Astera Labs, Inc. (ALAB) Earnings Call Transcript & Summary

Thanks, everyone, for joining. I want to welcome everyone to RBC's TIMT Conference. I'm Srini Pajjuri, I cover semis. Today, we're pleased to have Astera Labs. We have Nick Aberle, who heads the IR team there. So Nick, thanks for joining us. We have roughly 30 minutes. We're going to leave a few minutes for Q&A at the end as well. So Nick, well, it's been a little more than a year since you guys went public, very successful IPR. Maybe we can start with when you went public, I think you were targeting a high-growth market, roughly, I don't know, $10 billion-ish TAM. Things have changed quite a bit since the market has broaden and you expanded into new markets as well. So how are you thinking about in terms of the addressable market, what's -- how to think about your SAM and TAM and growth potential?

Yes. Got it. Yes. Well, thanks for inviting us as well. Srini, I appreciate it, and thanks for all the investor support. Yes, it's almost been almost 2 years. It will be 2 years in March. So at the time has flown by for sure. And a lot has changed, yes. So the initial market opportunity that we outlined around the IPO was around $7 billion, and that's grown substantially just over the course of the last, call it, 18 months or so. And the market opportunity for us is really a function of 2 big pieces. I mean, there's clearly a secular trend and tailwind that's at our back associated with just the build-out of cloud and AI infrastructure and new accelerators coming on the market. And obviously, the goal of Astera is to staple as many dollars as we can to each one of those accelerators shipping into the market. But we've been able to kind of further enhance the market opportunity on a company-specific basis, by bringing out new products, developing new technologies, expanding the road map and just broadening the overall portfolio. So if you look at where we stand today, there's the initial pieces of the business that we outlined during the IPO, the signal conditioning, pieces of the business, which entail the retimer opportunities, both for PCI Express and for Ethernet. We've outlined is about a $3 billion TAM or served addressable market because we serve that today. And both of those are poised to grow quite nicely over the next couple of years, both pieces of the business are going to be kind of graduating to the next level, going to PCI Express Gen6 for the PCIe side, ultimately, Gen7 as well, where the speeds will double again. And then same thing on the Ethernet side of the house as well. We're primarily shipping 50 gig per lane solutions today that will expand the 100-gig per lane starting next year for us and then obviously, 200 gig per land beyond that. So some pretty big drivers in that business where you see adoption especially on the Ethernet side, fairly limited today in corner cases, but much more adoption and attach expected as we move to these faster speeds. Leo, obviously, a big market opportunity as well. That business is still small for us today from a revenue standpoint. But I think we put out a press release last night talking about new engagement with Microsoft there. So hopefully, that those will start to drop some proof points that something starting to brew on that front. When you start to think about the market opportunity there in the entire CPU landscape of, ultimately, every CPU is CXL capable how many of those we can enable that market opportunity starts to be quite large as well. But it was at OCP last year where we launched the Scorpio product line, which boosted the TAM even further, made up of a couple of different pieces, and we'll talk about each of those pieces individually in a little bit. But the P-Series, we're calling it on or about $2.5 billion of incremental opportunity for head note connectivity. And then the X-Series, which is for scale-up clustering of accelerators we called that at the time at being around $2.5 billion to $3 billion market opportunity. And then most recently, talking about UAL being incremental on top of those kind of PCI Express or PC Express like protocols that X-Series addresses adding another couple of billion dollars on top. So yes, all folded in kind of looking at the $14 billion, $15 billion served addressable market today. Clearly, we're not sitting still either. We're developing and pushing forward, seeing some of the announcements on the optical side. So we would expect to continue to grow that TAM going forward.

That's great. So your largest business, the retimers, the chip that sits between CPU and GPU and sometimes CPU storage, et cetera. There were some concerns, I guess, a lot of concerns as the market moved from hopper generation to Blackwell, and that still continues to grow to this day, I think last quarter, it was up for this year looks like it's going to be up like 60% to 70%. And your expectation is for that business to continue to grow into next year.

