Good morning, everyone. Thanks for joining us on day two of the DB Technology Conference in Dana Point. I am Melissa Weathers. I am one of the lead semis analysts here at DB, and this morning we are grateful to have onsemi join us up on stage. We have Hassane El-Khoury, CEO, and Thad Trent, CFO. Thank you both for being here.
Good morning.
Thank you.
Thanks.
I think the best place to start here, let us talk about the cycle and your outlook on the cycle. On your last earnings call, you talked about some strengthening demand signals after finding the bottom in the first quarter of this year. Can you talk about the signals that you saw in the quarter, any booking visibility, lead time trends that you would call out?
Yeah, I mean, overall, we talked about, you mentioned in the first quarter, we talked about automotive kind of hitting the bottom. We do believe right now we're shipping to natural demand across all of our end markets. Natural demand doesn't mean strong or not strong. It's just at least we're at an equilibrium. What makes us positive about the outlook and where we are in the cycle is really if you look at the metrics that drive the business, we've had our book- to- bill above parity. Already, we've had a much better visibility. If I remember a year ago, we were talking about we're on a good quarter. We have visibility maybe 90 days out. Right now, we have much longer visibility into 2027, sometimes into 2028, depending on the product and the markets.
Sitting here today in 2026, we had better visibility about 2027 than we had while we were in the same spot in 2025 into 2026. Our lead times have been extending in overall, but we do have constraints in some technologies where we see the strength across a lot of the end markets, commonalities about the end markets. All of these are signs of an improving end market demand. Again, not all of it is equal. A lot of it is driven by our AI data center, which we've taken that number, that outlook up from doubling to more than doubling year-on-year. However, the business health across all of the metrics that we look at from an operational side have been moving all in the right direction.
If you had a crystal ball, as we think about kind of the slope of that cyclical recovery, should we be expecting chip shortages, COVID style magnitude of recovery? Or what shape do you think it's going to take?
Yeah, I think, well, that depends on how quickly some end markets wake up to what the realities are. What I mean by that is, we all know AI data center, we have the outlook, the demand is strong, and we're shipping to that demand, which is also creating some constraints in some technologies that are shared across a lot of the markets. We talked about some of our power technology, which is driving a lot of our demand in AI data center. We have some constraints. It's not broad-based constraint, but we do have technologies where we are constrained. We talked in the last quarter that we had to make some allocation trade-offs between auto and AI data center. But overall, to answer your question about the shape of it, we have the capacity. Let me just start with that. We're not worried about capacity.
You're not going to see a big CapEx cycle from onsemi. That's all behind us. We have the capacity. We're comfortable about the growth. We're comfortable about the capacity we've installed. The challenge is going to be how quickly does demand layer into the backlog. If it's going to be a snapback, there's always a cycle time. We do have some inventory in die bank we can launch into back end. That's a two-week cycle time, but that quickly depletes if the snapback is very steep. If the snapback is kind of a gradual increase where you have the AI data center strength and it layers in an outlook of an automotive strength over the last kind of two, three quarters, then we can start building for that, and we're able to support it.
If we get orders today and they need them in, I don't know, October, November, and you get a snapback of orders, it's like, yeah, I'll see you, the lead times are going to extend, allocation is going to start. It's going to be back like the COVID days. It's going to actually be worse than the COVID days, because back in COVID, AI data center was not a big consumer of capacity.
Today, data center is one of the largest consumers of capacity. So the call to action that I've been giving our customers across auto and industrial is put in your backlog, stand behind it, and we will together manage on how to build to it because we do have the capacity.
Melissa, just to put some numbers to what Hassane said, shape of that recovery. We exited last year with our utilization at 68%. Last quarter, we were at 83%. In a matter of six months, we've taken utilization up very quickly because of the demand that we're seeing. What we think is going to happen is, our cycle time, we think about from the time we launch a wafer to an end product is four to six months. I think, as we guided, we said, look, utilization is probably going to go plus or minus from here on. If the recovery gets steeper again, our utilization goes up further. That helps gross margin again. I think just that rate of us improving the utilization shows you the strength of the business, of what we've seen of that order pattern just getting that much stronger, that faster.
