Lattice Semiconductor Corporation (LSCC) Earnings Call Transcript & Summary

Great. Welcome back. I'm Joe Moore. You guys have seen me a lot of times. But -- and I do have to read this research disclosure again. But before, we are very happy to have with us the CEO of Lattice Semiconductor, Jim Anderson. So just to get this out of the way, for important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales rep. I didn't read that the first few times and somebody else...

We don't want to be in trouble.

Yes. So anyway, thank you so much for coming. I don't cover Lattice, but I definitely have been close to you guys over the years and have been impressed by your progress.

Maybe if you could just start out with kind of an overview of the strategy. You talk about kind of low power, low density FPGAs. What's -- what makes that a compelling area for you guys to...

Yes. Why do we like that market? Yes. So that is Lattice's historic market actually. So for decades, we've driven innovation in small FPGA. In fact, Lattice will hit our 40th anniversary this year. I'm a little bit further along than 40 years, right? But well, the company will hit its 40th anniversary. But if you look over those 4 decades, really all of the innovation that Lattice has driven over those 4 decades, really focused around, of course, FPGAs, adaptable, flexible technology. But we really specialize in power efficiency, ease of use, small physical form factor. And so that's where we've built up a tremendous amount of expertise. I believe, obviously, we're the world's absolute best at this type of technology in this type of market. And those products then get used across a number of really important markets, big growing secular growth markets, things like industrial automation and robotics, automotive electronics, all sorts of computing applications like data center servers, lots of different communications, infrastructure, applications, telecom, et cetera. So it's an area where we have, I think, the world's greatest expertise. We definitely in the world is now strongest product portfolio. I couldn't have said that 5 years ago when I joined the company. But hands down, we have incredibly strong product portfolio, and it's a market we love.

Yes. I mean it used to be, obviously, a lot of public companies in this domain and a lot of them have been acquired. I did the Actel IPO once in my career a long, long, long time ago, which I...

That [ date to you, Joe. ]

Yes, now have been acquired multiple times through the years. And then, of course, Altera, Xilinx, both having been acquired by microprocessor companies. What does that do for you? I mean it kind of -- you have this unique position of being the only one kind of solely focused on these markets in a market that it would seem to me kind of rewards that.

Yes. I think that -- so the competitive landscape definitely favorable for us. But I think if you look at, yes, the structure of the market, actually over those 4 decades, although a lot of start-up companies or even larger companies have tried to enter that market over that 40 years, really the 3 companies that originally kind of founded that space -- they were all founded back in the '80s, which was Lattice, Altera and Xilinx, are still primarily the providers of general-purpose FPGAs today. So it's a segment that has very, very high barriers to entry, and that's because of the hardware is unique, the software is unique and the interaction of the hardware and software, make the barriers to entry really, really high. And then within that space, we -- as I said, we specialize in that power efficiency, ease of use. We're the highest volume FPGA maker, we believe, by a long shot, and that's because of the part of the market that we're in, that technology gets diffused across so many applications, so many usage models.

Yes. I was -- also, in terms of [ dating myself, ] I was on the national semiconductor FPGA team, which I have to explain to people what national semiconductor is. And also, they never had made an FPGA, which is accurate, which is a testament to the barriers to entry that you're talking about. 2 years of attempting and then kind of never going anywhere. So a lot there. Maybe you could talk about the growth environment that you're in. I think you've had 11 straight quarters of growth. It seems a good market for everyone right now. We're seeing -- we're hearing good reports from FPGA businesses everywhere. Can you talk to what's driving that strength in an environment that has become so much happier?