That's correct. Yes. So Aries is still a big -- very big piece of our business. Yes. So if you look back 4, 6 quarters ago, it was over 90% of the total revenue. Today, it's closer to 2/3. And all the while, it's grown, yes, we'll grow close to 70% this year on the Aries business. So quite healthy growth, but also diversified kind of the business across the other product lines, just because they're growing very fast as well. So yes, I mean, there's a little bit of skepticism coming into the year about the transition to black wallet NVIDIA. Obviously, we had a big presence there with Hopper and you started to see some of the reference platforms that NVIDIA goes to market with on the rack scale side, use less retimer content, but we've also picked up a meaningful amount of retimer content in other areas. We've ramped aggressively on the custom ASIC accelerator market side. And we've also started to penetrate the scale-up opportunity as well for PCI Express retimers. And as these racks grow and clusters grow in size, distances become more challenging for these PC Express signals to traverse with high fidelity. So you start to need PCA Express retimers either chip down on a board or at the end of AEC cables. So that's been another big driver of our business. So yes, I mean, to your point, we expect that business to continue growing into next year. PCI Xpress Gen 6 transition is just underway. Blackwell is really the only endpoint today that supports Gen6. So we'll see a whole slew of incremental accelerators to start to adopt PCI Express Gen 6 starting next year going into '27. That will drive higher attach for retimers but also a higher ASP for us as well. I think on or about 20%, plus or minus, ASP lift on a generation-over-generation basis. So overall, very excited about the Aries business and the opportunity there.

That's great. I think this market took off of Gen AI with the Hopper generation. And then obviously, they're attracted some big competitors, right? I think Broadcom announced some products and Marvell announced some products. It looks like you have that visibility into next year in terms of your design win momentum also you're talking about ASP premium going forward?

Yes, that's exactly right. Yes. So I mean, just like any other big and growing market, you would expect competition, so not super surprising from that standpoint. And we respect all the guys that are coming into space. They're very well-known, established names in the networking arena. But yes, I mean, we have a substantial foothold in this business today. We're in the incumbent position. We're the only guy shipping PCI Express Gen6 solutions into the market today. We talked about PCI Express Gen6 making up a little over 20% of our business in total in Q3. And then to our knowledge, nobody else is shipping anything in volume today. So the head start is there. The learnings and kind of the gauntlet that we've gone through on Gen5 has built a lot of muscle for the organization in terms of understanding the nuances and the iterations that need to be made to kind of handle all these different workloads and interop between a wide variety of endpoints. So we feel very good about our position there.

Got it. So should we kind of think of retimer growth going forward kind of primarily associated with the ASIC side of the things? Or do you think there is actually more growth left in the GPU market as well?

Well, I think it's going to be both. So we've seen -- so the Gen 6 business that we're shipping today is shipping into NVL72-based platform. So that is a growing business. We would expect that to continue to expand next year. But to your point, I mean, I would say that the XPU side or the custom ASIC accelerator side is probably a little bit more fertile market opportunity for us because we can play on that scale outside in the peripheral connectivity side, in addition to the scale-up clustering side as well. So it's just a bigger dollar content per accelerate opportunity.

Got it. And then in terms of your visibility, I think the systems are designed well in advance, right, like, I don't know, 18 to 24 months. Is that the sort of visibility that you have in terms of design win moment?

Yes. I mean it's on or about 12 to 18 months, I think it's a good mark for sure. Yes. I mean the stuff that we'll ship for revenue and that will drive growth next year for us has been largely baked for quite a while. The things that we're focused on developing and we working with their customers on to launch is really a next '27, '28 at this point. So yes, good visibility on that front in terms of the design wins, the engagements, expectations for what forecasting and demand and units could look like. Obviously, not a perfect crystal ball, but pretty good about the trajectory.

Okay. Great. Just switching gears to the PCIe switch side. You have been shipping the Scorpio for scale-out. I think that seems to be ramping quite well. So let's talk about that. How do you see the opportunity there? How big is the SAM? And how many designs do you have? And how should we think about growth as we go into next year?

Yes. I mean it's clearly a large market opportunity. It's a multibillion-dollar SAM from our perspective. And when you think about how to build that market to opportunity out, you're really looking at all the accelerators that are shipping into the market. PCI expresses the need of language within an AI rack. We're an AI server and CPUs, GPUs, networking elements, storage, memory all needs to interface with each other over PCI Express. So there is a huge opportunity there, and we talked about some of the retiming opportunities for signal conditioning, but also for switching as well. And PC Express Gen6 is our kind of first foray into that market. Broadcom had a big play there for Gen5 and drove substantial revenue there. We were the first to market with the Gen6 switch, and we're the first ones to ship into volume on Gen6 today. So we do feel like we have a decent lead with a very optimized solution for that market opportunity. And when you think about it, there will be opportunities on the GPU side like we're seeing today on the NVL72, but there will be opportunities on custom ASIC businesses as well. We talked about on the call that we've picked up new design wins with incremental customers outside of our lead customer that we're shipping to today that will start to layer into our growth profile in 2026. So yes, I think the P-Series business is very healthy. It's grown very rapidly this year and is poised to continue to go -- grow even further next year.

