All right. Good morning. Welcome to TD Cowen's Technology, Media & Telecom Conference. I'm Joshua Buchalter, semiconductor analyst here. Very pleased to be joined by Jeff Palmer, EVP of IR from NXP Semis.
Sure. Hi, Josh.
Wow.
Thanks for having us.
Thanks for being here. I guess to start, we were just talking off to the side, there's a lot of positivity in the mature node semiconductor space, and there's, I think, some element of an AI and data center halo involved there, but also some legitimately improving cyclical elements.
Maybe you could just spend a couple of minutes talking through overall the backdrop, what you guys are seeing coming out of your Q1 earnings call.
Yeah. A great place to start. I'd say compared to 90, 180 days ago, I think we're more optimistic than we have been in a while. I think, with automotive being 58% of the company, when automotive starts feeling good, in general, we feel good, even though other parts of the company are doing well. We track a number of KPIs that give us an insight on how the business is trends are acting. Those KPIs are things like book-to-bill, solidly above one, end customer backlog through distribution, very nice and building. Late orders coming in and expedites going up, lead times stretching out. We have had to raise some prices in Q1. We're feeling some inflationary input costing, and our model is to gross up any inflationary input costs and pass to customers if we cannot digest it before doing so.
We've also looked like we're going to have to raise prices into the second half. This is primarily in the area of energy, transportation, precious metals, substrates, not on the wafer side as much.
Okay. I guess maybe on that topic, anything you can give us on the scope of the pricing increases, how they're being received by customers? Is it generally everyone went through this kind of already a few years ago, and there's empathy and understanding, or pushback?
Customers never like price increases. I think we've tried to be. We're consistent with our approach as we have been in the past. We're not trying to pad our margins. All we're looking to do is maintain our gross margins. As I said, our first course of action is to determine can we digest and operationally take care of any inflationary input costs. When we cannot, we do have to gross those up and pass them to customers. They don't like it. I can understand that, but we try to be transparent and fair about what we're doing.
Okay. I'm going to try to get some of the cycle stuff out of the way before we get to.
Sure
the more fun product discussion. I think Bill guided internal utilization rates to run in the low 80% range in the first half.
Yep
The mid-80s in the second half. Can you walk through maybe what are the key drivers of that and any major differences on your internal versus external loadings?
As you know, internal, external, so externally, we produce about 60% of our wafers, internal 40%. Utilization is really only in effect on the internal portion, right? Because we have no effect of utilization externally.
Remember, we have four main eight-inch factories, which are not fungible. They run different processes and products. Two in the U.S., one in the Netherlands, and a joint venture in Singapore with TSMC. While we are building some bridge stock on one of the factories, as we're planning to decommission it, the others are really doing better on fundamental and demand.
Okay. Is there any way to quantify how much of that utilization rate increase is from the bridge inventory versus the sell-through?
No, not really.
Okay. Last one on this topic. Correct me if I'm wrong, I believe auto is over-indexed to internal manufacturing. Is that right? Should we think about better auto demand equals higher share of internal utilization rates increasing?
That's not actually correct, Josh.
Okay.
A number of years ago, what we did was we moved all of our bulk CMOS products out of our internal factories. Our internal factories today are all primarily proprietary mixed signal factories. Anything that's on CMOS, including automotive, is done in the foundry now.
Okay.
No.
Okay.
The way to really think about it is if you hear us speak about any of our analog products or our mixed signal products, or RF products, those tend to be built internally.
Got it. Okay. I think there's been a lot of debate in the investment community on the China auto backdrop. People saw weak domestic consumption in the Q1 and were concerned, but exports have obviously been very strong. Given your vantage, could you maybe speak to us about what you guys are seeing in the China auto market and from both a unit and also content standpoint?
Yeah. I read that question last night and thought about it for a while. I think, Joshua, we have to take a step back. If you think about the auto industry, it's been running high 80s, 90 million cars a year for many, many years. Essentially flattish, right? Our auto business has been up nine percent CAGR over the last three years and up 13% in the last five years, which really reflects and underpins it's a content-per-vehicle story, and that's really what we're focused on. As China's become bigger, it's about a third of the total global production right now, it has its own unique seasonal trends. They tend to push really hard into the Q4 , take a little bit of pause into the Q1 , and then resume. That was contemplated in our guidance for Q1.
That's why we were not surprised when we saw a little bit of weakness in China. As you said, export is really what is driving that industry right now. I'd say we feel overall, compared to last year, very positive about the auto industry. I'd say auto in general will be up this year. I'd say China auto will be up this year.
When you say you're positive on auto, it sounds like certainly China. Do you feel similarly about North America and Europe?