Mm-hmm. I'm trying to think back to the past cycle. Where did utilizations peak in the last cycle?
We took some capacity offline. If you think about where we were before, it was kind of in the mid-80s. Fully utilized for us is in that kind of 92%, 93%.
Right.
You can see we're getting there.
Got it. We'll get into the data center side. It's been a little while since I've followed you guys closely, and I'm like, okay, onsemi's a data center company now. Outside of the data center, as we think about industrial automotive, some of the more cyclical businesses, A&D, those kinds of things. Can you walk us through where are the hotspots? Where are you still trying to see some recovery in demand?
Yeah. Even within every one of the end markets. Let me start with automotive. Automotive, we talked about the stability. We have not seen a replenishment cycle, nor have we seen demand uptick. If you look at the SAAR overall, SAAR is kind of actually slightly down to a first order. We expect automotive to grow about 6% or so year-on-year, which if you think about our target and the content growth that we talk about in automotive, we've always talked about high single digits above SAAR.
We're already delivering on that in a stable demand environment. Of course, electrification, which is driven today by China EVs, where we have a majority share, that's outgrowing automotive in general. Again, you look at automotive, you have the disruptors that are more skewed to automotive, whether it's regional or by customer. Those are doing much better than the rest of the automotive market, which is lifting the content for us. Industrial, same thing. We talk about industrial being kind of that, in general, over a longer period of time, being the GDP kind of plus.
However, within industrial, we have secular markets, which is about 60% of our industrial market. You can think about energy storage system, energy infrastructure, ADAS, medical, a lot of these. Our energy infrastructure business is growing 40% year-on-year, and that secular side of the business, I think, posted a 25% sequential. You can see a lot of that strength is being driven by that. Of course, the industrial side, the energy infrastructure, we call it the AI halo.
The strength that we're getting is really a build-out of industrial that is driven by the AI data center build-out. But those are secular, where we can really put a finger on and say, "This is exactly true demand. It's a healthy demand, and it's a demand signal that is out in the end market, and we are supporting it." That is kind of the auto and industrial, and a lot of the growth segments and the growth applications or the sub-markets that we are seeing, but they're all tangible, they're all secular, and they're all end market demand. Net of any replenishment or any uptick in demand, for example, in auto, that's all going to be tailwinds for us.
Just to clarify, on the inventory side, whether it's at distributors or at the end customers, how would you characterize the health of overall industry inventories at this point?
Look, our inventory in the channel is right where we want it to be, right? We want to have 10- 11 weeks. We're in that sweet spot. We've been there for a while. We effectively manage the inventory for our distributors, right? They would take more inventory if we'd ship it to them. We hold it back.
Okay.
I think at the end customer, obviously, that's harder for us to track. I think in automotive, you know Hassane said, we think we're shipping back to natural demand, so we've gotten through the inventory digestion. In some cases, automotive is still dangerously low. They haven't changed their behavior yet.
But we don't think there's a lot of inventory out with the customers. Our inventory on our balance sheet is actually very healthy as well, so I think we're in a good spot and we're lean. And what we've been saying, back to kind of that utilization, is we can match the utilization to the demand recovery, right? Given that we don't have to bleed through inventory at our customers, in the channel, or on our balance sheet before we start ramping up manufacturing.
On the pricing side, can you summarize how you guys are approaching pricing at this point? With some of these end markets getting a little bit tighter, what has been ON's approach, and how do you guys plan to manage through it going forward?
Yeah. So look, we did our first price action in April. We're doing another one now. We've been very clear on that. We have seen our input costs go up significantly because of just the inflation, right? That's already hit the P&L, where I think we haven't really seen the pricing hit the P&L yet. This next second round, it'll start to layer in over the next couple of quarters. So we've done that across the board, across the entire portfolio. We also have supply constraints on certain lanes and certain products, so we're surgically raising pricing there as well.
The pricing environment is a good environment, obviously, right now, but it's primarily us passing cost on that we've already been incurring.