Got it. Then Nick, one of the questions that we always get is that PCI is not a new market, right? There's a lot of players in this market. And what is that you bring to the table that gives you guys this competitive advantage especially on the retimer side, how do you sustain that on also and you're making progress on the switch side. What is that -- what are the...

Yes. I mean, so -- totally right, not a new market, not a new standard. And I think some of this was identifying early on the trend within AI servers and this heterogeneous compute environment, we're starting to use CPUs and GPUs within the same system. And by default, PCI Express became very important within AI servers I think the right approach that was taken by the team, if you look back 3, 4 years ago was clearly the hardware is important the capability of being able to condition the signal is important, having good SerDes is important, but focusing on how to add value specifically for fleet managers and hyperscalers trying to optimize productivity within the data center was really the key that set it off. So there is a very intense focus on building a software team and a software stack that could be provided to customers in order to proactively analyze what is happening within the data center. If you think about them hyperscalers spending tens of billions of dollars on CapEx and having hundreds and hundreds of thousands of AI servers within a data center that are having millions and millions of PCI Express links running through them. having knowledge of exactly what's happening on those links and what's happening on each of those endpoints could be a very valuable thing. So we put sensors into the product. We developed APIs that we provided to our customers. And those customers then take those APIs, build it into their operating stacks and can more effectively manage their fleets. So they can identify if there is a problem in any specific area within the link or between endpoints within the infrastructure that they can go in and fix something proactively. And at the end of the day, again, where they're spending tens of billions of dollars in CapEx, they want to utilize that and make it as productive as possible. And so if we can help them make it just a little bit more productive than our solutions end up being free for them.

Got it. Got it. Then switching gears to Scorpio X, the Switch product for scale up. Obviously, you've been pretty bullish about the opportunity. It's a large market, but also there is a healthy skepticism about the longer-term sustainability, but it hasn't even started to ramp it -- let's start toward the ramp and how do you see into 2026 in terms of design wins and the momentum.

Yes. I mean we're super excited about the prospects and opportunities around X-Series. And if you think about it, I mean, rack scale infrastructure is just really in the early stages deployment NVIDIA has been the first guy to kind of bring rack scale infrastructure, AI infrastructure to market at scale. You've had a couple of others just starting to do similar things over the course of the last couple of quarters. So just the market opportunity and that secular trend is very early in the game right now. So X-Series was built in order to enable clustering of accelerators or clustering of GPUs and a kind of back-end scale-up topology that allows each of those accelerators to speak with each other and create one cohesive compute unit. And we've seen initial traction with customers wanting to leverage PCI Express to make that happen. That PCI Express is a very low latency standard that allows you to do kind of memory semantics and optimize those back-end scale-up connections. So we're in preproduction today. So we'll ship a little bit -- shift a little bit of it in Q3, a little bit more in Q4. We hope to go into volume production early half of next year. And we have multiple customers, again, multiple platforms that are poised to start kind of pulling and demanding X-Series solutions starting next year. So very excited about the market opportunity there. And again, it's still in its very early stages of expanding. So we're going to see, just on PCI Express alone, at least the next 3 to 4 years, we have programs and designs that will ramp into volume production, and there's new ones that are still getting converted that we talked about a pipeline of over 10 unique customers looking to adopt and leverage PCI Express for scale up, of which only a handful of those are actual design wins today. So there could be incremental conversions that even push that out even further. So very excited about that. I'm sure you'll ask about UALink next, but that would be the next evolution of what we are trying to provide on the scale-up side.

Yes. We'll get to you in a bit. Just want to understand, as you go from retimers to Scorpio P and then Scorpio X. Just kind of talk to us about your content per XPU because I think that's a pretty significant increase that you're seeing.

Yes. So yes, I mean, we're fortunate, like I mentioned earlier, to be kind of pointed in the right direction from a secular standpoint. And we've kind of cut our teeth with the retimer business and then engaged very tightly with some of the kind of biggest hyperscalers in the world, the biggest merchant GPU vendors in the world. There's not too many boxes or AI servers that have been shipped into the market over the last couple of years that don't have some type of Astera content in them. So that's afforded us some very tight relationships and a sneak peek at exactly what things are coming down the pipe. So I think that -- when we think about the growth that's available to us by just expanding the content and again, earlier talking about stapling as many dollars as we can to each accelerator going out the door, it really started with Aries which probably allowed us $50 to $100 worth of content per accelerator roughly introduced Taurus. If you add both of those solutions within a certain platform that kind of gets you up into kind of $100, $150 worth of content per accelerator. Scorpio is really the game changer where you started to see multiple hundreds of dollars of incremental opportunity per accelerator and then X Series kind of kicks up the next the next notch where it's going to be a very high value, high protein anchor socket that's within the platform. So if you have the same of the Astera content across the entire platform, probably get pretty close to $1,000 of content per accelerator. And like I mentioned earlier, we're not stopping there, UAL will very likely take it above that number.

