Credo Technology Group Holding Ltd (CRDO) Earnings Call Transcript & Summary

Good morning, everyone. I'm Toshiya Hari. I cover the semiconductor space at Goldman Sachs. Very honored to have the team from Credo with us this morning. We have Bill Brennan, CEO in the center. We've got Dan Fleming, CFO as well. Bill and Dan, thank you so much for coming. .

Absolutely. Thanks for having us.

I've got a bunch of longer-term questions, but I did want to start relatively short term. You're fresh off of your earnings call from last week. If you can kind of recap key highlights in the quarter. Did anything surprise you? You're guiding October revenue up 11%, I believe, at the midpoint. Where are some of the drivers that went into that outlook for October?

Yes, I think generally, I'm really happy with the execution of the team. I think one of the big focuses for us over the last several quarters has been product ramp, and we saw real results in the last quarter. So that's reflective of strong execution across the board from our team. The -- so no surprises there. And I think it's part of the theme that we've been talking about, the narrative that we've really talked about for this year. . So going forward, I think we feel good. I think we're -- we feel very good about our largest customer ramping strongly. We see that continuing throughout the year. I think generally, if we look kind of across our business, we see strength coming from really every area from a product standpoint. So feeling good about the upcoming quarter as well as the balance of the year. .

That's great. Last February, you surprised us all in....

In what sense?

I don't want to rub it in.

You mean 2 Februaries ago.

That's correct. Not this February, but the last February.

Got you.

Since then, to your point, you've executed really well, both in terms of delivering results and guiding to, if not above Street consensus. Is the improved visibility a function of environment being favorable? Or do you feel like you have a deeper look into what customers are working on? Are they more open with their plans, their forecasts, et cetera? Has something permanently changed? Or is this sort of you guys benefiting from a more favorable backdrop or both?

Sure. So just to go back and touch on 2 Februaries ago. It's really important to talk about what happened because we have a full view of what happened at this point. The big shift was the shift towards AI. Our largest customer, Microsoft was the first mover. And I couldn't really describe 2 Februaries ago as to what the shift was that we saw that caused us to talk about a much lower expectation for that largest customer at the time. But it's super important from a theme standpoint because we've been involved in different AI developments for going on more than 5 years. Tesla, for example. We've been working on the Dojo supercomputer for more than 5 years. So we've always known that AI is going to be wind in our sales. It's really going to really propel our growth. And that was an inflection point that happened. And at first, it looked like it's super tough for us. But very quickly, we recognize that many, many programs were going to be pulled forward. And when we think about AI clusters and the difference there is that the back-end network is critical to the performance of the cluster. So you look at the density of connections and you look at the speed of connections and that moved the entire market all the way to the front, all the way the leading-edge technology, where we do very well. And so I think from that moment, really probably 3 or 4 months after that, it was very clear that there's this massive pull from the market for higher speed connectivity and more connectivity associated with AI clusters. And we've seen nothing but that play out over the last several quarters, and we see it continuing to play out over the next several quarters. So that's definitely fueling our optimism and the activity that we see that's going to lead to continuing to grow our revenue in the future.

Okay. Got it. That's helpful. A couple of questions on AEC, which obviously is the largest business you guys run today. For starters, how big is the AEC market today, according to what you see? What percentage of the TAM have you been able to penetrate? I think on last week's call, you guys talked about AECs becoming sort of the de facto standard for inter-rack connectivity? What about other applications? What sort of the catalyst to get you going on, on some of those newer applications, if you will?

Yes. The first part of the question is how big is the market. And I think we're very much in the early innings of the market developing. If you look at where is the market coming from? Typically, short in-rack connections have been made with passive copper cables. And so because of the speed of the connection and really the physics involved with communicating over copper, there has always been a feeling in the market that eventually, there will be a move away from passive copper cables and you have to go to an active solution. For a long time, the market was thinking they were assuming that it was going to be an optical transition. And that's where we pioneered this market in a sense, let's keep the connection electrical on copper. You can use much thinner gauge wires and the signal integrity no longer becomes an issue. And so that market is growing -- our market is growing as the transition away from passive copper is happening. And so we're early stages. It's hard for me to gauge what the market size is. We're very near 100% of the market. The market forecasters see this market being a multibillion-dollar market long term. And so what we're really focused on is converting one customer at a time. When you say the de facto for inter-rack, yes, we've mentioned that. So especially as speeds go to 100 gig per lane, using passive copper really is not an option. And so AEC is becoming a de facto standard in a sense that we have customers that now call us when they hit that point where they need to find another solution. It's important to note that with optical, there's also trade-offs. And they're not necessarily favorable, not just from a cost or power standpoint, but also from an MTBF standpoint, when you're talking about mission-critical connectivity. Looking at an AI cluster, it's interesting because if one, if one connection within the cluster fails, just intermittently, the industry refers to this as a link flap. We have one connection that needs to be reset, it will take the entire cluster down. And so if you look at an AEC, link flap is not an issue. You just don't see those fares. It's really related to the optics. And so this is another thing that heavily favors AEC connections where they can be made. On the call, I mentioned that we're seeing more activity around rack-to-rack connections. So meter cable can span 2 racks, 7-meter cable can span 3. And so as people are looking at -- now that the industry has had more than a year of deploying AI clusters, there are a group of people that look at failures that are happening within the clusters and talk about how do they address those failures over time. And this has been really recently fueling conversations about how to rearchitect and in the industry, really recently, we've seen super impressive things from a power sourcing and cooling perspective that enables people to do more in a smaller footprint. And so making these 2 rack or 3 rack span connections is becoming possible. And if you can do that, then you can reduce the number of link flaps, you're going to improve the quality of the entire cluster and ultimately, the efficiency across the board. So that's kind of a new thread that leads us to be optimistic about the AECs. .