Yes. I think -- in general, I think just FPG technology overall, there are some things going on across multiple end markets that are really beneficial to FPGA technology. If you look at almost all of our customers are trying to figure out, hey, how do I make my system more adaptable, more reconfigurable, more software programmable, because I want to design the hardware once and then be able to change that hardware over time. That -- obviously, that plays into FPGA strengths, right? So we're adaptable, we're reconfigurable. We help the customer future-proof their platforms. So I think in general, FPGA overall, there's trends that really favor the technology growth moving forward. And then if you look at Lattice, in particular, definitely, we've seen really strong growth, 28% growth last year that followed 26% growth the prior year. I think we see a lot of great opportunities to continue to drive that strong double-digit growth. Our major market segments have grown double digits multiple years in a row. And I think it's a combination of -- first of all, we're going through a lot of new product cycles -- our new small FPGA platform is still early in its ramp. And we just announced a new mid-range FPGA platform that doubles the addressable market of the company and creates an entirely new revenue stream. So look -- certainly look, we're proud of the growth that we've driven over the last few years, but we're way more excited about the growth opportunities for the company moving forward.

Great. Maybe we could talk a little bit to the product families that you've introduced. And before we get to the newest one, can you talk about the Nexus product that you introduced in 2019? What's -- that business seems to be going well. Can you talk about what's driven the successes there?

Yes. So as I mentioned, our stronghold or where we've always driven great innovation is in the small FPGA part of the market. So we launched Nexus in 2019. That's our newest platform for small FPGA. Really proud of the progress we've made on that platform. First of all, it's incredibly competitive, 2 to 3x better power efficiency than our competition, much smaller physical device size. So it brings incredible differentiation to our customers, which is really important. But beyond that, we've launched multiple different versions or product families based on the Nexus platform. We've actually launched 5 different device families to date, 4 are ramping into production. The fifth goes into production first half of this year. We've committed to 2 additional device families that will launch this year. And then more device families on the road map. So we're -- with each new device family, we're layering on a new Nexus revenue stream. So we're still early in the overall ramp of this product line. This is only the first or the third year of full year revenue growth for the platform. So we expect this platform to continue to grow for the next few years and probably further than that. So we're excited. We're seeing really good customer adoption of it.

And when you see the design wins in those markets, is that areas that are traditionally served by FPGAs? Or are you going into areas that have -- may compete with microcontrollers? Or is there a different competitive...

Yes, it's a great question. So I would say there's kind of 3 different parts of growth that we're seeing. If you take like the industrial market, for example, which for us, that market grew 41% last year. So definitely, we're displacing competition. So we're displacing competitor FPGAs in favor of Lattice FPGAs. That's one part of the growth. Second part is there's cases where we're displacing other semiconductor devices like microcontrollers, where customers are switching to our Lattice FPGAs because of some of the parallel processing benefits, especially for AI algorithms that we offer. And then the third is where we're adding just entirely new content where we're bringing new capabilities, new features that have never been there in that platform before. So that's really TAM expansion. So it's really a combination of all of those.

Okay. Great. And then can you talk about the Avant platform you mentioned?

Yes, definitely.

You recently launched it.

Yes. Yes. So Avant, we launched that in December. Avant is a totally new purpose-built FPGA platform for mid-range FPGAs. So Lattice again, historically has focused on small FPGAs. We're now extending our technology into midrange FPGA. So this will double the addressable market of the company. So it raises the addressable market from $3 billion to $6 billion, creates a totally new revenue stream for the company. So Avant is -- doesn't cannibalize in any way the existing products, it creates a brand-new fresh revenue stream. When we launched the platform in December, we shared live silicon demonstrations of its competitiveness. It's up to 2.5x better power efficiency than the competition. So that's not like 10% or 20% better, that's 2.5x better. That's a big deal to customers because power consumption is usually one of their primary design constraints. So 2.5x better power efficiency, twice the performance. And I think this statistic is really amazing. It's up to 6x smaller physical size, so 6x. That's a big deal. And so -- and then the other thing I would share is that the target customers for Avant -- and this is really beautiful. I love this is -- the target customers for Avant, if you look at them, 90% of them are already customers of Lattice today. So we don't have to go win a bunch of new customers. All we're doing is selling another product line to existing customers. And the software that they'll use with Avant, and Joe, you know how important software is in an FPGA. The software that they'll use is the same software that they're already using on our Nexus products today. So we've made the adoption, I think, really easy because same software customers that already know and love Lattice and are familiar with us. So we feel really good about the adoption of the product line moving forward.

