Lattice Semiconductor Corporation (LSCC) Earnings Call Transcript & Summary

Good afternoon. Thank you all for coming. I'm Stacy Rasgon. I cover the U.S. semiconductor and semi-cap space here at Bernstein. And it is my honor to have our guest here today, Jim Anderson, the President and CEO; and Sherri Luther, the CFO of Lattice Semi. Before we start, I want to mention if you have questions on the back of your program, there's a QR code you can scan. It will take you to our Q&A portal. It's called Pigeonhole. You can submit those questions there. We'll have time for Q&A at the end. I don't cover Lattice. It does, however, come up pretty frequently in our conversations and actually more recently, investors ask how it fits into the broader FPGA space, especially since some of the larger competitors got acquired and how they differentiate those -- against those larger peers. They ask the new efforts to move up that FPGA stack into higher performance regimes. And they ask how much further the company can take things as they continue to extend their financial targets upwards as they did again recently, and we'll talk about that, I'm sure. So to answer those and many other questions, it gives me great pleasure to welcome Jim and Sherri. So thank you so much for joining us today.

Thank you, Stacy. Appreciate it. Yes, I appreciate that.

Thank you, Stacy.

Maybe just to touch on -- you guys did just have an Analyst Day, right? If you have to summarize it like 2 or 3 sentences, like what would you say?

I think I can do it in one sentence. Lattice is going through the biggest product portfolio expansion that we've ever done in the company's history, and we're 40 years old. So that's a pretty big deal, 40 years this year. And that's a really big deal for us, but it's an even bigger deal for our customers. Customers are really excited.

Got it. Maybe just as a high level, can you talk about that product portfolio in those announcing -- we'll go deeper certainly as we go. But just to the high level, like what's new?

Yes. So over those 40 years, where Lattice has historically innovated is in what we call the small FPGA part of the market. So think of these as very power efficient, very easy to use, small-sized FPGAs and going to just all sorts of different devices.

You wanted to talk about what an FPGA is, by the way, just given our...

Yes, I was assuming that was given. But -- so an FPGA is -- think about that as just a really flexible silicon device, where you can program your own personal algorithm, your own personal kind of innovation, you can program that right on to the chip itself. And this has been a technology programmable logic that's been around since the '80s. And it's super useful in many different applications because you may need a device that's really accustomed to your particular system. And you might want to be able to change and reprogram that device over time and an FPGA is a perfect fit.

Got it. And so you've been historically playing in the lower end part of the market?

No, not lower end, small FPGA.

Small FPGA, excuse me.

There's nothing low end about 70% gross margin. Well, we've been doing support [indiscernible].

[indiscernible].

It's mostly Sherri. She gets the credit. But if you look over to Lattice's history, it's really been the innovator in that small -- small form factor, power efficiency. And so we're definitely continuing to innovate there. In fact, we've really cranked up the rate and pace of new product introductions over the last few years basically tripled our rate of new product introduction. But then the newest news is in December of this past year, we introduced a new product line that's targeted at the midrange part of the market.

This is Avant.

This is Avant. Okay. So that Avant is the name of the platform. We introduced the first device family based on Avant and we'll bring out a series of device families basing on Avant. Avant doubles our addressable market and it basically allow...

[indiscernible] what is the...

In the most -- in our Investor Day, we estimated that the market size 5 years from now for us, our TAM is about $10 billion. Okay. And so Avant doubles that within that $10 billion. And the great thing is it's largely all the same customers. So if you look at the target customer list for Avant, 90% of those customers are already customers of Lattice today. They're already using the software that's used on the small FPGAs, they're already using that today from us. So it's easy to adopt.

So the software scales like directly across...

[indiscernible] the software so that our customers could leverage that onto Avant to try to make it really easy for them to adopt those new devices.

Got it. And the prior family, was it Nexus? Is that the small FPGA?