Let's get into the data center side. Very exciting business for you guys. I think you'll admit you were a little bit late to the market, but now it's growing nice and fast. So talk about your entry into this market. What parts are you most excited about? What kind of growth can we expect?
Sure. Overall, in the market, if I would describe the opportunity in the AI data center, we have our IR deck post on our website that shows what we call the power tree, which is grid to core, and all of the opportunities onsemi has across every one of these kind of step conversions to get us to the core. Last year, we talked about as we penetrate that market with new products, we are generating revenue across every single part of that power tree. So of course, people think about onsemi as a high voltage power play. Yes, that's true. That's how we started the AI data center. But today, revenue is coming in pretty nice and even from all the way to high voltage, which is closer to the plug, all the way to what we call the smart power stage, which is right at the XPU.
It is broad, meaning it is end customer diversified, and it is starting to be regionally diversified as other regions build out their AI data center infrastructure. We are very happy with the baseline that we built in 2025. We entered the year and said we were going to double that revenue in 2026. Last quarter, the first two quarters came in ahead of where we expected at the time we guided. We increased our outlook for 2026 to we are going to more than double. We have not given a target because, of course, it is going to be lumpy from a growth perspective, but we do expect for the year as things are starting to ramp for the second half, so we are very excited about that. Forward-looking, the opportunity gets larger for us. Of course, the TAM continues to grow with the buildup.
What I am excited about is on top of that baseline or strong baseline that we have done, you are going to start seeing, whether it is end of 2027, beginning of 2028, or somewhere in that range, that transition to 800 V. The transition to 800 V is actually right up our alley. We have been doing 800 V in automotive for over five years. We are number one in that market. We have the technology. We have differentiated technology no other peer has. As that transition gets there, we are able to deliver compelling high-value solutions to our customers, and we are starting to see that in the design-in and what we sample. One of them you hear me talk about is vertical GaN.
Which is a very different technology than any one of my peers talk about GaN. I say that not because we have a better mousetrap. Vertical GaN as a technology, we are the only ones today in the world that is able to manufacture that because it is a very hard, complex device, physics and material, to manufacture. We have been able to do it. We are sampling in AI data center and in automotive already.
We already have the manufacturing sites here in North America. We are very bullish about the capabilities that we as a company, not just vertical GaN, but the breadth of the portfolio of power can bring to bear as we start seeing that 800 V transition.
I definitely want to touch on the GaN side, but as we think about the go-to market for the 800 V and for these new data center power applications, talk about the go-to market, talk about your engagements, how the design-in process is going, and also the competitive intensity.
Yeah. For the high voltage, or call it whether it is 800 V, because we are shipping 800 V today, it is just not back at the rack side. But in the power tree. If you think about it at a high level, it depends on the go-to market for our customers. Anything that lands on the board is going to be with the design-in primarily with the hyperscaler or the GPU vendor, XPU vendor, because that is very tightly coupled to the XPU itself.
When we are talking about the power from plug to core, that is more you can think about as the Deltas of the world, the Lite-On, the Great Wall, the Flex Power. That is a very different go-to- market where they deliver the power conversion, call it box, that goes in the rack or on the side of the rack and so on. The go-to market is different depending on where.
Because for us, I mentioned we're covering the whole power tree, which means that the go-to market is very different depending on what stage we're in. As that transitions to an 800 V. The 800 V on the back plane of the rack, then that's going to be more of an XPU or a hyperscaler. It's going to shift a little bit as the go-to market changes, but we're able to do, and we have been doing today across the board, and that's part of the diversification strategy we've had where we wanted that kind of the coverage to be much broader.
We've done this in automotive already. A few years ago, you hear me talk about the customer concentration, and the only way out of it is by geographical distribution and customer distribution of revenue. We're in that same playbook for AI data center, and we're very happy with the strength of that business across multiple end customers, multiple end regions. That's going to continue to fuel that growth.
Those engagements have started and are-
Those engagements have started. That is what is driving more than doubling of this year, and that will continue in the next year.