Great. And can you give us an idea then of -- when you talk about midrange, how are you defining that? Whether it's [ gig ] count, dollar amount, things like that? And how much now reach do you have across the broader FPGA space? How much of it are you covering?

Yes. So if you look at -- so on the first part of your question, we roughly define small FPGAs as up to 100,000 logic cells. So logic cells is -- think of that as that's the fundamental capacity measure for the size of an FPGA -- so up to 100,000 logic cells being small FPGA and 100,000 to 500,000 being midrange. And ASPs would be -- for small FPGAs, ASPs would be like in the dollar or -- single dollars, up to tens of dollars. Midrange would be kind of tens of dollars up to hundreds of dollars. And then large FPGAs would be hundreds to thousands of dollars. And so where Lattice is now with Avant plan is the roughly $6 million of market opportunity in small and mid-range. Now the -- if you look at the volume of the FPGAs, that's the vast majority of volume of FPGAs. And if you look at one of our favorite markets is industrial applications, industrial electronics. Actually, most industrial customers use midrange and small FPGAs. They don't use a lot of large FPGAs. Large FPGAs are mostly for data center compute, for high-end communication systems. So for most of our customers, we can service almost their entire portfolio with the combination of Avant and Nexus. And that's why our customers are really excited about this. And it was actually the customers that pushed us to invest in Avant back in 2019 when we started developing it because with Avant, we can basically cover most or if not entirely their portfolio.

Okay. Great. And can you talk about the time frame to drive revenue through these...

Yes. Yes. That question comes up quite a bit, right? So, all right. So with Avant, we're modeling it basically off of the last platform that we launched. So we launched Nexus in 2019, that's well into its ramp. So we're basically modeling Avant revenue timing similar to Nexus. So that would put maybe a little bit of Avant revenue by the end of this year, but more significant contribution in 2024, and then it would ramp in the years after that as well. And Avant -- we will do the same thing in Avant that we did with Nexus, which is this is a platform where we'll bring out device families at regular beat rates over time, so there's a layering effect as each new device family is launched and enters into production, it creates a layering and a building effect of the revenue. And again, I'll stress that Avant, totally greenfield new revenue stream for the company. It's additive. We expect to continue to grow in small FPGA. We still have quite a bit of room to run in terms of the small FPGA market. If you look at our revenue relative to total market size, so we expect to continue to grow in small FPGA and layer this new growth factor on top.

So quite a bit of TAM expansion there. All right. Can you talk about competition, maybe starting with small scale and then with Avant? How do you see yourself positioned relative to Altera, Xilinx? And you mentioned some of the Avant characteristics relative to those. But can you just talk about how you're...

Yes. So if you look at -- so the competitive environment, I would say, is favorable. If you look at the competitiveness of our product line in small FPGA today, it is absolutely -- and any customer would tell you this and any employee that's been with the company for a long time, the product line is absolutely the strongest it's ever been in the company's history. And that wasn't the case 4 or 5 years ago. So we've got incredibly strong product line. Now we've always assumed that in our business planning and our road map planning that our 2 historic competitors, Altera and Xilinx, now part of bigger companies, that they'll continue to bring on new devices, that there will be robust competition. So we've always assumed that we build all of our road map plans assuming there's robust competition. And then look, if the competition doesn't show up, great, that's a tailwind. And if they do, we're ready, right? The product line is ready and the team is ready for that. So with the assumption of competition, we still -- when you look at the lineup that we have today with NEXUS, Avant coming, and the road map in front of us, it's the strongest product portfolio and road map that the company has ever had. And I think based on what customers have told us, we feel very, very well positioned competitively.