Okay. Got it. And obviously, you guys made a comment that Scorpio could be biggest product. sometime next year. Retimer is obviously much bigger now. So that kind of, I guess, tells us how confident you are about be in?

Yes. I mean with all the caveats and disclaimers that we always post, I mean, this very, very driven by our customers and their deployments, the timing of their on and the pace of their deployments. So certainly a possibility that we could see crossover late next year, if not, then certainly in early part of 2027. But yes, I mean, it is a function of that dollar content opportunity when you're able to kind of attach those kind of higher-value solutions to these accelerators, then you can grow revenue pretty meaningfully and expand your opportunity.

Got it. Got it. And then UALink, there's a lot of debate about. So as we go from Scorpio X, which is PCI Express-based UALink I would imagine that there is additional content opportunity, right? So where are we in that in terms of the standards. And actually, maybe you can talk about what does UALink bring to the table that PCI doesn't have today? And what do you -- how do you see the timeline of that ramp?

Yes. So if you look at the landscape today of what customers or GPU guys or accelerator manufacturers, AI provider platform providers are looking to leverage to scale up within Iraq, multiple accelerators. You have a couple of options. You have PCI Express, you have standard Ethernet, you have some proprietary protocols out there as well. And then most recently, NVLink Fusion has been entered into the mix. So you have really those -- that set of opportunities. PCI Express and Standard Ethernet, as you know, have been around for some time, decades. So there are by no means optimized for AI workloads and AI applications, but it's what we have today and customers will that, and we'll help them to make those efficient as possible. Looking out over the long term and seeing the opportunity in that secular trend moving to kind of a rack scale architecture. There is a big need for something that's going to be more purpose-built for AI. So the consortium, the UAL consortium was put together, I guess, late last year to solve that exact problem to kind of bring a standardized, open ecosystem approach to scale up networking and to provide solutions that are purpose-built for that application. The standard was ratified. The 1.0 standard was ratified back in March of this year. And ever since that standard was ratified. You've started to see development happen where clearly hustling quite quickly to develop and build out a portfolio of solutions. It won't just be a switch. It will be a switch. It would be signal conditioning capabilities cabling. It could be chiplets, there could be a whole variety of opportunities there from a portfolio standpoint to pick up business on that side of the house. So all that is tracking well. Our expectation is that we'll be able to put products, UAL products in customers' hands at the end of next year, back half of next year. for system qualifications, initial bring-ups and deployments in 2027. So that's the expectation today. I mean we continue to grind along and build these things up we're hoping for others to do the same to really kind of promote this openness of the ecosystem to provide customers and hyperscalers with an opportunity to have a broad and diverse set of suppliers.

I guess the debate is about UALink versus -- UALink is somewhat new, whereas Ethernet has been around forever. So Broadcom is pushing Ethernet, and it looks like they got pretty good industry-wide support, just like UALink. So I guess, how can investors get conviction that UALink opportunity is going to be as big as you think longer term, given the contribution from Ethernet?