On the GaN side, GaN was another one that I was surprised. I am like, "Oh, ON is a GaN company too now." That was all organic. It was all grown organically within ON's business, or were there acquisitions associated with that?
A couple of things. If you think about GaN as a market, there are two different types of GaN. Without getting too technical, I will just summarize it into you have the vertical GaN and you got the lateral GaN. Everybody in the world that talks about GaN is referring to lateral GaN, which is GaN-on-silicon.
We do that. We are doing that with a partnership. We have announced a partnership with Innoscience and a partnership with GlobalFoundries. We are doing that because the differentiation for lateral GaN is really on the control and the driver, which is what we are doing internally with Treo.
We are partnering with very strong partners externally to get access to the GaN-on-silicon devices. Again, a very competitive approach that is kind of the best of both. We are very strong in the control and so on because we have invested organically in Treo. We combine it with competitive GaN devices, and we go to market that way. We have actually announced the complete go-to market at PCIM a few months ago from a power conference. So that is on the lateral GaN. We did that with partnership and organically internally. The vertical GaN, that development has been all organically.
We have done a small tuck-in acquisition that came with some of the IP. So we are very happy with the IP moat that we have and the patent moat that we have. But the whole world, when we talk about vertical GaN, it depends on if you talk about people that understand the power of vertical GaN, they say this is the holy grail for power. For people that do not have vertical GaN, they will tell you it will never work. That is kind of how it works. I am here to tell you we have the fab. It is up and running. We are sampling, so we have physical samples at customers on their boards being tested.
The proof of life is non-debatable anymore. We are very happy with the results that we are getting, and it is a very competitive approach to power delivery. Now, the next question people ask is, "Well, what does that mean for silicon carbide?
It is actually complementary technology to silicon carbide. What I mean by that is if you can solve it with silicon carbide, we are going to solve it with silicon carbide. We have a very competitive silicon carbide JFET portfolio and MOSFET portfolio that is growing very nicely in AI data center. But there are segments in the market where size is actually a constraint, or density is a constraint, and we need a vertical GaN to be able to resolve that. So it is incremental. It gives us a much broader portfolio than any of our peers have today, and it puts us in a very innovative solution where we can engage at a much higher level with customers on forward-looking designs.
Where is that GaN fab?
The fab is actually in Syracuse, New York i n [North America].
Got it. Then the competitive intensity, are there other vertical GaN players who you're-
Nobody has been able to deliver a vertical GaN device to date.
Okay.
By the time they figure it out, they have to work around our patents.
Perfect.
That's always the good first to market.
Also on the data center side, but outside the walls of the data center, I think that the power infrastructure outside those walls, it's harder for investors to kind of picture what the opportunity looks like there. Can you talk about what's the TAM on the energy side? How fast is that growing? Is it bigger or smaller than within the data center?
Yeah. We will be talking a lot about that in a lot more detail coming up at our Analyst Day. If you haven't seen the announcement, we are having our Analyst Day on September 16th.
I will be there.
Which is when we are going to be talking about a lot of the opportunity and the content both across our end markets. But specifically to your question, that is what I refer to as the AI halo. In industrial, that business, the energy infrastructure, has been strong historically for onsemi. We are number one in energy storage systems.
We're extending that into solid state transformers. The growth is coming back specifically in 2026. That's the 40% growth I talked about, pulled by the AI halo. That, for us, the opportunity is incremental to what we've been doing in that market. That energy infrastructure or the secular side of industrial is about 60% of our total industrial. It's been growing nicely over the last few years to become a major part of our industrial segment, and that will continue to grow. If you think about solid state transformers to support that 800V transition, historically, transformers have had no electronics content. When we go to solid state, that's all incremental for us.
Energy storage, that's a growth business. That market is already there. It started, I would say, four or five years ago, and it's going to accelerate from an end demand perspective driven by AI data center. Very different on the industrial. Nevertheless, it is all high voltage play, which is where a lot of our strength already lies. A lot of things I talked about from GaN and silicon carbide and even IGBT, all of that lands into that AI halo or the industrial secular side of it that we're servicing today.