Got it. And Bill, you talked about converting one customer at a time.

Right.

You had a big #1 customer, the #2 customer ramping very aggressively. You've talked about #3 and beyond. How should we think about sort of the dollar opportunity, dollar content opportunity evolving as you go from one customer to the next because not all data centers obviously are created equal, right?

I think a good way to look at that is to look at the application within a given customer. The first customer that we ramped was really connections were being made from servers to the front-end network. And this is traditionally what we all thought about as the network. Now that AI clusters are being deployed in a big way, the real large opportunity is the back-end network. And within the back end, there's 2 opportunities, there's a scale-out network and the scale-up network. And so I think if you look at dollar content, you can -- it's a hard question to answer. You've got to be specific about are we comparing apples-to-apples. But as we ramp into AI clusters, the TAM increases for sure, and it's -- the effect is twofold. The number of connections is significantly larger, but all speed of the connections are faster. And so there's a content difference whether you're selling 100-gig cable or an 800-gig cable. Naturally, the ASPs for an 800-gig cable are going to be higher, but also the effect is there's many more connections. So that's adding to the dollar content for a given customer. .

Okay. Got it. That's really helpful. The advantages of AEC is pretty clear. Data rates are increasing. We had [ Jensen ] this morning. He talked about the densification of data centers and all that stuff is really good for you. You've got to reduce power. To the extent you've had episodes where you engage with the customer, but ultimately, they decide not to transition to AECs and maybe there hasn't been any situation. But to the extent there have been, what was the pushback? Is it passive copper cable continuing to improve? Is it optical? What's sort of getting in the way?

Yes. So we've never had a situation where somebody has considered an AEC and then gone back to passive. And I can say that comfortably as we talk about 100-gig per lane. It's not a conversation. I would say that the only time that people have looked at it and gone a different direction, it's purely based on an architectural approach. And there are deployments where you'll have racks of appliances and no switching within the rack, where the switch has happened at the Leaf-Spine. And if you eliminate top of rack, there isn't a need for in-rack cabling. And so that's when people have looked at different architectures and chosen to go with a [ torless ] cluster architecture, that's where the opportunity hasn't ultimately materialized. But we haven't seen a situation where a customer is considering passive AEC and something optical, every single time he has landed on AEC.

Okay. Got it. Maybe on customer diversification. I'm sure you guys hear this directly from investors a lot as well. The concentration is something that some investors struggle with. I think it's really a product of your success in sort of pioneering the....

A very smart way of looking at it. .

But the question that we get from investors is when do we start to see a Credo that is more diversified? I think you talked a little bit about Tier 2 or, I guess, emerging CSPs. Maybe talk about the forward outlook from a customer profile standpoint? .

