NXP Semiconductors N.V. ($NXPI)

Good morning all. I'm Matthew Prisco, analyst at Cantor Fitzgerald, covering semis and semi-cap equipment. And today, I have the pleasure of hosting this fireside chat with Jeff Palmer, Senior VP of IR for NXP. Thank you so much for joining us today.

Good morning. Thanks, Matt.

So we'll be opening with a little near term. Okay. The analog story has been one of head fakes over the past few years. And last quarter, you highlighted a number of positive trends across your tracked KPIs. Are there characteristics in these trends today that make you believe we're perhaps truly on the other side of the cycle this time? And are there any surprise in trends, thus far, through 1Q?

So we were joking about this at the beginning, Matt. We're not going to be updating guidance today, sorry. But in terms of the KPIs that we do track, they do look like we have passed the bottom in terms of the cycle. Direct and disti backlog continues to build nicely. Customer escalations are continuing, so that kind of tells you that people maybe didn't lay in the right amount of material. We are starting to see short-term orders increase, escalations and expedites. EDI feeds from our direct Tier 1 customers continue to build, so those are all positive things. But the only challenge -- and I know these all sound positives. We don't see a V-shaped recovery. We think this is a bit of a grind higher from here.

Perfect. And then maybe digging into that, would be great to walk through your thoughts on the recovery side set of across maybe individual end market, starting with auto. As of last quarter, essentially back at peak levels, but you guys guided 1Q down mid-single digit sequentially, a somewhat seasonal guide despite tailwinds of now shipping closer to end demand and secular business strength, which we'll dig into a little bit later. Obviously, exited MEMS business, a headwind here. But why should we not see more of a cyclical snapback given that setup?

Well, I think a snapback makes you feel like the amount of inventory that's with customers is below what they need. What we've actually seen over the last 9 quarters is a lot of our Tier 1s in automotive burn down inventory to appropriate levels. Our view of appropriate levels is 10 to 12 weeks. And we track about 25 different Tier 1s. I'd say a large percentage of them are at that kind of appropriate level. We have a few that are actually below that, quite a bit below and -- but we've just not seen a snapback in terms of a V-shaped type of recovery. I think in terms of your question about auto guidance into Q1, you have to remember, you can look at it 1 [ or ] 2 ways. You can either rationalize out the MEMS business for Q1, where we have 1 month in the guidance, and then you rationalize it off of Q1 of last year in Q4. And you see it's pretty decent year-on-year growth.

Perfect. And then in industrial, I believe we're still about 10% to 15% below prior peak. So how should we think about the slope of this recovery, PMI trends and your kind of cyclical KPIs improved, just be more robust than an automotive growth from here?

I think so. I mean we think that the industrial business has some really good trends behind it. One of the things -- our Industrial & IoT business, about 80% of that goes to the channel. So it's a very much a channel-driven business, very long tail. One of the things that we look at there is end customer backlogs in distribution. So if you were a customer of ours, an industrial customer and buying through distribution, we can see your backlog. And what we've been noticing is backlogs have been building consistently month on month, week on week. That gives us 2 pieces of data. At least, we interpret it that way. If you're a small- or mid-sized industrial company and you continue to build backlog, you probably don't have any excess inventory. Two, you're probably feeling more optimistic about your business. Otherwise, why would you be putting backlog on. So we're feeling optimistic about what we're seeing. We're also seeing very good trends. I think we're going to go in just a little later on, Matt, for some of our newer products in the industrial IoT business.

All right. Perfect. Maybe hitting on the channel fill dynamic here a little bit. Last quarter, you talked about hitting the 11 weeks in 1Q and kind of holding steady from there, which, by my calculation, adds about $100 million to the 1Q revenue and sets up for a bit of a challenging 2Q growth compare. So as most conversations today kind of center around growth versus seasonality and you have a little bit more visibility into 2Q versus 5 weeks ago, how do you think about this dynamic where consensus currently modeling 2Q revenues kind of up high single digits when we adjust for that channel fill versus pre- and post-COVID seasonality kind of plus low single digits?