And do you see new product families supporting these markets from those guys? Particularly on the Nexus side of things, do you see an equivalent to that?

We've certainly assumed that our 2 main competitors will bring out new product refreshes. So we've assumed that. Now whether they do or not, you have to ask them, right? But in all of our road map planning, we've assumed that. But I think in general, if you look at those 2 main competitors, I think their primary R&D innovation focus is in a different part of the market now, right? I think it's more focused on data center compute, other applications mostly around large FPGAs.

Yes. Okay. Okay. You talked about software. Obviously, really important aspect of the FPGA business. Can you talk about where you guys are with your software capability and where you see that going?

Yes. This is actually a really important part of our strategy. And I think it's a part of our strategy that if you said what's the most underappreciated part of our story or strategy, this is probably the most underappreciated piece. I still feel like we're doing some education on this part of the strategy. But back about 5 years ago when myself and the new management team joined, one of the things that we did is put in place a new investment stream with respect to software. And so what we wanted to do is build out a portfolio of application-specific software solution stacks that basically make it really easy for a customer to design Lattice products in. So this is not the development environment. So the development environment software we've always had, every FPGA company has it. But this is higher level software that sits above the FPGA. And we've -- to date, we brought out 5 different solution stacks, each for different common usage models across our customer base. And as you can imagine, we'll continue to bring out new software stacks as well. Now the main goal of this software is, again, to make it really easy for a customer to design us in to either switch from a competitor's FPGA over to us or to switch from a microcontroller to us or even if they're -- the customers totally new to FPGAs, just really easy to -- for them to design us in and get to market quickly. It speeds up their time to revenue. It speeds up our time to revenue. Now there's side effect benefits of this as customers adopt this software, as you can imagine, it makes our solutions much more sticky over the long term. So it creates really good multigenerational stickiness. And so this is a key part of our strategy. And there's a couple of metrics around this. We track very carefully adoption of the software stacks. And if you go back 4 or 5 years ago, the adoption was 0, right? We were just starting. Now if you look at the, say, design wins over the last 12 to 18 months, we're now at an adoption rate of over 50%. So over half the time customers are using this new -- these new software stacks in their design. So that -- we view that as really positive. The other thing that we mention -- that we measure is when we do get a software attached to our silicon, are we seeing higher ASPs? And on average, we see higher ASPs. So we know that customers are valuing that, they were getting paid for that. And so this is an important part of our strategy. It's something that we're continuing to build out. But we really -- again, back to -- it's really about driving that faster time to revenue and ease of design in for our customers.

Yes. I mean I've talked to a number of people who do design work in the automotive industry and in hyperscale that have kind of said, we don't really know how to do this stuff. We don't -- and if we did, there would be a lot more opportunity for FPGAs. And like how do you guys -- you put resources into those types of design wins. Obviously, the software helps get you there. Like, is there a resource constraint where like if we had more people to put on this, we could grow faster?

Yes. And this is less about putting bodies on the problem. I mean that can solve it, but that's not scalable, right? So this is more about trying to provide a prebuilt software that our customers can just take off the shelf or they can at least take it as a starting point, right? So making even a little bit more self-service for our customers. But I'll give you a specific example, like let's use industrial customers. So we've got a lot of industrial and automation customers. And one of the things they were trying to do is add more inference capability to their systems. What they were trying to do is make their systems more autonomous, make their industrial systems, basically be able to make decisions on their own that those systems be more autonomous at the edge of the network because they can't always rely on a connection back to a big data center, right? There may be a failure in the connection point. So they want some level of autonomy and intelligence in the industrial equipment itself. So as we're working with those customers, what we're finding is, well, really what they're talking about is they're trying to add some level of artificial intelligence inference processing. So first of all, that's a great match for our products. FPGAs are just a naturally good match for inference processing because you can program them as a custom parallel processor and you can reconfigure them as your algorithm changes. But those industrial customers, they really had no idea how to add AI algorithms to their systems. So that's where this -- in fact, our first software package was around artificial intelligence inference processing at the edge, it's called SenseAI. And that's where that software package came in incredibly helpful because we gave that to the customers, that -- some customers took that as is right off the shelf and took that into production and some customers took that and customized that. And then where necessary for the big strategic customers, we'd definitely add some application engineers or people from Lattice to help them with that design, make it easy. So we'll certainly do that when we need to. But the idea of the software is to make it easy for customers to adopt the silicon even without the application engineering support, right?