Maybe the next leg of AI people are saying is physical AI. Can you talk about what are your offerings today in physical AI? When do you see this market inflecting in a big way? Maybe we can tie that into Synaptics as well.
Sure. If you think about physical AI, today, we all consume. The opportunity for AI is in the data center or adjacencies or AI halo, but it's driven by the AI data center, and we all in this room consume AI, if that's a word, I guess, consuming AI. We consume it through the computer or through tool sets, so it's a digital consumption of the AI. The inflection point for physical AI, which is purely the AI capabilities going out of the walls of the AI data center, that's the physical AI, and that's going to be with an articulating application. What I mean by articulating is application that need power, they need sensing to sense the world, they need compute or capabilities of compute or connected compute, and the control side. Those are the four pillars of what physical AI.
If you think about what does that describe today, the market is with robotics. I am not talking necessarily humanoids. That is a forward-looking robotics segment. But today, we are very strong in robotics. It started with the factory automation.
The cobots and so on. It went to AMR, which is factory floor that required increase the level of sensing because now it is kind of going around humans, so the safety requires a lot more sensing and control. Today, onsemi's opportunity is the power, sense, and control. We deliver ultrasonic sensing, we deliver inductive sensing, we deliver image sensing, and we deliver the whole power domain capabilities in order to power those robots. Forward-looking, of course, that is going to extend into the humanoids, and we are starting to see that. I was, this morning, watching the World Humanoid Robot Games, which was a lot of fun. But you are starting to see that becoming now. Of course, the market is not going to be with robots for Olympics.
This is just the capability showcase of what robotics usability of humanoids are going to be, whether it is a factory automation upgrade, AMR upgrade, that is going to deliver more content for us just because you have more motors, more of everything in a humanoid. So that is an opportunity moving forward. But that is today we are a power, sense, and control play in these, and we are very happy with the access that we have to those markets, and really, already the growth that we are seeing in these markets.
Do you have a sense of timing, like when we could see that business inflect for you guys?
Look, I don't see an inflection, meaning it's going to be this way, because robotics is already growing. Actually, the largest segment, the largest TAM today in robotics is really the AMR and the factory automation.
Yeah.
Humanoids is going to grow, but not to a level to overcome the rest of it. That's why you don't see an inflection. When you're layering a fast-growing business, but a smaller scale business, you're not going to see an inflection at a high level. But you are going to see the inflection within those sub-markets.
Humanoids will have an inflection, depending on the headlines and what you believe. From a volume perspective, probably one or two years out. Because we're seeing all the design-ins, we're seeing the architectural changes that we're making today. The beauty of it is, all of the products we are delivering in automotive, we are servicing the robotics business with.
You heard me talk about ultrasonic sensing. We're number one in that in automotive. Inductive sensing, we're number one in automotive. Most humanoid companies or robotics companies today are car companies, whether it's China or North America, with a few startups that are disrupting that business. We're very happy with our penetration and exposure in that market. We're going to grow as that market starts to inflect.
I want to get to the automotive side, but as we think 10 years from now, what's the bigger opportunity for ON? Is it data center, AI, or physical AI?
10 years from now is a very long time.
Hopefully from now.
When we are talking about the pace of change, I would love to know kind of two to three years. But overall, I think AI capability. What is AI going to look like 10 years from now, or even five years from now, is the exciting part that we have. How do we address that? If I do not have an answer, which I do not, how do we address it as a company? We address it with capabilities. I just gave the example of a lot of the capabilities we have had in automotive have translated to industrial, and when AI data center came out, it is all the same capabilities we have had in auto.
A lot of the AI halo is capabilities we have had in industrial. Our focus is really on capabilities and the pace of delivering those capabilities. My view is for onsemi perspective, whatever that pivot is, whatever the next inflection, whenever it happens, we have the capabilities to be able to service it. You mentioned, which you are right, on the AI data center, we were later to the party than most, but look where we are today.