I know this isn't going to scratch the itch, Matt, but we're not going to talk about Q2 sitting here today. Our target long term in the channel has been to get to 11 weeks. We've been very clear about that. We've been below our 11-week target. I did the math last night. We've been below 11 weeks for 37 quarters. So that's how tightly we've managed the channel. And all we want to do is get back to 11 weeks. And once we get there, it's steady state from there. .

Perfect. How about on the memory side? Do you have any update on customer conversations there, maybe impacting end demand? I know you last talked about being an area of concern for customers into the back half, but since then, supply-demand imbalance and pricing really have only become more challenging. So have you seen any headwinds kind of surfacing there? Any conversations?

Yes. So we're not a buyer of memory. So we don't repackage memory in any of our products like some of our peers do. The topic of memory comes up in probably every customer engagement conversation. And it's kind of 2 flavors. You have certain customers who I would call very advanced and strategic in their supply chain management. Those conversations, they're more just annoyed of what they have to pay for the memory, but they've got their supply locked in. You then have other customers who, I think, take a maybe not as a strategic view of their supply chain and they're annoyed that they can't get product and they're having to pay a lot for it. But we're not yet, at this time, seeing any demand destruction from the challenges in the memory market.

Perfect. And then maybe lastly on the near-term side, pricing. We've already seen a few analog price increases year-to-date from peers. Should we expect the same from NXP? And how are you thinking about the pricing dynamics for the year?

So for our large direct customers and mostly in the automotive marketplace, we do annual price negotiations. They take place in December and take effect in January of the new year. For '26, they're all complete. They range in the low single-digit range like we thought they would. So that's good. We are seeing some increasing inflationary costs. There's kind of a breakpoint where if we can't operationally kind of digest some of those increases, we do have to pass on increases to our customers as we've done in the past, and we will do so again. But nothing to announce today about raising prices.

Perfect. Maybe moving on to the bigger picture setup, and once again, I'd like to start with kind of Auto here. You've got your core business and then the accelerated growth drivers. But in order to hit the midpoint of your target model in '27, excluding the MEMS business exit, you'd have to see this business grow in aggregate about 13% CAGR over the next 2 years. So this would also have to occur within a backdrop where end demand remains relatively subdued. The EV mix is a bit of a headwind. So can you discuss the drivers that give you confidence behind that growth rate despite these headwinds?

Yes. So what we've been most focused on, the core business really does tend to track production, right? And we said our core business should grow in the low single-digit range over the '24 to '27 horizon. Well, we've really been focused on our accelerated growth drivers. Now the first part of '25, there were some headwinds that impacted the accelerated growth drivers. But as we move through the second half, they did start to perform as we would expect. I think the math you're pointing to is correct. We're not backing away from our target for Auto growing 8% to 12%, which would mean you would have to be above that rate in '26 and '27. But we're not backing away from it right now. It's really a content per vehicle story. It's not about hoping that units, global production grows astronomically. It's about taking more share of wallet. And that's really around higher ASPs, higher value products that we've been awarded projects, and that really is what underpins a lot of our long-term growth outlooks.

All right. Perfect. And then maybe focusing in on the software-defined vehicle, which I would think, over those 2 years, is probably 1 of the leading growth drivers for the company, so correct me if I'm wrong there. But we know processing is at the core here for NXP. But what is it that differentiates the offering here? What drives the wins? Why is somebody choosing an NXP processor over the competition and maybe how to think about the stickiness there and ability to leverage those processor wins into further analog connectivity, auto attach?