Yes. This AI application for FPGAs is pretty exciting. And I guess, a lot of energy has been put into making that -- a data center hyperscale product. And a lot of the real-world inference is actually happening at the edge in industrial markets, in compute markets. So can you talk about your opportunities to broaden that out? And I guess as you -- the current wave of enthusiasm is around large language models and those types of inferences, is there an opportunity for Lattice to participate in the inference model?

Definitely, we already are. I would say there's kind of 2 buckets of opportunity. And let me come back to the inference at the edge. But the other bucket of opportunity is in the data center itself. So -- if you look at servers today, Lattice has -- there's Lattice silicon on almost every server that ships today. And in fact, our attach rate in servers is over 1x, which means on average -- if you look at total global server shipments, on average, servers are using more than one Lattice piece of silicon per server. And that attach rate has grown really steady over the last 5 years, and we expect that to continue to go up in the newest server generation in future as well. And we've also been adding more content -- in addition to the attach rate, just adding more content with each new generation of server. Definitely generative AI, the large language models, that's certainly going to drive tremendous demand for compute cycles in the data center, whether that's on a traditional CPU compute or graphics, GPU compute, or offload on very large FPGAs. So Lattice will naturally benefit from that as there's more servers, more accelerator plug-in cards where you would also find Lattice silicon there, more appliances that are built specifically for AI processing. That's absolutely will benefit from that increased growth of data center service. That's why I told my teenage son when he asked me, "Can I use ChatGPT to do my homework? "I said, "Absolutely. And tell all your friends to do their homework on it, do it, drives Lattice silicon demand." So -- I'm sure I'll get a call from his teacher. But so -- that will definitely benefit from the -- basically the data center build-out that generative AI will drive, right? But then the other -- at the other end of the spectrum is there will be inference processing -- more and more inference processing being done at the edge of the network, we're already participating in that today. In fact, that industrial example that I gave you, that's exactly what they're doing is they're adding inference technology at the edge of the network. And we see this across almost every different end market that we're in, automotive electronics, the industrial we already talked about, computing and client computing. We also see that, in fact, if you use the Lenovo ThinkPad X1 Carbon, there's a Lattice piece of silicon and software in there, that is doing artificial intelligence processing in that platform, which adds some really nice user security features. So we see that inference being -- that inference technology being added across multiple markets, and Lattice solutions are great -- again, a great solution for that because we're adaptable. We're reconfigurable over time. People know their AI algorithms are going to change over time and then they just reprogram the FPGA.

Great. So I do want to leave time for questions from the audience. I would maybe ask one more end market question first. Comm infrastructure, can you talk about the opportunities for Lattice...

You said comms?

Communications.

Yes. So today, where you would find Lattice and communications infrastructure is usually in the control plane. So like 5G wireless equipment, we have a great position in 5G wireless equipment. In fact, in a 5G base station, we have about 30% more content than in a 4G base station. But today, mostly in control plane. Also in data center networking, similar, mostly in control plane. The great thing about Avant is it expands our ability to address even more control plane applications, but Avant will also address data plane applications. So in communications infrastructure, data center networking, we'll start to see Avant get into new data plane applications for us. So we see that as another big growth area for us.

Yes. Great. Let me pause there and see if we have questions from the audience.

