NXP Semiconductors N.V. (NXPI) Earnings Call Transcript & Summary

Great. Good morning, everybody. My name is John Vinh. I cover semis here at KeyBanc Capital Markets. We're very fortunate this morning to have NXPI with us. We've got Jeff Palmer, Senior VP of Investor Relations. Congratulations on your retirement, Jeff. It's been great to work with you all these years. I'm very glad that you were able to include us in your farewell tour and...

I'm still here for a few more months, John.

Welcome Mike, back to the world of semis. I'm sure you're -- so anyways, why don't we just kick things off, and just talk about the cycle, right? Maybe, Jeff, can you just walk us through kind of what are the key trends that you're seeing right now?

Yes. So we reported earnings just a couple of weeks ago, so there's not a lot of new news. But what I would say is our view of the environment is considerably better now than it was 90 days ago and even 12 months ago. So things have clearly gotten better. Lead times are stretching out a little bit in certain areas, book-to-bill is solidly above 1 across all end markets, which is good. I would say distribution is in great shape. We're running at 11 weeks, which is our target. We have been running under that for a couple of years during the downturn, but we're back to 11 weeks in the channel. We are seeing escalations go up, so these are folks that come in late inside of lead time and place orders. And so we monitor those on a kind of event type of basis and they're up. We have had to do a little bit of price increases as we are seeing some inflationary input costs come in. It's been kind of targeted here in the first half and immaterial to the overall financials, maybe a little bit more so in the second half, we'll know when we -- the second half is all done. But all in all, I'd say we feel very good about where things are at. One of the questions you had earlier, John, I'll come -- maybe circumvent this, about restocking. Now as you know, in our auto business, about 60% of our auto business is we do direct with the big Tier 1s in North America and in Europe. And then the other 40% we manage through the channel is primarily in Asia, Korea, Japan, China and Southeast Asia. I would say the challenge we still face is some of the large Tier 1s are still holding very low levels of NXP inventory. Our ideals for them to hold somewhere between 10 to 12 weeks, and there are a few who do that, and that's finally we have a great relationship with them, but there are quite a few of large Tier 1s who are still between 3 and 6 weeks. And I just think the challenge is some of them don't have great working capital metrics and are going to just hope that we will manage the inventory for them.

So yes, that's a great, interesting commentary that you guys had talked about. So maybe we can just maybe dig into that a little bit. So how do you think this ultimately plays out, Jeff? And then are you planning to maybe hold a little bit more inventory for them? And then talking to one of your peers last night and they said that they're actually seeing signs that some of the OEMs are actually starting to hold inventory on behalf of the Tier 1s. Is that a trend that you're also seeing?

Yes, but it's not broad-based. It's very specific and targeted because the OEMs know fully that their Tier 1 partners are challenged and don't want to bring in too much inventory, but it's not broad-based. So what triggers it to change? I think the first time one of the -- here's the calculus, I think a Tier 1 is going through. They look at NXP and they say, lead times are sort of reasonable. They look at our balance sheet and they say, they've got 156 days of inventory, so things look good. That will work fine if you come in and we have a product that can finish goods. It will even work fine if we have the die and we can put it through the back end within the quarter. But there's going to be a day where something doesn't come in with an order, and we're going to be like going to have to go to raw die, right? You're going to have to build a wafer. And that's 16 weeks minimum. And I think that will be the wake up call.

Okay. Just relative to previous cycles, are you planning to hold higher -- slightly higher levels of inventory going into this up cycle here?

So we're holding -- our target is 110 days. So we're holding 156 days today. Probably by the end of the year, about 15 to 20 days will be buffer stock for some of our fab rationalizations we're going through. So we'd actually like to get inventory down a little bit, so there's not a plan for us to hold inventory on behalf of Tier 1s. I think one of the fallacies that people don't realize is our business is fundamentally a build-to-order model, right? So we don't generically build products and hope you come and buy it. If you, as a customer, don't forecast your demands on us, we're not going to make a guess. We have just too many SKUs, too many performance targets to hit, and so we're just not going to do it.

I would imagine with lead times extending that you are getting better visibility at this point. How far out do you have visibility to at this point?

It depends. It's Customer by customer. I mean you get some customers that give you a rolling 6-month forecast and you roll into that at all time. So you have other people give you a month forecast. So it varies. I don't want to give you a flat number, John. It's very end market specific, I would say one of the things that's unique to NXP is the size of our channel business through global distribution. We manage it very tightly. We have the ability to see what we ship into the channel, what the distributors ship at on a part level, customer level, but we also have the ability to see customers' backlog into distribution. So if, let's say, you are a customer buying from one of the large distributors, I could see your backlog building. I can't touch it, but I can see this building and I can see when you're ebbing and flowing. And that gives us a good insight on how the channel health is. And we feel very good about it.

Being a veteran of many cycles, Jeff, anything you'd want to call out in terms of differences with what you're seeing in the current cycle versus past cycles?

They all seem different. I mean the last one was post-COVID and it was -- I don't think any of us want to go through that again. But no, nothing like that.

Okay. Maybe you can dive into auto in a little bit more detail. Can you just talk about the trends you're seeing in automotive by regions? And maybe between what you're seeing in EVs versus non-EVs?