Yes. Holistically, the business has improved. As you remember, last year, one of the headwinds we faced was in the Western markets, North America and Europe. We had quite a few of the tier ones who were over-inventoried. We thought that over-inventory situation would digest in 90 days, maybe 180 tops. It went on Q9 . It was very brutal. Right? That's finally behind us. Most of them are buying what I'd say two end demands. Some of them are still highly under-inventoried. They've run very lean. We're not expecting any restocking to occur. We think we're just going to have to accept that many of the tier ones are going to run lower working capital metrics than they have in the past.
Yeah. Is that something that surprised you? After the shortages in 2021 and 2022, maybe I was a bit naive, but I assumed that the tier ones wouldn't necessarily put themselves in that position again to be under-inventoried. Can you maybe speak to what's driving that? Also, what could potentially break that and force them to change the trends of carrying too little inventory?
First, we should be clear. While the tier one situation is in a Western market, it's North American, Europe.
That's about 60% of automotive. It's not a one-size-fits-all. We have some very large tier ones who are sitting right at the number of weeks of inventory that we would expect and giving us forecasts as we would expect, everything's fine. We have a few other large tier ones who are running three and four weeks, and the view from their perspective is they just have very thin margins. They just aren't getting any compensation from the OEMs, and they're going to run at just very low metrics.
going into the cycle. What could change that? Well, as you know, we put a letter out to customers last year, said, "Look, our cycle times are three to six months fab to finished product. Get at least your forecast to us so we can get you in the queue." If a customer decides they're just not going to do that, they are probably going to go lying down at some point.
Got you.
I know you had another question there, but while we have some shortages on substrates and precious metals are high, and labor and transportation are high, wafers, the actual wafers themselves, we have a fairly well-defined envelope of what we can expect this year. If we have a customer who comes into us and says, "Oh, I didn't forecast," and they take us outside of that envelope, we will be charged for those additional wafers, and we will have to pass that on to those customers.
I guess on that topic, do you think there's enough capacity out there in the short term to handle an increase in expedite orders? Maybe, perhaps more importantly, I think there's increasing concern that there has not been enough investment in mature node semis capacity. You guys are sort of doing that with your partners in some of these JVs. How do you feel about structurally the industry's ability to support nine percent auto CAGR and then also seemingly some data center demand as well?
I think companies like our peers probably have to plan out a roadmap for wafer supply quite farther out in time. We started working on VSMC, our joint venture in Singapore, three, four years ago, and it's not even up and fully running yet. It takes time. Our partners, TSMC and GlobalFoundries, are great partners. They try to accommodate us whenever we can. When you say short term, you have to remember there's a certain amount of physics that go into this. If you don't forecast and you come to us and we do not have the product, you're probably looking at a full three - six month cycle time.
There's nothing we can do about that. That doesn't even take into the fact that there's demand from other customers on that capacity.
Got you. Sorry, one more on the auto cyclicality piece. One of your peers said on their earnings call recently that in particular in China, things strengthened at the end of the quarter. I know you can't speak to what happened after the print, but linearly, can you maybe speak to what happened in 1Q with China Auto? Was February super weak?
Not that granular, Josh.
Yeah.
What I will say is, contrary to popular, China auto for NXP in Q1 was actually up. It wasn't up great. There were other regions up stronger, and we've guided auto to be up again here in Q2, of which China is a participant of that.
Okay.
That's probably the closest I'm going to get to that topic.
Okay. Understood. It's a good segue, though. I think you and your peers, in particular this earnings call, but for years have been leaning into, "You shouldn't think about us as tied to SAAR. We're much more a content store.
Yep.
I think with inventory now normalized, we hopefully should be able to see the evidence of that in your and your peers' models.
Could you maybe walk through what are the key vectors for content growth that you see and you're most excited about at NXP?
Yeah. The way we've defined it, Josh, as I think you know this, we've looked at our auto business and we've said there's a certain portion of it, which we call our core business. These are franchises where we have a high market share, and it's unlikely that we're going to outgrow the market.
That core we think will grow at about low single digits. There's a layer on top of that, which we call our accelerated growth drivers. There's four very well-defined accelerated growth drivers. The largest one is being the effort towards software-defined vehicles. Which is really around our S32 MPU family, our K1 series zonal processors, auto Ethernet, and software. That business was $1 billion in 2024. It was over $1 billion in 2025 and is expected to be about $2 billion by the end of 2027. In that horizon, Josh, that is not design wins we go get. We have the design wins. We have to just wait for customers to get ready to go to production. That's a good chunk of that. Our radar business, 77 GHz radar, that's probably going to grow at 15%-20%.