Just on market share, where do you think the market share can get to in the low-power range and then mid-power range? Would you expect to be able to get to a similar level of market share? And just on the barriers to doing that, is it down to more product families coming out, more software stacks or just waiting for products -- for the customers to be refreshed?

Yes. I've always told our sales team until we're at 100% market share, we're not done. So -- but it's actually -- it's actually difficult for us to measure exactly what our market share is because as Joe mentioned, our 2 traditional competitors are now part of bigger companies. I don't really break it out separately. So -- and definitely that revenue is also not broken out by the small, mid, large. It's hard for us to estimate market share. But we definitely feel in the small FPGA space, we certainly feel like we've got a very good chunk of that market today. But I think it's probably still less than 50% of that market. So we still have a lot of headroom for additional share gain in that market. And I think our growth in that market will be driven by not just share gain, but there's also places where, like I said before, we're adding new content that hasn't existed before. So that's TAM expansion. So I think our growth in small FPGAs is a combination of share gain, but also TAM expansion through either new usage models or where we're displacing say, a microcontroller in favor of an FPGA. So I think those are the growth drivers for us in small FPGA. And in midrange, yes, I see no reason why we can't grow significantly in midrange. We're -- if you look at the competitiveness of the Avant platform, when we launched it in December, we did live silicon demonstrations of the competitiveness, very, very competitive platform. We're selling this to existing customers with software they already know. So -- Yes, I think our opportunity to grow the Avant revenue stream over time is very, very strong. So we certainly -- actually, every dollar of Avant is, by definition, share gain. So -- because we don't have any position there today, right? So -- but we feel very good about the long-term growth of Avant in midrange.

There's such a long tail on these designs that market share is probably harder to measure than like design win activity and things like that. But it seems like you have a pretty high market share when it comes to new design wins in those smaller scale?

Yes. And we -- you're right. The -- actually, the fundamental limiter usually to us is, of course, it's the new products that we're bringing out in midrange, but the bigger limiter is just the rate at which our customers redesign their platforms. So the socket will -- or the opportunity opens up when the customer redesigns their next system. So that's usually the main limiter to our growth trajectory.

We have time for one more quick question if there is one.

Yes. Maybe on the auto part of the business, replacing medium microcontrollers, where do you see your sweet spot there? What are you replacing? How to think about the attach rate there?

Yes. So in general, in auto, that's been a really good growth area for us, but it is a smaller part of our revenue today. So if you look at industrial, and automotive electronics. Last year, it grew 41% year-over-year. So it was a really good growth area for us. Actually, that's the third year in a row that, that segment has grown double digits overall. But within industrial and automotive electronics, auto is the smaller piece of it, but it actually grew more than -- higher than 41% last year. So it's the smaller but faster-growing piece of it. And where we're seeing growth is replacing traditional FPGA competitors, some growth coming from replacing microcontrollers and then also -- some growth also from basically new capabilities that didn't really exist on those automotive platforms. I think that, in general, where we've seen -- if your question is specifically about replacing microcontrollers, where we've seen that happen is where there's a customer that, let's say, they historically used microcontrollers but they're now trying to add more -- actually inference processing is a great example. They're trying to add more inference processing. Inference algorithms are inherently parallel algorithms. Microcontrollers are sequential processors. They don't generally run well on microcontrollers. And so we've had customers that are looking for, hey, how do I take the next leap forward in terms of the performance of my system? And we've showed them that you can take an FPGA and you can program it to be basically a custom AI processor for your specific inference algorithm. And you can see an order of magnitude performance per watt speed up versus trying to run it on a microcontroller. And the other great thing is those customers know that their algorithm is going to change over time. No problem. They don't have to change the hardware. They just reprogram the FPGA. So that's an example of where we've typically seen us displacing microcontrollers.

Okay. Well, thanks for the good questions. Unfortunately, we're out of time. Jim, all right?

Thanks, Joe. Appreciate it. Thank you.