Yes. I mean I think just -- just to provide some context at a high level to start out. I mean, clearly, this is a huge market opportunity. So we've sized it out to be tens of billions of dollars when you think about the entire universe of accelerators and the connectivity content that's going to be needed on the back end to scale and cluster all of these accelerators together. And that's across NVLink, Ethernet PC Express, UAL, eSUN whatever else comes to the market, that is the pie that everybody is trying to grab. And we've always said that there will be a role for PCIe, UAL, Ethernet, NVLink, NVLink fusions. All these things will play into that mix. and customers will choose based upon which standard on a multigenerational basis, fits best with their topology, their architecture, their accelerators, what they're trying to accomplish within their data centers and within their infrastructure. So we see a place for all of these standards. Like I mentioned, UAL kind of set out on the path with taking a blank sheet of paper and figuring out, hey, what are the top 10 things we need within an AI scale up topology and everything else is kind of thrown in the recycle bin you want a highly optimized solution that has no overhead and is just focused on that application specifically. So that is the incremental evolution from like PCI Express to Standard Ethernet today is serviceable, but it does have a lot of that overhead. So I believe at first SUE, then eSUN has been the natural evolution for Ethernet to say, hey, we realize that this is not an optimal solution for this market or these applications, so we need to provide a purpose-built solution for this market. So it's very much going down that same path that you all identified a little over a year ago and kind of set course to kind of go solve. So I think it's a step in the right direction for sure. And we'll see how everything plays out. Like I said, I think there will be customers that adopt different standards. And we -- like we've said on our calls and previously, we have multiple hyperscalers, multiple kind of customers within that group of 10 plus PCI Express engagements that are looking to develop UAL based solutions and are pushing us in that direction. So we feel very confident that based upon those kind of pushes in those road maps that we're headed in the right direction and there are going to be a meaningful market opportunity. So when I talk about multibillion-dollar UAL opportunity kind of as a stand-alone, I mean, it's really a fraction of that total tens of billions of dollars of market opportunity. Hopefully, that's actually conservative and it's much bigger than that.

Yes, makes sense. We're not going to know for a while, right? I mean, by the time UALink is shipping and by the time Ethernet is shipping is probably '27, '28. But let me ask you, I mean, is there anything preventing you from, let's say, hypothetically, Ethernet is the way to go. Is there anything preventing you from entrant that the Ethernet market? I mean you have the skill set, right, today?

Yes. We have the skill set. And as I mentioned, I mean, it's a step in the right direction, right? We're all about open ecosystems, open standards I think it's a great step in the right direction for Ethernet to go down this path and open things up and create a level playing field for all the reasons that UAL has already done for collaboration for a diverse supply chain. So yes, I mean -- so it's starting to make a little bit more sense. I mean it's still very early. Like you mentioned, it's a work stream at OCP right now. They'll set forth on building a spec or a standard that will be released at some point. And if and when it makes sense for us to do that, and we hear customers telling us that that's something that they want us to pursue that there's an opportunity there, then it's something we would definitely consider.

Yes, makes sense. I want to see if there are any questions from the audience?

So one of the things I kind of think I learned here was the AI at the edge seems to be something that one can discern from vague press releases by some of the large hyperscalers who have consumer end markets as well is kind of probably in the next 18 months to 2 years. All of it is going to be CPU. The thing I think it was kind of new to me is that the CPU at the edge probably hands off to a CPU into the data center. And there's -- really little architecturally, it's better to speak CPU to CPU on many of these devices rather than handoff to a GPU or something else? Am I underestimating Leo and CXL opportunity? Could this be a significantly larger opportunity as AI pushes to the edge and it's more of a CPU to CPU type of market.

Yes. So yes, I mean, Leo has applications within general purpose servers and it's got applications within AI. I think both are very viable and make a lot of sense in terms of return on investment for customers. There are a variety of different customers and designs that we've either already won or that we are chasing that kind of fit into some of the buckets that you described. So it's -- for us, it's -- we expected this market to already develop a little quicker than it has. So we continue to want to be a bit conservative on kind of disseminating exactly how that curve or that growth might layer into the model. We expect it initially for it to happen this year and it has not. So we're confident that we will see some incremental growth from LEO and CXL next year across kind of both of those applications. As I referenced earlier, a nice press release this yesterday after the market closed, talking about our relationship with Microsoft to start bringing CXL into the market, leveraging and supporting the SAP HANA kind of large database applications. So I would say stay tuned. We're optimistic that it will continue to develop and drive some growth. But we want to just be a little bit tempered just given that it's taken a little longer to date. Not necessarily, yes. I mean, so I mean it could a little bit. So I would say that when we built out that TAM, we basically looked at the entire universe that ship annually. And we assume that close to 100% of those CPUs are CXL capable by, I think it was like 2028. So that's what yielded the $4 billion market opportunity. And again, we're using, I would say, somewhat conservative kind of attach rates and ASPs and those types of assumptions. And then -- but when you talk about the immediate market opportunity, it's like, okay, we have all these capable CPUs out there, CXL capable CPUs, how many will actually enable CXL. And that's obviously going to start at a much lower level and then hopefully grow as adoption takes place over time. So I don't know that it materially changes it. That's already a pretty giant number for us to go prosecute. And to the extent that we can even grab 5% to 10% of that over the next couple of years, that would be a pretty nice win for us, I would say.

Great. That's all the time we have. Thanks, everyone, for joining.

All right. Thanks.