I think what we're experiencing right now in the auto electronics market is kind of a shift in architectures. Historically, most cars were kind of these flat point-to-point architectures with many different vendors supplying different processors in different parts of the car. And that was fine for many years, but that did limit the car companies from being able to develop kind of a software upgrade type of a model, sort of like what maybe a Tesla would do. And we've seen, over the last couple of years, companies wanting to go towards this more software-defined type of model. And so what they want to do is build a hierarchy of processing fabric, if you will, starting with kind of a vehicle computer and the core electronics of the car, then kind of domains and then zones and edge nodes. The NXP SDV product line is a broad portfolio of MPUs, zone processors and domain processors. They are very high performance. It's a breadth of the products. It's also the other products that we can offer to the customer like automotive Ethernet. It's the software enablement, the digital twins we're providing. So it's really we're basically working with the auto OEMs to help them implement this new architecture of the cars. It's a much bigger software effort than, let's say, just selling a socket to a customer. And a lot of what we're guaranteeing is the performance at the system level, not just the performance at our socket level but saying, okay, if you think about a zone, a complete zone in a car, we can guarantee the performance on its own because we do all the products within it, that makes sense. It tends to be a much stickier business because if you're selling a processor to a customer, they're committing to writing software on your platform for a number of years. It's not something that gets swapped in and out every year.

Do you find that -- I mean, when you think about competition in this type, are there other guys you're competing with on the system-level approach really tying this all together? Or are you more competing today with discrete solutions and trying to displace those?

So the way we'd ask you to think about the modern car, kind of SDV, there's kind of 3 big functional areas. There's the IVI space or in-vehicle infotainment; and in that space, we have to give it to our peer Qualcomm. They've done very well with the immersive dashboards. Then, the other big kind of space is ADAS. And in that area, we participate in radar type of solutions. We don't do cameras. We don't do LiDAR. But there is also an opportunity for something they called ADAS fusion. And that's usually something like Qualcomm or NVIDIA. Those are tougher areas given our product portfolio and where we're focused in. The third area is what kind of we termed the core electronics of the car, what makes a car a car. And that's where the S32 family is really focused on.

Okay. And maybe digging into that, how do we think about the product road map here and design win traction? Early feedback on that S32, I think you guys have the N5 (sic) [ N55 ] and N7, K5 iterations, maybe timing of those ramps, relative magnitude of contribution. And any step function increases we should be thinking about as these products ramp and kind of open new opportunities?

Sure. So the S32 family, we said would grow kind of a 20% to 30% CAGR '24 to '27. The baseline of that in '24 is about $1 billion of business, and these were primarily our 16-nanometer and larger type of domain processors. In '25, that business, the SDV business, grew in the low teens year-on-year. It was a little weaker in the first half, accelerated above that in the second half. So we feel the trajectory is coming back correctly. The design win traction for the 5-nanometer products is very good, but they won't go into production till probably late '26, early '27. There's still a little ways away. The K5, which is the industry's first 16-nanometer auto microcontroller family, zonal microcontroller family, we just released to sampling to customers. Traction has been very, very good. But again, given our design-to-revenue cycle, it's not going to impact materially our '27 target.

Okay. Makes sense. So I guess moving to industrial and the intelligent edge. I feel like the industrial story historically been a bit more difficult to conceptualize versus auto, but few would argue today the prospects of the industrial edge within this new AI paradigm. So maybe we could start with what drives the wins for NXP here. Are there anchor products that we should be greater focused on? And how important is it to think about the opportunity being at a system level versus selling these more discrete components to customers?

Sure. So in the Industrial & IoT business as well as for the complete company, over 50% of the revenue is processor-based. In Industrial & IoT specifically, we have a broad portfolio of microcontroller families, the MCX microcontroller family; the RT family, which is kind of a hybrid microcontroller apps product family; and then we have the i.MX application processor, which is the industry-leading industrial application processor. When our teams go into these industrial and IoT markets, they always lead with the processor first. If we can win that, then you as a customer are making a commitment to write software on my platform. Once I've been awarded the processor, I pull along PMICs because most of these products are built with companion devices that are matched with them. So you win the PMU. You then look to pull in other things like connectivity and security and other analog type components. And you really do build reference designs for customers. These are a long-tail business, tens of thousands of customers, and they're very, very smart in their end markets, but they're not real smart in picking and wanting to choose individual semiconductor socket, so to speak. They really want a company to come to them and say, "Here, we've taken that kind of reference design problem off your shoulders." And some of the companies pick up the reference designs as is. Some relay them out, but that's how we go to market.