We did not have to reinvent the wheel. We just had to shift some of our capabilities to that market. That is the moat we have been building over the last five years is building that company that is a versatile technology company that is able to service these segments. Because what we want is the inflection does not have to be big for us to be able to go after it. A lot of companies go, "Well, it is a small SAM or TAM, we are not going to go after it." We need to be able to go after all of these, and that is by building a capability baseline of technology that you are able to do all of that.
Maybe a quick comment on Synaptics with that?
Yeah, if you think about everything I said about our play in robotics and human or the physical AI realm, the fourth pillar that I didn't mention we're doing organically is that connected compute side, where at the center of every one of these systems is a connected compute, where we do the other three around it. We do the sensing, that drives the compute. We do the power, that powers the compute, and we do the control, that is the output of the compute. So Synaptics brings a strategic capability at the center of the physical AI realm. But it's not forward-looking. It's in the robots that we're talking about today and where majority of the market is today. They're at the center of it, and that brings that nice cross-selling capability that as a combined company, we were able to support our customers with.
Now, why Synaptics is very important, although I talk about compute, people say, "Yeah, but there's a lot of companies that have MCUs," which is very true. But talking about inflection, MCUs are where the market has been.
If you talk about where the market is going and the capabilities of compute that are needed in the systems of tomorrow, back to working on the inflection now as a capability to be able to leverage the inflection. That is an AI-first compute. You need to have an AI engine, not a microcontroller pretending to be an AI because you put a label on it. If you read between the lines or the writing on the wall, most microcontroller vendors are trying to figure that out because they are coming from a legacy and everybody is talking about it, everybody needs it. Synaptics brings that. They have done very good strategic choices and that I am very happy about. I am happy about the strategic combination of the two companies. We expect that deal to close in mid of 2027. More to come on that.
From a technology perspective, it does bring the fourth pillar. Of course, as we started working together back at the diligence side, I was very happy to see that a lot of their other business, the non-Astra or the non-connected compute business, is already a strong business in an inflection point, which is a tactile sensing that people do not appreciate yet. Every robot we talked about, every sensor we talk about needs tactile sensing to grab things. They do that. They have very compelling technology and capability. That was a pleasant surprise, I would say.
One more on the top line, and then I promise, Thad, you will get involved in the conversation too. On the automotive side, you touched on this a little bit earlier about SAAR being kind of flat to slightly down, but it is still a big part of your business and you guys have a lot of opportunities in automotive. I guess rank order, where are you most excited about in auto? What kind of growth do you think you can deliver?
Yeah. I think for automotive growth, you are going to see us continue to deliver on that content drivers for us, that high single-digit growth above SAAR. We have been able to deliver that. If you look at the last five years, from end of 2019 till, I think, [beginning of 2026], our growth was 72%. If you take that and you turn it into a CAGR, it is a 9.5% CAGR over a five-year period in an area where SAAR was not that exciting either. So we have been able to deliver that content. What is driving that content today and what will continue to drive that content, of course, the top is the electrification.
I call it electrification because it is not just battery electric vehicles. You have trends in there, which is plug-in hybrids. Plug-in hybrids are becoming a higher volume of the electrification, but the opportunity for us, four years ago, those were IGBT driven or silicon power driven. As OEMs push a higher range of plug-in hybrids to 100-mile range, we have already have two design in China and North America for a plug-in hybrid with silicon carbide. So one is electrification and the economies that it drives around and back to the energy infrastructure. The second one, which I am very excited about because we started ramping production now, is a software-defined vehicle or the architectural that allows that with zonal. Part of our Treo platform, we have introduced our 10BASE-T1S Ethernet connection.
Our smart power, which is at every zone as well. Those are all ramping starting this year. That will continue to provide that growth that I talk about, which is the content growth. So we are participating in those inflection points in automotive on top of an already strong base of electrification. So we will continue to deliver those. Of course, we have our sensing franchise that will continue to deliver, and I can go down the list of incrementally more and more content that we are introducing that will continue to deliver that high single-digit growth above SAAR.
Sorry, Thad. One more. On the Treo side, can you walk us through what is the reception initially to Treo? What was the motivation to start something like Treo? And how can that change how you are engaging with customers?