Yes. Nexus is the name of the platform of device families that we built based on that platform for the small FPGA segment. We introduced that in, I think, 2019. And so we've been introducing new device families based on that platform. We've got 5 now that are -- the fifth one went into production in this past quarter. All 5 of those device families are ramping. We just launched the sixth one, and we announced the seventh one will be out later this year. So we've got, again, a steady beat rate of new devices for the small FPGA segment. And then Avant is totally additive. It doesn't cannibalize small FPGA in any way, it's added.

Got it. What are your competitors -- so you have larger competitors that have mostly been acquired. You had [indiscernible], do they not play in these markets at all?

No, they do, but I think their strategy and focus is just different than ours, right? So as I said, we've typically focused our innovation around power efficiency, small size, ease of use, and if you look at midrange FPGAs, a lot of those same characteristics are really important to mid-range applications as well. In fact, most of our customers only use midrange and small FPGA. So like industrial and automotive customers, primarily just use mid and small, they really don't use the large FPGAs. But our competitors have generally been more focused on building very large FPGAs. That's a different architectural optimization point and so that we're just focused on a few different parts of the market.

So if you're moving into the midrange and that was, I guess, historically underserved, like what were those customers buying for those applications already today?

Yes, they would be buying Altera or Xilinx FPGAs. In fact, actually, it was the customers that told us to invest in Avant. So in 2019, we had just joined within the -- this was within the first year we were at the company. So we were showing them the new product road map for small FPGAs. So we're showing that to our customers, showing them all the Nexus devices that we're going to bring out and the competitiveness and the customers looked at that and said, hey, it looks like a great road map, but why are you guys just limiting yourself to small FPGAs. Why don't you just take that same power efficient architecture and build midrange FPGAs as well because again, we have thousands of customers. Most of our customers only use midrange and small FPGAs. So their point was by introducing a midrange product line, you could basically cover all of their needs. And so it was really customers that gave us the -- I wish I could take credit, but it was customers that gave us the idea to invest in Avant, and they're really pleased because we've launched it on time. We had promised Avant before the end of last year, and so now it's launched, and we're designing that in with our customers.

Got it. That's not being taken up like by microcontrollers or anything at this point or...

No, there's -- FPGAs have always been used in certain applications. again, because you can program the silicon itself and you can reprogram that over time. So there's just -- there's applications where an FPGA is just an inherently better fit. One of the places that we're seeing FPGAs grow in some of these Edge applications is when people are trying to add some level of artificial intelligence like inference at Edge...

We'll definitely talk about...

You're going to ask about that. Okay, yes.

What I would like to know was you mentioned you launched the first product based on the Avant platform. What was that product -- what application, what end market is it going into?

Yes. It's the first device family. It's called the Avant-E, E is for Edge, and so the product itself is more optimized for edge applications. So for Edge AI, which I guess we'll talk about talk about it -- but -- so one of the things that we see customers doing is -- and we've been talking about this for many years, so long before it became fashionable to talk about AI, right? But we've seen customers trying to add more intelligence, more decision-making capability to their Edge devices like industrial robots or automation equipment, where we've seen tremendous growth. And so in those applications, Lattice FPGAs are a really good fit because those algorithms are inherently parallel. They can inference algorithms or parallel. They can program our devices to be basically custom parallel processor specific for their algorithm. And they know those algorithms are going to change over time. And so the FPGA, they just reprogram and so we've seen good adoption of our small FPGAs in those applications. And so we did a version of Avant, our first version of Avant for those same type of applications.

It's interesting because you usually think about artificial intelligence as being like massively compute-intensive, and yet you're talking about small [indiscernible].

This is inference, right? There's training of the network, and then they're using the network to make decisions. So in Edge devices, what's usually the processing that's usually being done is inference processing. And so you want to be able to do inference processing with very low latency, very high power efficiency. And there are some tasks that fit very well into a small FPGA device from us that is incredibly low power and is a naturally good fit. And then that first Avant device that we're launching is targeted there.