It was about a little under $900 million in 2024. Had a little bit of a soft year last year. Grew, but it's a little softer as we were going through a product transition towards imaging radar, but that's, I think, really a good franchise for us. Electric vehicles, primarily focus there is battery management systems and gate drivers. Last year, we did have a program that kind of went sideways on us. Coming back this year, we feel good about that. That's probably, it was about $500 million in 2024. Should grow at about a 15%-20% CAGR. Lastly is connectivity, and by connectivity, it's really two types. There's in-cabin connectivity, Wi-Fi, Bluetooth for the passengers and things like that. There's ultra-wideband connectivity, which goes into your phone as well as into the car.
Okay, a lot there.
I know.
Let's double-click on-
I know a lot there, but I wanted to give you the laundry list.
No, it's very helpful. Let's start with S32.
Right.
Your guys' approach in investments in your S32 and five-nanometer MPU family is a bit different than a lot of the other legacy auto microcontroller players. Can you walk through why you made those investments in more advanced process geometry, microcontrollers, SOCs, and microprocessors, what you're seeing in the trend of the software-defined vehicle, and why that's important?
Sure. About seven years ago, we made the decision. We said, "Where's the puck going to go, and where is it today?" If we wanted to just continue along with our peers, we would've invested in another auto microcontroller family. We made the decision based on conversations with OEMs that the current flat point-to-point architecture in a car had kind of reached its useful life, and that OEMs were starting to contemplate how they could move more towards a hierarchical switching fabric in a car. We made the decision at that time to invest in a whole family of MPUs. These are high-performance, primarily 16 and five-nanometer products on MPUs. This ranges from the S32N series, which is not even in production yet, the S32G, which is a gateway product, S32R, which is a radar product, and other product families.
The idea here is to create a hierarchy of processing, and different companies are approaching it different ways. Some companies are going more of a zonal approach, where you'll have powerful microcontrollers in four or five different zones around the car, and maybe a gateway as a kind of a traffic cop type of a thing, whereas others are going full-on central compute with lighter IO aggregator zones around the edge. Our view is not to dictate to our customers what architecture is better or worse. We try to have a portfolio of products of both processors, microcontrollers, ethernet products, and software enablement to allow them to develop the car they want to do.
Okay. Maybe can you help us, where are we in that evolution? I think we've been talking about software-defined vehicle for a while, but it feels like this is the year when it's supposedly hitting the knee in the curve. What's taken so long, and what, I guess, are the benefits that can be unlocked by the software-defined vehicle from your customer standpoint?
I think what's taken so long from your perspective is the software lifting that had to take place, especially for those customers who are thinking about going towards a central compute model. It is just such a different way to build a car. If you think back to how cars had been built just a few short years ago, the awards were given to different tier ones who would develop different parts of the car, different ECUs independently, and then put the car together and deliver it to the OEM and say, "Here you go. Figure out how to write software for these islands of different computing architectures." The idea of software-defined vehicles is to have a holistic processing architecture so they can have software that can be pushed down to any layer of the vehicle. You park your car at night, you get an update overnight.
Over the time you're driving the car, the car tends to learn your behaviors, maybe update software to your driving behaviors. I'd say that when you start to see Western car makers rolling out SDV cars, probably the 2028 model year, go on sale late 2027.
That's our current plan. Our current business has been primarily driven by a number of the Chinese and Korean OEMs.
Okay. Maybe can you talk about the five-nanometer MPU? I believe a couple of years ago you talked about having one or two lead customers.
Yeah.
Any updates there you're able to share?
No
how material that could be?
What I can say is that probably the last time we spoke, it was probably a more fragmented market globally. I would say today, fast-forward, every auto OEM in the world has an SDV program. They're either running a program today actively, and we're engaging with them trying to win business. They're architecting it. We're trying to influence that architecture or they're thinking about how they're going to roll one out. This will be the secular change in how cars will be built in the future.
Okay. Maybe two more on this topic, then I'll finally leave it alone. You guys also recently made a few acquisitions, I think, to bolster your portfolio, specifically on the software side and middleware for software-defined vehicles.
Right.
Can you maybe walk through the rationale of those deals and how they differentiate you? Then I'll just squeeze both questions in at once. You described this hierarchical architecture in the vehicle. How synergistic is it? For instance, if you own the central compute, does that mean you will get higher share in the ECUs, or is that not really how it works?
