customers and then the broader OEM customer set with respect to taking on more inventory on the balance sheet? Thank you.
Yeah. First, I'll characterize it as specific to OEMs. It's not all OEMs are behaving the same way, so I can only reference the ones that we've been engaging with that have demand. They're looking at it as they see the writing on the wall. AI and industrial is recovering, and they're trying to place demand so they're not caught in traffic as far as capacity and capacity constraint, because they see it across the board, not just on power semis that we see, but across analog compute and, of course, the memory. So they're looking at their 2027 plans, and they want to think about it as have a, what I would call, a business continuity plan, or BCP, as known in the industry.
Those are strategic plans that we do with our strategic OEMs where we sit down and figure out kind of their capacity and their pull rate, and we build ahead of that. Normal standard procedure as far as replenishment ahead of a ramp. That's why we're having these conversations as far as we're not going to build it all in Q3 and Q4, which we made that decision. We will continue to ship demand knowing that their backlog is now on the books and we'll service that, and that's what we built the new capacity model towards. That's why it's temporary in nature, it's short-term, and we'll catch up in 2027.
Thank you. Then maybe just as a follow-up, your remaining other subcategory seems to be actually growing very, very well, even when you take away the data center, re-bucket it into its own category. Can you maybe talk about some of the what's driving the other category that remains, and can you kind of extend that growth rate into 2027, or why or why not? Thank you.
Yeah, the other bucket for us, we play opportunistically. In the prepared remarks, we talked about Q3 as being seasonal with the wireless end market that is in that bucket as well. When we look into Q4, we have a large customer ramping a program, so it is a little lumpy. I think that is why Q4 is up to mid to high teens. As we go into next year, no, I would not expect it to grow at that. I think you are going to see the AI data center and the automotive and industrial grow at a much faster rate, and that to be a little bit more in line with normal historical rates there as we focus on the other markets.
By the way, the Q4 ramp that I am talking about is manufactured externally, so you are not seeing an impact of that. That is why we are able to support that.
Very clear. Thank you.
Thank you. As a reminder, to ask a question, please press star one one on your telephone. Again, that is star one one to ask a question. Our next question comes from Tristan Gerra with Baird. Your line is open.
Hi, this is Tyler Bomba for Tristan. Thanks for taking the questions. Is there any inventory replenishment embedded in the strength in industrial? You talked about not allowing replenishment in auto.
Yep.
We are curious what you are seeing in industrial.
No. The strength in industrial is really coming from the AI halo, as I mentioned in a prior answer. The AI halo, we see end product going out in the market. So we can see the sell-through, not just the sell-through from us to customer, but the sell-through from customers to their end customers. If you look at the large industrial names that do kind of grid to data center, they have all had really good quarters for the ones that announced so far. And we see that as well as far as what they are building. So that is why we are allocating when I say it is going to truly end demand, which is stuff getting deployed in on sites. So that is not an inventory replenishment. It is purely an increase in demand that is getting installed as fast as possible in end sites.
Okay, great. For my follow-up, how should we think about the geographic mix of silicon carbide growth in 2027 and the underlying contribution from new platforms and content gains versus just a recovery in the underlying EV production?
I think it is very broad in nature. Of course, from an EV perspective, China remains a strong pull for silicon carbide. But also, a big growth and new products on silicon carbide is coming from North America with AI data centers. And of course, JFET specifically that I called out in my prepared remarks, that primarily is going into AI data center. So I would say, since the EV diversification we have had geographically and end customer, we are adding now a very broad AI-diversified market for silicon carbide, both by application, customer, and geography.
Thank you. Our next question comes from Tore Svanberg with Stifel. Your line is open.
Yes. Thank you. Hassane, I wanted to go back to what you mentioned about the business continuity plans. I mean, we all know that the capacity especially for AI data centers being tight for a while, right? I am just curious what is driving the auto OEMs to willing to restock right now, and it must be a sign that their demand is really starting to accelerate, or do I get that wrong?
No, I mean, I do not know about accelerate, but I think there is demand that they are working on. Remember, for us, it is not just the SAAR, it is the content. As you heard in my prepared remarks, Europe is moving into the 48 volt and zonal. That creates a higher demand opportunity from a content perspective, not necessarily a vehicle units. Mix of EVs is higher, which is more content. You can think about it, demand that I refer to from our side is content. When we talk about resuming to our long-term model, which is high single digit over SAAR, that is the one I am referring to. That is what the OEMs place orders for. They see their content growth in their new platforms that they are introducing, not necessarily a SAAR or unit. That is what I want to be very clear.
Inventory for us is content in the car.
Yeah, that's very helpful. As my follow-up on Ferric, I think there's a lot of companies positioning in IVR right now, and I was just hoping you could share with us a little bit about what's the uniqueness that Ferric brings? I know in your presentation you talked about better magnetics, smaller footprint, and so on and so forth. Yeah, anything else that you could share with us, why their technology is sort of the winning formula versus what we've seen from other companies? Thank you.
Yeah, so if you think about what the solution, what the end solution needs in order for it to be viable from an IVR perspective, there is a thermal, a mechanical, and an electrical. From an electrical perspective, we know Ferric has a superior transient performance than anything else in the market. That covers the electrical. That's important because of the dynamic load that the high-end processors are driving, where you got a lot of transient, and the voltages have to regulate or respond to it. So, that provides a superior electrical performance versus peers or competitors. Then you have the mechanical side of it, which is the power density, where there are other solutions that are not necessarily magnetic.
There are other innovative solutions out there, but the solution from Ferric, with the electro and magnetic components of it, with the controller, that gives you the most dense from an area and the most compact from a Z-height. You can think about everybody's Z-height is in the 2.5 millimeter. Ferric blows that away. That is extremely important when you're talking about truly integrated, or IVR specifically, not outside of the package and so on. That's key. The last one is the thermal. With the efficiency that the first two provide, the thermal become a lesser of a problem to resolve with the GPU or the compute vendors. Those three components, the Venn diagram, if you put all of them together, Ferric is the one in the middle. That's why we like it.
That's why we wanted that asset rather than comparing to everything else.
Great, Khoury. Thank you.
Thank you. This concludes the question and answer session. I would now like to turn it back to Hassane El-Khoury, President and CEO, for closing remarks.
Thank you for your questions. We are focused on the areas where the challenges are getting harder and where our differentiation matters most. Power density is becoming critical across AI data center, automotive, and industrial, and our investments in high voltage technologies, Treo-based solutions, and EPP enable us to solve those challenges at the device, platform, and system level. Combined with the leverage in our operating model, we believe that positions us to drive profitable growth, stronger cash generation, and long-term shareholder value. Thank you for joining us today.
This concludes today's conference call. Thank you for participating. You may now disconnect.