Yeah. So back on the capabilities. So Treo in itself is a technology and a design platform. It is not really the product. It is a BCD, it is a 65 nm mixed signal analog platform with a very high voltage range. I say that because there is no other technology or no other peer today that is able to deliver a 65 nm up to 90 V. Why that is important, because when you have 90 V, now you are talking about high voltage applications that we can do monolithic. We are able to do analog mixed signal at high voltage, which is typically today the high voltage is outside.
People focus on the analog kind of as a device. We are able to combine the two. So that is the capability. From that capability, we are delivering all of these products. Ethernet. onsemi historically has not been in automotive Ethernet connectivity. We are already shipping that first to market to a lot of the OEMs that we have.
Was that organic?
That's all organic.
That's all based on the Treo capabilities that we have. Back to we invest in capabilities so we can deliver a lot of these products. That's point number one. Point number two is it delivers a pace that we are able to deliver. We talked about six to nine months to deliver a product on Treo. That used to be two years for a mixed signal analog.
The pace of capability back to capturing inflection points, even that we don't know are going to happen, you need to make sure your technology is solid and your pace is fast. That's what Treo is delivering for us. On top of that, of course, we have all the drivers, we have all the power, we have all the controllers. All of that is based on Treo. To answer your question on adoption, the fact that we created Treo, then we delivered something like 10BASE-T1S, and we are already in automotive production, tells you the compelling value that customers see to pull on the adoption forward. We're already ramping in medical devices, in the industrial. We're already ramping in automotive, and we have the AI data center as well.
Okay.
If you think about it, what does that provide? It provides a franchise that is high growth in all of our end markets, and more importantly, a margin expansion. The margin profile on Treo-based products is in that 60%-70% gross margin. We've talked about 2030 being a $1 billion revenue. You can start seeing part of the margin bridge that Thad, I'm sure, will talk about, is that mix of new products. Part of it is Treo delivering 60%- 70% as a larger percent of mix towards the 2030.
Thad, you are up. On that gross margin bridge, you have your target, I think 50-ish percent. You are now in the 40s. Maybe give us your bridge of how we get from current levels back up to 50%.
Yeah, let us bring it home. Let us talk about all the cool differentiated products that Hassane has talked about and where it takes us financially. Just to calibrate you, in Q2, we had a step function in our margin improvement, and in Q3 with our guide, another step function. All the work that we have been doing in our manufacturing footprint, in our portfolio, rationalizing our portfolio, exiting low-margin business, Fab Right, taking cost out, the mix shift that we are seeing, we are now starting to see that leverage come through. Even in Q2, you saw that our earnings grew at a much faster rate than revenue. Through the downturn, we have been investing. Now, as you think about as we go forward, the biggest mover in the short term is utilization. I was touching on that earlier.
In Q2, there is 650 basis points of headwind just from under-absorption of the utilization. You can think about every point of utilization is 25 basis points- 30 basis points of gross margin improvement. You can do the math. We are at 83%, do the math, getting up to 92, there is your 650 basis points. That is one of the key drivers. We have got more Fab Right initiatives that we are driving towards. That is about another 200 basis points. Last quarter, we announced the divestiture of two additional manufacturing locations, so that is 50 basis points of that 200 basis points that we will start to see some of that in 2027 and most of that out in 2028. The products that Hassane talked about, Treo, the others as they start to ramp, we get favorable mix. Think about that as being another 200 basis points.
The last piece of it is the fab divestitures we did a few years ago.
Yes.
We built bridge inventory to support those divestitures that we are burning through. As we start manufacturing that inside, that is about another 200 basis points. That gets you with a five handle, well above a five handle, and our target is a milestone, not a destination. Everything we have been investing in through this downturn, it is in differentiated products. We have a core that is very solid that allows us to build on that, and as we project forward, we have this nice trajectory going forward. Our OpEx, we have been investing through that downturn, so we are not going to grow OpEx anywhere near what the revenue growth will look like for the next several years. There is a ton of fall-through from margin, leveraging the OpEx investments, and that all falls through.