Right. Okay, let's start with the software. We acquired two kind of what I'll call auto-centric companies. One was a company called TTTech Auto out of Austria. Their product was something called MotionWise, which was a middleware operating system for auto OEMs. Their expertise was in the area of functional safety and security. We looked at them, and it was really a make versus buy. We knew we were going to need more software engineers going forward than we currently had. By acquiring them and we engaged with them, it was 1,200 very, very well-versed, very well industry-respected auto software security people, and was really make versus buy.
The idea here is to not only have them continue to focus on MotionWise, which is the product, but to also help us enable the software on top of the S32 families. That's going well.
It was really a make versus buy decision. The second acquisition was something called Aviva Links. What Aviva Links offers is a multi-gigabit asymmetric SerDes technology. A lot of geeky words there. The idea is you have certain applications in a car where the data from the end node, the sensor, is higher upstream than the amount of data you have to push down to the center, so asymmetric. It's perfect for things like ADAS applications where you have multiple cameras or multiple radars or even lidar in certain situations. It also works very well for things like in-cabin, where you may have multiple screens and you're only looking to push data one way to those screens. Early acquisition. The nice thing is a couple of our customers kind of nudged us towards it, so we do already have design wins.
We probably won't see revenue till next year at the earliest.
Okay.
Your second question was?
Synergies between-
Yeah. If you win the central computers.
Yes.
The thing to understand about the S32N product, it is very interesting. It is effectively a virtual ECU. It has 16 independent ECUs built into the same die, with independent resources, switching capability. What you are really doing is you are aggregating all the ECUs from around a car into a single device, and you will probably have more than one for redundancy. What you tend to see in those applications is customers then going to a lighter, more of a zonal aggregator around the edge, grabbing the data from different sensors, packaging it up, putting it up to the central compute. The interesting thing about the S32N is it can dynamically change its function depending on how the software is. It is not fixed like ECU one through N only does one type of function. It can actually be dynamically programmed by the car company.
Does it give us a lead on pulling other data, other sockets along? Sure, of course. It would be naive of me to say that it's a slam dunk. I will say SDV is probably more like the opportunities are bigger, the opportunities to win are bigger, but if you don't win them, if they decide to go a different direction, it can be big as well.
Okay
to the negative.
I'll stop on auto, finally.
It's all right.
You guys, I think for the first time, called out your data center business on your most recent earnings call. $200 million, expected to go to $500 million this year. Can you walk through the key components of that? There's a few different buckets that are contributing to that revenue.
Yeah. It's really two kind of functional areas. I should be real clear up front here. You are probably smarter on data center than I am, Josh.
I don't know about that.
We're a bit new on this topic. Fundamentally, we only focus on the control plane. We're not in the data plane at all. We don't compete with the accelerator guys. It's really in control plane. Think about those functions that are doing kind of housekeeping monitoring across the rack and communicating between the rack. There's really two types of products. There's the Layerscape-based products, which are top-of-rack switches. This is technology we've had for many years. We've continued to invest in it in other parts of the business. It's a 16 nanometer product. It's a 16-core, 16 64-bit Arm cores with a really big switch fabric in it. That's sold into top-of-rack switches and also an eight-core version, which go into NICs. That's which we reflect in our digital networking business.
We didn't talk about it until recently because we had been awarded the design win several years ago. They just weren't going anywhere. Like every company, we get design wins. Until they turn to revenue, they're not worth the paper they're printed on. It wasn't until late last year that they started to accelerate and contributed to the $200 million last year, and have really accelerated this year, contributing to the $500 million we talked about. That's kind of one half of the business. The other half of the business in data center is what the industry terms board management control, and these are functions on the different line cards in a rack that manage cooling, power management, security, inter-card security, zone-of-trust security. It's similar to the type of function that we do in factory automation.
It's really kind of in our wheelhouse of strength, if you will. Products there are our i.MX application processors, also our MCX root-of-trust security microcontrollers. We have a pretty decent portfolio of what's known as I2C and I3C for high-speed data movement, things like that. Different sensor products. I would say on the board management control, it's more akin to what we do in core industrial. It's a broader customer base. We have engagements with different hyperscalers, different server OEMs in Taiwan, and certain key reference design OEMs in the U.S.
Okay. Is the digital networking piece in comm infrastructure and then the board management in IoT?
Correct.
Okay.
Yeah. We know it's confusing, and we are thinking about maybe looking at how we may change the segments a little bit.
All right. I'll look out for it.
because it is confusing. We understand.
I guess, you mentioned it started to take off end of last year into this year.
Why that timing? Was there specific programs? Is it tied more to accelerated computing, or is it general purpose data center builds?