Perfect. And maybe to finish on this product or this topic of product differentiation, can you perhaps walk us through recent acquisitions and how these potentially allow for further differentiation of NXP's offering?

Sure. There was 3 acquisitions last year. They all happened very quickly, and I think some folks misinterpreted that maybe we were changing our M&A strategy. It was really more around deal timing that they just all happened in a short time. Starting first off in automotive. TTTech Auto is a software company that we acquired. The reason we acquired them, they were about 1,000 very experienced software engineers in the automotive market focused on functional safety and security, and we really wanted to use that capability to complement our S32 CoreRide platform. So it's really around -- as I said at the beginning of the conversation, Matt, this SDV effort is much more software oriented, and so having that capability in-house and being able to engage with our customer software teams is key to being successful. So that's why we bought TTTech Auto. They also have a product that's called motion-wise, which is a middleware product that kind of sits above the operating system but below the application layer in the car. And it's early days there. It will take a number of years to really kind of productize that to where customers are very excited about it. So TTTech, while very strategic, will not yield any real revenue before '28. The second one in automotive is a company called Aviva Links. Aviva Links is a small startup in the Valley that delivers high-speed asynchronous SerDes type technology. Think about an ADAS system where you'll have multiple modalities, cameras, LiDAR, radar, all generating a lot of data that they need to force back to the kind of a central processor, but they don't need the same amount of data back from the processor to the edge node, so kind of asynchronous type of a structure. And that's what this Aviva Links product does. We acquired the company very much at the complement or kind of pushing of a customer. And so there's already design wins that have come with it. It complements our in-vehicle networking technology, revenue probably late '27, early '28. And then lastly, in the industrial and IoT space, we acquired a company called Kinara. And Kinara's another Silicon Valley startup doing neural processing engines for running large language models in an inference environment at the edge. And so we've been working with them for a while. These products have to slave off of kind of the central processor of the system. And so we've been showing some really interesting engagements with customers. Customers are coming to us with very innovative and creative ideas how they want to use this technology. So this is not doing learning at the edge. This is running actually the inference at the edge. A lot of the customers, the reason why they want to do that is they don't want to go to the cloud and come back and incur the latency of the round trip to the cloud and back. They also want a much more higher security environment. And so early engagements are very, very good, but like a lot of things in industrial, the time to revenue is probably at least a year or 2 away.

Sure. All right. Help moving to the operating leverage side. You guys have undergone a structural transition over the past 5, 10 years, resulting in meaningful resiliency through this past cycle with operating margins contracted roughly 5 points peak to trough for you guys versus peers over 20 points. Can you walk us through what has changed to drive that resiliency, the sustainability of that dynamic moving forward? And any changes to come with new CEO now at the helm?

So the biggest thing that changed kind of the financial structure of the company occurred over the last decade where when I started with the company probably over a decade now, our fixed versus variable cost structure was we were about 70% fixed cost and only 30% variable, so very IDM-like, right? Over the last decade, we've flipped that 180 degrees. We're now currently 70% variable cost and only 30% fixed. And that 70-30 mix will probably go to 80-20 over the next number of years as we rationalize our legacy 8-inch factories in the U.S. and in Europe. So we are moving more and more towards a variable cost model. The question about sustainability, we believe it's sustainable. We believe it's -- the model has shown to be fairly resilient in the last cycle, where our peak to trough was much less than many of our peers. We did not take our utilization of our factories really, really way down, so we're able to maintain utilization at reasonable levels. And in terms of the new CEO, he's been with the company now 11 years. And so the strategy you've seen out of NXP or the messaging, he's been part of that, right? He was one of Kurt's direct reports, Kurt Sievers, our prior CEO. So I don't think you should expect [ brand-new ] data. I don't think you should expect any change in the financial model because Rafael is now at the helm.