The other thing I would tell you is through a downturn last year, our free cash flow margin was 24%. That is another kind of nice launching point that we will continue to improve on as the top line goes up.
Any sneak preview of the Analyst Day that you want to give us?
You must be present to win.
On that bridge inventory, how long does that take to burn down? Is that a matter of qualifications, or is that just a matter of-
It is a matter of time. We peaked that inventory mid-2025, so we built that inventory. We are now bleeding through it. It has taken us longer to bleed through it than we anticipated, just because the market got softer, right? But, look, last quarter, I think we were down 8 days- 67 days, something like that. You can see we are burning through it, and then that is when we will start manufacturing that inside of our footprint as well. All is planned. It will take us a couple more years to bleed through it all. The good news is we paid for that, so that is a tailwind of free cash flow again.
Mm-hmm. East Fishkill, where are utilization levels at East Fishkill at this point?
It is below the corporate average. It is improving. That is our 300 mm fab. Treo goes in there. The big investment in East Fishkill is behind us. All the capacity is there to support the growth that Hassane talked about on these new and emerging products. It is just a matter of time as we fill that up, and as these new products ramp, as the revenue grows, we will continue to do that. It is not necessarily a qualification, it is just the new products ramping.
On the OpEx side, you kind of touched on this already, but with all these new projects, with getting into the data center, with trying to address these things very dynamically, how do we think about the trajectory of spending going forward?
You should think about this as where we are currently is our new baseline.
Okay.
Going forward, think about that growing at a much slower rate than revenue. But we are above our model today. We have got to grow back into our model. So that is why I am saying that the big investments are there. You should think about this going forward. Think about it as kind of low single-digit percentage increases for the next several years. Not a big step function.
Business exits, that was a big priority for you guys when you took over as a management team. I think you are very, very close to completing the $900 million.
We are done. We are done.
All done?
Yeah.
Okay.
We are done. $900 million over the course of several years, $900 million annualized. Done. We said that we were completed at the end of Q2, so that is all behind us. You can think about Q2 is now kind of the new base from that, has all the exits gone. That was business that was highly volatile. It took us longer to exit than we thought because the way we were exiting was we raised pricing. Customers stuck around a lot longer. We had the capacity, so we kept it. As the market downturn lasted longer, competitors started coming in, cutting price, and we just let it go.
As you look at your portfolio today, are there any soft spots, areas that maybe you would want to optimize further or?
Product-wise, no. In the manufacturing footprint, yes, we're still optimizing there. That's that other 200 basis points. But in terms of the portfolio, no, we got the right portfolio at this point.
Great. Have you sized recently where the consumer electronics exposure is? I'm asking because memory pricing has obviously gotten a lot higher than where people were thinking, so I'm just trying to assess the risk.
It's a very small piece of our business. It's in the other other, right? Because we put the AI data center other.
Right.
It's in the other other, which is a very small portion of our business, and so small that we're really not seeing any impact from the memory shortage.
A lot of it is more of the enterprise than the consumer electronics. A lot of it is enterprise, which we are not seeing the softer because they are getting the supplies that they need.
On the CapEx side, your existing footprint, are you happy with where you are? How do we think about CapEx? Is there any maintenance CapEx we need to look out for?
Look, we are in maintenance mode at this point. Like I said, we invested heavily in capacity, right? For silicon carbide, for EFK. We brought in all the capacity to support the Treo ramp. We have been running in that kind of mid-single-digit percentage CapEx intensity. I do not see that changing for several years at this point, right?
We can expand that footprint within our manufacturing, the footprint that we have today, brownfield investments without having to do any significant CapEx investments. Our revenue has to get to about 30% higher than where it is today for us even to think about adding capacity.
Okay.
We have the ability to flex inside and outside as well. About 35% of our production is managed outside. As you come out of this, the first thing you always do is load up your internal fabs, and then you start flexing to the outside. We have plenty of capacity, and just don't look at having to bring in any meaningful. We're not breaking ground on a new fab anytime soon, just put it that way.
Okay, great. Well, thank you both so much for joining us, and everybody have a great rest of your conference.
Thank you.