I would say on the Layerscape products, the top-of-rack switches, that is a few hyperscalers who are building out their own proprietary AI kind of racks, if you will. We've been awarded the designs a number of years ago. It just took them a long time to go to production. I couldn't give you more insight than that, Josh.
Okay.
If it had gone to market sooner, we probably would've announced it sooner.
Yep. Investors like big, round numbers. Have you spoken to how big this opportunity could be long term, any kind of TAM?
We've said the SAM, and I will tell you that a few investors laughed at us when I shared this data, but we think the SAM is about $4 billion growing at about a 10% CAGR. Right? It's not the whole AI data center, but it's a decent size. We think we can grow our $500 million at some multiple of that SAM. If SAM's at 10, and our ideal sweet spot is to grow 2x of SAM. You can kind of do a matrix and figure between 2x - 1x of SAM, get a sense of how it should go.
Okay. Thank you for the color there. Switching to the industrial IoT business.
There's a lot of questions of when AI is going to migrate out of the data center into edge environments.
Yeah.
I think you and your peers have talked about this trend for a while. Can you maybe talk about what you're seeing with your i.MX family and how we should think about edge AI or edge computing, whatever buzzword you want to throw around it, as contributing and changing the long-term growth trajectory of your broader industrial business?
Yeah. What we're seeing in the core industrial part is more and more of our customers are looking to run distilled models locally without having to go to the cloud. We're starting to see them be more forward-thinking and future-proofing their processor decisions. They want to have more headroom in the processor choice that they're picking. Now, today, most of our industrial and IoT processor families do have variants that have embedded, I'll call, moderate performance NPUs in them. The brand name's eIQ. It's a smaller performance. Most of the i.MX, and we're seeing customers starting to use that lower performance NPU to kind of sandbox ideas. We acquired a company called Kinara, which is a much higher performance Ara-based, the family's called Ara, NPUs, that can be multiple ganged up on one i.MX.
The way to think about it is the i.MX views the NPU from Kinara as just a resource, and it can use up to three different NPUs with one instance. We're really kind of amazed at the ideas customers are coming up with. I'd say they range in all sorts of forms and fashions. Fundamentally, the idea is they're taking large language models, they want to distill those and have them run locally without having to go to the cloud on an ongoing basis.
Okay. All right. We're bumping up on time. I did my best to avoid talking about gross margins, but I have to get there.
Okay.
You have a 57%-63% gross margin target. You're sort of operating in the middle of that right now. Your internal utilization rates, as you mentioned, 80% range. There's a little room there, but can you maybe walk through some of the drivers that'll get you further into the top half of that range?
Sure
How we should think about the timing of that?
Sure. Given our hybrid manufacturing that 40% of our wafers come internally, which are affected by utilization, the other 60%.
external are not affected. The rule of thumb that we've given is for every $1 billion of incremental revenue, we believe we can throw off 100 basis points of gross margin expansion. You heard us very clearly on the Q1 call, and we tried on our Q4 call to clearly state that we believe we can grow in the low double digits this year and next year. You can do the math. That kind of points, if you get to $15 billion-$15.5 billion, you should be about 60% gross margin using the rule of thumb. The one thing I know I've warned you on is don't take the rule of thumb and divide it by four and think you're going to get it every quarter. Mix does play an influence on gross margin. That gets you into the 2027 horizon.
The benefits from VSMC, which is the joint venture in Singapore, don't come to the table until 2028. At full load in 2028, it should add about 200 basis points of incremental gross margin at the corporate level. If we exit 2027 at, let's just do the math, at 60%, that would say that VSMC in its first full year of loading would add another 200 basis points.
Okay. Last question. I think NXP perpetually seems to represent an underappreciated capital return story. Can you speak to your priorities for uses of cash the next few years with those investments in mind?
Yeah. Our capital allocation policy hasn't changed. It's to return all excess free cash flow to our owners in the form of either dividends or buybacks. Now, the key word there is excess free cash flow not invested in the business.
I think most investors would agree the best thing we can do with your free cash flow is to help build the business. The last few years, we have had a number of requirements on the cash we generated, the three acquisitions we talked about. We also have the joint venture fab in Singapore. These have been all calls on cash. Those are starting to tail off. I think we're going to be past the big demands on cash here in 2026 and 2027, and we believe we can get back to that 100% free cash flow return. We've returned, I think, almost $23 billion in the last 10 years, which represented almost 95%, 96% of all free cash flow we've generated.
I think in the long term, we're consistent, but we will have periods when we're going to look at decisions to invest in the business first, giving cash back to owners.
Got it. All right. We are out of time. Jeff, appreciate you joining us.
Thanks.
It's nice to be on the other side of the cycle this time.
Yeah. Thanks.