Perfect. And then as we contemplate the leverage into the up cycle, you've talked about for every $1 billion in revenues, should drive about 100 basis points gross margin expansion. Is that still the right bogey to be using today? And then maybe outside the dynamic, what are kind of the rank order primary levers we should be thinking about to get those gross margins towards the higher end of the targeted 57% to 63% range?

Right. So yes, the $1 billion driving about 100 basis points of gross margin expansion is still valid. That rule of thumb contemplates revenue growth, improving utilization, mix and pricing. So it's kind of altogether, right? That -- what that encompasses. In terms of how to get it towards the higher end, well, as you know, each one of our end markets has a range for revenue growth. And so if you want to drive to the higher gross profit, you have to drive the higher end of revenue, very simple.

All right. So revenue, primary driver to drive [indiscernible].

Revenue is our friend.

Okay. Perfect. And then maybe as we start to think beyond that target model, can you walk us through the VSMC dynamics, the latest thoughts on progress versus plans, P&L impact ahead of the ramp and maybe the timing of the ramp in subsequent 200 basis points flow-through to gross margins?

So the VSMC joint venture is going very well. I'd say they're ahead of schedule. For those who maybe are not aware, VSMC is a joint venture we've developed with Vanguard in Singapore. They're building out a 2-phase 300-millimeter factory for trailing and mixed signal. First phase is already -- the facility, all the shelves are built. The utilities are in. They're starting to load equipment in, so the progress is going very, very well. I think what we really want to highlight is how appreciative we are of TSMC's effort in this. If you know, TSMC owns 30% of Vanguard, and so they're kind of an invisible hand in this JV, if you will. Because of that, the factory that's being built in Singapore is a match exact to a factor you might see in TSMC's network. We are actually licensing TSMC's process flows to run in that factory, so it should be very match exact for what we already buy from TSMC. In terms of the investment, our original cost of the investment was $2.8 billion. We're about 50% through that investment. The other 50% will go out probably in '26 and '27, is the current plan. And then I think you had a question you want to talk about the equity accounted investees related to that. So as you know, since we will be a 40% owner of this JV, there's 2 ways we get a benefit. One, we're able to access lower-cost wafers. So our -- that will flow through our COGS on the P&L, and that is what gives you the 200 basis point gross margin expansion. But below the EBIT line, we have an entry called equity accounted investees. This is where we would recognize our proportionate share of losses as well as profits. Now the losses occur because you're building a new factory. Now when we originally announced this JV, we thought that those losses would amount to about $200 million in aggregate between '24 and '27. In '25, the losses were only $4 million. We think this year, there'll be about $20 million total. And so when we look out into '27, we actually think the losses will only amount to about $100 million, so about $75 million below what we originally thought. So you put that all together, instead of $200 million of start-up losses, we think it's about $125 million in aggregate over the 3-year period.

And does the remainder of that get pushed out? Or are there efficiencies that you've experienced and now the costs are lower than expected or...

No, to be upfront, Matt, we were probably too conservative when we gave the $200 million, right? It was our best estimate at that time. That was even before shovel was in the ground, right? We looked at other projects, consulted with other folks in our operational team. And we thought, all right, $200 million is a good placeholder. But now here, we're moving through it and it's just moving along better than we thought.

Got you. And then maybe just one more on that, that 200 basis points benefit from when VSMC ramps, how do we think about timing there? How quickly can we get up to that level of benefit?

Well, the Vanguard team wants to get the factory online very quickly because while we own 40% of it, the other 60% is a commercial foundry for mixed signal. And there's a lot of demand for that, especially with TSMC's messaging. They don't want to do trailing edge that much anymore. Our current plan of record is that factory will be up and online in '28 and will be fully loaded in '28. And so over the course of '28, it's how you'll get the 200 basis point benefit.

So hopefully exiting that year.

Yes. So if you think -- if you kind of put it all together, if we continue to execute to our long-term revenue growth, right, between now and '27, we should hit roughly $15 billion, plus or minus, in '27. Given our rule of thumb of $1 billion, 100 basis points, that means you're going to be at about 60% gross margin exiting -- or in '27. Once the fab in Singapore is up and fully running, you'd get another 200 basis points on top of that.