Is this the Lattice AI play? Or is the AI play broader than that? Because I know you sell a lot of stuff like Intel servers and GPU servers and...

Yes, that's a good point. Okay. So one place that we're seeing adoption is in the Edge, but the other place is in the data set. So set aside the AI for a second, I'll come back to that. But in the data center, which is primarily servers in the data center. When we started, Lattice had a little bit of a toehold in the server segment, and we've grown that toehold significantly over the past 4 or 5 years. Now what we're doing in the data center on servers is different than what others are doing. What we're doing is control management and security of the platform. So we're not -- when I talk about FPGAs or Lattice chips getting used in the data center, it's not the compute but it's making sure that the system can be remotely managed, controlled and that it's secure. In fact, one of the first devices that boots up in a server is oftentimes a Lattice device. And so the Lattice FPGA will go out and make sure that, that system is secure, that it boots up in a secure way, that the firmware hasn't been corrupted and if the firmware has been corrupted it can repair that, et cetera. So those are the type of tasks that we're doing in servers. And we've seen that position within servers significantly increase over the past years. In fact, that's one of the main growth drivers of our comms and computing segment, which has grown double digits for 4 years in a row.

What's your attach in servers?

Yes. I was just going to mention, so that revenue growth that we've seen is a combination of both attach rate and ASPs. Our attach rate years ago, it would have been, say, 20%, it's now over 1x, which means almost every Lattice server ships with at least one...

Mostly more than one.

Yes, exactly. Almost every server ship with at least one Lattice chip and there's a lot of servers now that are shipping with multiple Lattice chips, right? Because they may have a chip. A Lattice chip on the motherboard and on the plug-in boards like the security plug-inboard, et cetera. So our attach rate has gone up significantly. And we brought out new capabilities, new functions that have increased our ASP over. And so what we measure is the dollars of content per server and we've seen that significantly go. And in the new -- and the new generation of servers, that's just starting to ramp in the second half of this year and into next year. There's, again, another big step up in our dollars of content. For sure, we estimate about 50% increase in dollars of content per server. And then you asked about -- we kind of go off on the spread on AI. So things like generative AI, large language model processing or just in general, machine learning that's certainly driving a tremendous amount of compute cycles into the data center that drives growth of just data center infrastructure in general, but growth of general purpose servers, but also servers -- you see more servers that are AI optimized, right, that have extra GPU content, et cetera. And so we're benefiting from the growth of AI as well. If you look at the...

Those are -- also your attach rates on those are also high?

Yes. So again, we look at dollars of content. But if we look at dollars of content, per more AI optimized server versus general-purpose server. It's either -- depending on the system configuration, it's either the same or [indiscernible] -- so yes, so we see benefit from the growth in that as well.

I want to ask a near-term because I'm not asking for any kind of guidance update or anything?

Okay, good because that would be [indiscernible].

What -- what are you seeing in data center right now? Because clearly, we're seeing a big surge in demand from these AI servers. At this point, it doesn't necessarily look like overall CapEx budgets are going up. So just what are you seeing in server and data center.

Yes. We certainly saw -- let me go a little bit back over the past quarters, we certainly saw a little bit of softness in the deployments of servers. We still grew year-over-year because of not the underlying units, but because we were growing dollars of content per server sell, for instance, in Q1, our comms and computing segment grew year-over-year, right? But yes, we certainly saw some of that softness. We've definitely seen a pickup in any server volume that's related to AI and that's not a surprise, right? So generative AI, et cetera. So we've...

[indiscernible] Last week, it's us. So maybe...

Well, we were seeing that even before, maybe not a surprise to us. But so we've definitely seen a pickup in that. But I think server demand in general will start to improve in the second half of the year.

New platforms.

Yes, new platforms, new platforms always drive some additional uptick in consumption.

Got it! Sherri, I wanted to ask you a few questions. So we talked a lot about the new platform introduction of new products. So clearly, it's improving the financial model. Can you maybe just give us a review of the changes to the model and maybe talk a little bit about some of these new product introductions that are actually influencing that?