Sure. So overall, auto really started to turn the corner for us exiting Q3 of last year and into Q4, notwithstanding the low order rates or inventory levels, some of our Tier 1s have, I'd say auto is in very good shape. In this most recent quarter, all geographies were up, all product categories were up. We're feeling very good about product categories. So I think we feel good about auto other than just the issue we have with a few Tier 1s.

Okay. What about between Europe and China. It seems like the Chinese automakers are doing extremely well, and the European automakers are struggling a little bit. Are you seeing a little bit of that?

We see the same news as you see, John. I mean our view of auto was -- we think the auto industry fundamentally is a 90 million unit marketplace. You might have a little above that, some years, a little below, but that's about steady state. We try to win with every partner we can and then we let the consumer decide which car they want to buy. We don't obsess too much about car sales. We've obsessed a bit about global production, so we see how it's going. And we see exactly what you just mentioned, John. We see the Chinese clearly have won the EV battle from our perspective. Domestic China sales are a bit weaker, but some of the bigger guys are offsetting that with exports. So the market in China has not been that bad in totality, right? The Europeans, I think are trying to figure out the next phase for themselves. It's challenging, but I think they're doing well. I think the big brands, the luxury brands will continue to sell on the brand loyalty.

Great. So if you think about the growth for NXP and autos over the next 2 years, what do you see as the main key drivers? Is it going to be the same kind of company specific drivers that you guys have called out in the past, such as SDV, radar, BMS? Or are there other kind of emerging opportunities that could start to move the needle for you over the next 2 years?

Given the dynamics of the auto market where it's usually a 2- to 3-year design to revenue cycle, right? And I know you guys probably get tired of hearing us talk about SUVs and radar and stuff like that. But we're in the very early innings of that up cycle. We feel very good about it. Of this last quarter, it looks like the accelerated growth drivers were just under 50% of our auto revenue, up very strongly year-on-year. The nice thing also is the core business was up as well, so overall, things are good. We have engagements with customers that won't go to production until probably late '27 into '28 and '30, which are very interesting, like our 5-nanometer S32 end product. The alpha customer and that actually goes to production, probably model year '28, so they'll start taking some product late '27. The K5, which is our next-generation 16-nanometer zonal product is just, no better way to put it, but on fire in terms of customer engagements, but that won't turn into revenue for a couple of years. We are seeing -- one of your other questions was, are we seeing any interest in AI in the auto industry? And we are. It's kind of we're seeing some in-cabin applications, which are kind of interesting where if you sit in the car today, you have a lot of electromechanical switches. I think the vision some of them had is why have you actually turned the button on for your windshield wiper. Why not just say, turn windshield wipers on. And so there's some ideas around using large language models distilled to run inside the cabin that we're working on. That's kind of with our canonical product. But there's nothing in revenue just yet.

Yes. Maybe following up on that, right? You talked a little bit about what this AI car thing is. Is it simply just an LLM running inside the car? Are there other kind of attributes of what people are trying to define what this AI car is all about?

I think there are, John. I had to get front. I think it's very early days. I don't have enough of an insight to be the spokesman for it. I just know the one application I heard of is this kind of compressor distilled LLM that the car OEMs want to own themselves, which has to be able to interpret language from around the world and they want to be able to own that model, and they want to own the kind of, let's call it, the voice box, if you will. They don't want it owned by someone else, right? So there are some interesting ideas a couple of guys are working out.

That's great. Any questions? Great. So you mentioned Canara, right? So I think there was -- it seemed like a pretty big pivot from a messaging perspective in those last earnings call that you guys are pretty excited about fiscal AI. It sounds like you've got a compelling NPU from Canara which you acquired quite a few years ago. Can you just talk about the NPU that you have from Canara and how it's differentiated?

Yes. So I want to take a little step back. So Canara not the only NPE we have. So back in '23, we developed a smaller NPE that's embedded into our industrial microcontrollers, our Conatus family, the MCX family, i.MX family, and it's a good small NPU. It's about two tops, not super powerful, but it gives customers the ability to start testing the water, right? What we've seen with those microcontrollers and processors that have that small NPU in it, in '25, they made up about 6% of our total industrial IoT processors. In '26, it looks like it will be about 15%. So clearly, the trajectory is up. And so what we're seeing when we talk to customers is they're trying to future proof their systems, right? So that's on the kind of integrated NP of small performance. Canara is a much higher performance. It's about a 40 tops to us today. And the way it works is Canara device has to slave off an application process, or in our case, the i.MX. And so the i.MX views the Canara NPU as a compute resource, and you can gain up to three on one i.MX, and so there's good ASP uplift. But also from a customer perspective, they can scale their performance as they need it. And so the message we're hearing is customers want to be able to run large language models that are distilled to run in a non-cloud-connected environment, very much an industrial type of environment. The Canard opportunity pipeline has been the fastest-growing pipeline we've had in the history of the company. It was $1 billion last year, exiting last year. It's about $1.5 billion now. That's an opportunity pipeline. We start to convert proof of concepts and opportunities to design wins, and design wins to revenue. But it's a trend early on.

Yes. When you -- you seem pretty excited about physical -- when you look at the physical AI TAM what's the most exciting thing to you guys about physical AI?