Perfect. On the OpEx side, you guys have been very clear about your focus managing the business intensity of 23% through cycle. But as revenue growth theoretically outpaces that target model over the next couple of years and intensity is already sitting at that target, how should we be thinking about midterm OpEx growth and potential leverage from that dynamic?

Yes. So the model doesn't change, Matt. So the model still is -- we think 16% is the right amount to invest. 7% on SG&A is correct. You'll get a little bit of leverage on G&A. And yes, I understand the question that, okay, if we're growing above our target rates, shouldn't we get lower intensity of R&D dollars. Possibly, it might be because we won't be able to hire as many people. But the model has not changed, 16% and 7%.

Okay. I was going to say, basically, can OpEx even keep up.

Well, that's really what it is, right? I mean, if we really can drive above trend line growth, can we hire enough people at a 16% hiring rate? I don't know.

All right. Perfect. So China, historically, I believe NXP's exposure here, oftentimes misunderstood. And hopefully, that dynamic has remedied a bit by accounting for the region not changing to a headquarter basis. I think at '25, 17% of revenue's from China, slightly below China's contribution to GDP. So it kind of makes sense there. But thinking big picture, as China works to minimize U.S. exposure and source domestically as much as possible, how do you view your positioning in the region as a Netherlands-based company? How defensible is your portfolio? And are revenues here something that should grow in line with the company over time or perhaps even decline as China looks to source more and more domestically?

So from our perspective, China is a strategic market for us, very important for us in multiple end markets, automotive, industrial and other areas. So we are going to stay engaged. In terms of the competitive dynamics in China, the reality is we continue to compete with our Western peers in China. Kind of paraphrase one of the large electronic -- electric car companies in China, they say to us, "Hey, if you can innovate at our rate and stay on our design cycles with us and continue to bring us value, you will always have a place here at our company. You'll always win awards. If you fall behind or you don't innovate like we think you should, we'll choose another vendor." But we don't really see today local players doing the same type of products that we do.

Do you think there's an advantage to being the Netherlands-based in this conversation versus the U.S. peers? Or does that not really factor in?

It's not a hall pass. I mean it's nice, but at the end of the day, in the Chinese market, it is very much an innovation game. If you bring innovative products and you can keep up with this kind of development cycles that the Chinese customers are running at, that's what matters, not that you're a -- hold a Dutch passport.

Perfect. Last quarter, team reiterated long-term model remains intact, which suggests revenues, by my calculation, roughly [ $15.5 billion ] in '27 when excluding the MEMS business impact. So can you maybe provide an update on how you see the individual end markets tracking versus the prior growth ranges? And has anything structurally changed for NXP since initiating this target back in 2024, either operating custom partnerships or secular growth opportunities?

No, nothing has changed in terms of our focus. In our businesses, especially in automotive and industrial, which is over 80% of the revenue, these are very long design-to-revenue cycles. And as more and more software is -- these design cycles are software dependent, those -- that design-to-revenue cycle stretches and becomes longer and longer in nature. There's no change to our growth rates. Automotive, we think will grow 8% to 12% on a 3-year CAGR, Industrial & IoT at 8% to 12%, mobile at 0% to 4%, and we think our comm infra business kind of stays flattish. Depending on what corner of that kind of ranges drives the total in the 6% to 10% rate, Matt. But there's nothing really new to update from that perspective.

Perfect. So looks like a minute left. It's been a busy conference season for you. Across all your conversations, anything you think the Street is underappreciating from a portfolio quality standpoint or just the overall NXP that you'd kind of highlight?

Well, I mean, we have kind of a simple thesis that we offer, and that is we believe we can grow in the high single digits over a number of years. We think we can drive our gross margins into that north of 60% range over the next number of years. We'll throw off very robust free cash flow, which will return to our owners, and we think that results in our EPS doubling between 2024 and 2030.

Perfect. Perfect wrap-up. Thank you, Jeff. I really appreciate it.

Thanks, Matt.