Sure. So at our Investor Day actually on May 15, a couple of weeks back at NASDAQ, we laid out our new financial model, our long-term model where we raised our targets. We raised our target for revenue growth. Revenue previously was sort of in the low double-digit range with our prior target. We raised that to 15% to 20% revenue growth and again, this model is over a 3- to 4-year time frame. We raised our gross margin target to the low 70s.

Where are you running now? It's pretty close to that, right?

Yes. Last quarter was 70.3%. Yes. But low 70s. So we raised that target. Previously, it was 65%, our target. And we -- for OpEx, we put out a target of 30% for our OpEx spend, which incorporates you mentioned product -- new product development, expansion of that, and certainly, that incorporates it because believe investing in the long-term growth of our business is really important. So investing in our long-term product portfolio, investing in demand creation, customer support, all those areas we view as very important for the long-term growth of the business, and we want to continue to do that but do it in a disciplined way. And then we also raised our target for operating income. We raised that to the low 40s, and we put out a new target for free cash flow margin of greater than 30%.

So this is the first time you've had a free cash flow target.

Yes. The first time we put out a free cash flow margin target. This business generates a lot of cash. And so we've been focusing on cash for the past 5 years. I mean, it certainly was an area of focus for me when I joined the company definitely. And so in recognition of that, we want to continue to maintain that focus while investing in the business. But with record operating income that we've generated and that disciplined focus on investing, we're really happy with our free cash flow target, again, of greater than 30%.

And the growth target is -- what was the baseline for that? Was it last year? Or like what was...

You know that for the revenue growth you're referring to, I mean, it's based upon what we see out there for our products. We've got a lot of -- there's a lot of underlying secular growth drivers that we see for our business. We talked a lot about that at our Investor Day and Jim talked about that here within our strategic core market segments of comms & compute and industrial and automotive. But there are also Lattice-specific growth drivers. Our -- we also mentioned our seventh device will be announced later this year and for Nexus platform and on Avant. So -- we see a lot of underlying growth drivers there.

So the new product side.

So Nexus is not like a stale platform like...

No. Nexus -- we would expect Nexus to ramp for years.

So what is the typical life cycle of when you introduce the product. What is that [indiscernible].

[indiscernible] they last forever.

15, even 20 years, last for very long. That's why a pre-Nexus has been growing. It certainly has been a contributor. And we expect Nexus to also be a long-term contributor as well.

Yes. We'll get design wins on products that have been in production for 10 years or more. We'll get new design wins, so customers designing that into new applications. So FPGA is -- it's different than -- traditionally, you think of like memory or CPUs where the new version of memory or CPU comes out and the old one drops off and the new one picks up. FPGAs, even in older products last for a very long time. And it goes back to the nature of the chip itself being reprogrammable, very flexible. So even older products can continue to get designed into new applications -- so -- and we love -- well, Sherri loves to win pretty Nexus products and get designed in because the incremental investment is 0 or close to 0.

Have you guys ever given like a view of like how your revenue breaks out by age of the products. ADI used to give this great chart where they would show the revenue. And they'd always use this bar at the bottom and the products are like 30 years old, it's 10% of the revenue, whatever it is, still selling...

Well, actually, as you mentioned, ADI, I think our FPGA products in terms of their longevity are more similar to analog-type products than to processors or memory. But we've never broken that out, but we -- in the Investor Day, we did say that even our pre-Nexus products, we expect those to continue to grow over the coming years. Modest growth rates. That's not the fastest growth area, but we expect those to continue to grow. And actually, some of the new software packages that we've developed have helped us find new applications for those devices and kind of extended the longevity of those devices, which is a really good thing.

Let's talk about software -- actually it's a good segue because FPGAs historically have a reputation of being very problematic and difficult to program. Some people have pointed to that reason for why GPUs have like taken over in terms of [ accelerators ]. I know -- some of your larger peers were targeting those markets with like various degrees and mostly non-success. What are your thoughts on that? Are you doing things to make the programming environment easier and likely what is that -- what does that look like?

Yes. Short answer is absolutely. So when I joined the company, I'm not an FPGA guy.

You were an AMD but...

AMD, right. So I was doing computing and graphics at AMD and so when I joined Lattice, I didn't have a history in FPGAs. And I knew I had that same perception that, hey, these things are difficult to program, et cetera. And so one of the things that we really ramped up in that first year, and we've continued to ramp up the investment is really ramping up investment in software. And what we've specifically have been focused on and our strategy is to develop software that makes it just really easy for our customers to design our chips into their systems to either switch from a competitor's device to our device or if they've never used an FPGA before, to make it really easy to adapt. And so the specific strategy that we took is we've been developing application-specific solution stacks. So these are targeted at common applications that our customers would use the chip for and we basically give them a prebuilt set of libraries, tools, reference designs that make it much easier for them...

All different, though, based on the specific application...

Yes, exactly. I mean they may have some components that are similar across but they're specifically targeted at common usage models, so like the most recent one that we announced, which we'll launch this year is our automotive solution stack. So again, this is software that makes it easy to design our chips into automotive [ ethic ].

To do what? What kind of like...

Sensor aggregation in cars, video display, a number of other type of applications. The other one that we just introduced last year was around factory automation and robotics. We've got one -- I mentioned security before. Remember in the context of servers we introduced a software stack specifically for platform security. So we're seeing that now not just adopted in servers, but adopted in other types of equipment like data center networking, things like that. So we've got now 5 that are already launched and the sixth that we'll launch this year.

You haven't talked about what that sixth is.

No, no, that was the automotive. Yes, that's the automotive one. And so -- and we've said we're continuing to develop additional ones. And even those ones that we've already launched, we continue to refine and improve over time. And if -- the great thing is, we measure adoption rate and if you look at, well, the adoption, when I started was 0, but now our adoption rate is over 50%.

Of solutions.

Of solutions stacks, yes. This isn't the development environment, right? So solution stacks, meaning that when a customer chooses a piece of Lattice silicon, over half the time, they're now adopting one of those solution stacks. And the metric that Sherri likes, in particular, is we measure the ASP difference.

Well, that was my next question. How do you monetize it?

Exactly. That'd be -- good question. So how do we monetize it? We bake it into the cost of the chip and software together. We give it a single price and when we measure the ASP for the design wins that have software attached versus not, there's a higher ASP, and it's significantly higher. And that because that's software driven, the COGS is 0, so the gross margin is higher. And so as those design wins convert into revenue, they are naturally...

50% of your sales now are solution...

Design wins. Yes, [indiscernible] design wins of the last...

How much of the revenue is solutions?

That's still a smaller percentage, right? Because these are solution stacks that we brought out over the course of multiple years, right? But those will convert to revenue over the coming years, and it will be a [indiscernible]

Just interesting because it's higher gross margin presumably and higher ASP and your target is low 70s on gross margin and you're kind of there. And it sounds like a pretty small fraction of your revenue can currently...

Yes, that's true. So this would be a tailwind to the gross margin.

I can kind of see how it's -- where it's coming from.

Yes. But actually, the thing that we like the best about it is we've measured that this helps customers get to revenue quicker or get to production quicker, which drives our revenue faster. And then the other thing that we believe is that it creates much more long-term stickiness of our products because if customers integrate that software those solutions into their system level software that creates multigenerational stickiness. So there's a lot of positive effects of this. This is why -- if you look at our R&D spend, both spend on hardware versus software, actually, our spending on software has been growing faster..

That's interesting. So that's like where -- I was going to ask you, where are you -- how do you think about allocating that R&D? It sounds like that software is huge.

It's definitely. Yes. It's definitely the fastest-growing part of our R&D spend.