Sure. Sure. So I mean talking about our largest customer first, I feel very good about where we are. I think we're in the early innings of ramping that. And I mentioned that I don't expect it to be linear. There's a surge. But ultimately, if we look at the several quarter outlook, we see that increasing. So they'll be our largest customer this year. There's no question about it. And so measuring from a concentration standpoint, yes, you'll be able to point to that. But I also feel very comfortable about how the company is executing across our product lines and really with customers. So I expect that the first large customer we talked about, they'll return to historic levels in the -- in, say, the out quarters and going into fiscal '26. I mentioned the third hyperscaler that we're engaged with. My confidence is growing that we're going to see a nice ramp in fiscal '26. We may catch a little bit at the end of this fiscal year. But the confidence is driven just by the progress that we're making, working with them on qualification and talking about more efficient architectures, and we're just going deeper. So I feel like that will add to our diversity in fiscal '26. I mentioned this category that we're calling emerging hyperscalers. This is not new. We've talked in the past about this class of companies that's talking about large CapEx. And you're seeing companies out there spending a lot of money that you wouldn't consider being in the top 5 hyperscalers in the U.S. We've been designed into those architectures for some time. And so we've talked about being a beneficiary as they spend. We mentioned that this quarter, we'll have a new 10% customer. That's not a new customer for us, but it is evidence of this class of new companies that's spending actually following through on that. Hard for us to say how that looks, but typically, somebody is not going to build 1 data center and stop, right? So if anything, that will become a part that adds to the diversity of the company. We've talked about optical and I'm sure we'll get into it. But for our team, about 12 months ago, I said, look, we need to achieve 10% of our revenue in fiscal '25, 10% or higher with optical DSPs. And I think we're well on track for that. So I see that adding to diversity as we look into '26 and '27 as we continue to grow that business. We've talked about entering the PCIe market, and you'll see that we'll make announcements later this year for PCIe Gen 6. So what we'll bring to market is solutions that will be used inside appliances. So retimers that will be put on those boards will be delivering a set of AEC solutions along with it as the scale-up architecture goes rack-scale. And then we're also going to have options that could be integrated into optical solutions as the scale-up network goes cluster-scale. And so that will add to our diversity over time as well. So it's a combination of expanding our customer base as well as expanding our footprint within the different business lines that we're pursuing. And that's going to take place over really the next few years as we grow as a company. .

Got it. Okay. I'll definitely come back to optical. Maybe on competition in AEC, I think ever since the IPO, you guys have talked about you expecting competition, if anything, competition sort of being an endorsement of what you've been trying to do. A couple of your competitors are talking more about AECs. I don't think there in full production or they've won significant designs. What are you seeing from a competitive standpoint? And what are your expectations going forward? .

I think it's encouraging. When a market that you've pioneered gets to the point where we've got other companies are spending money in targeting that market. That's encouraging, and it's evidence that I think the market forecasters are on the right path and thinking that this is a very large market. I think that we haven't seen a big shift competitively from the last time we talked. We see competitors coming in with samples for qualification. Going through a qualification process, it's not a fun thing. It's a difficult thing to get through. And so I think that that's probably the stage. We haven't seen anybody reach volume production yet. But I understand, this is a fast-moving target. And so the way I think about things competitively, what we're trying to do is ultimately deliver to our customers in a way that's hard to compete with. But that's not really saying I got to be better than my competition. It's really how do we serve our customers in a better way. And we're organized in a way that I think we're able to do that. So we've taken complete ownership of the system. It is not as simple as putting a retimer into a connector, that's a gross underestimation of what's required. What we've seen is that our customers -- if we open the door for them to innovate, they'll come through the door and come in with features that they request, both hardware and firmware features that make the rack design that much more efficient. And so we're organized in a way that I can work on 20 or more different SKUs at the same time for different customers. We've built a capability to not only deliver the system solution but qualify it internally. And we're actually giving qualifications for our customers now. That's a level that simply you don't think about if you're just trying to sell a chip to a cable company. So I think the approach that we're taking allows us to respond more quickly to the demands of the customer as well as I think there's -- I think it's pretty natural that if you're going to take something into high-volume production, you really want to buy that from one person. You want one party to be responsible from the qualification all the way through any kind of issues that you have in production. And that's why I feel more confident, the further we go that we've picked the right model to be on. And ultimately, is measured by feedback from the customer. If we deliver flawlessly and predictably, I think that's encouraging from the standpoint of like trusting the model you've chosen. .

Right. Right. And your model at this point is still very unique. The competitors do not follow it?

I think it's unique in a sense that we fully embraced the accountability for the product line. But I don't think it's unique in a sense that the work needs to be done. And if it's one party doing the work, it's way more predictable than a group of companies doing the work and potentially looking at each other as the shortfall if there is one, right?

Makes sense. Optical DSPs, to your point earlier, great momentum, 10% plus of fiscal year '25 revenue? Where are you seeing the traction? What's resonating with customers? I know it's still a small base, but where do you see the .

This has been a long road for us, right? And hats off to the leader in the market for doing such a great job developing the market and delivering. We're on, I think, our fourth generation of DSP now. I feel better than ever about our ability to deliver good performance, measured by bidder rates, interoperability, power and ultimately having a disruptive position from a cost standpoint. This market is not a -- it's not a rich market from a gross margin standpoint, our customer base, the neighborhood that our customers live in, if they are in a really nice neighborhood within that market is in the 30% to 40% gross margin area. And so naturally, there's an opportunity for disruption. And so that's the -- that's really our focus, right? And I feel great about our ability to deliver on all fronts. The market is interesting in a sense that the market doesn't move at -- all at once from 100 gig to 400 gig and 400 gigs to 800 gigs and 800 gigs to 1.6T. There's a large 50 gig per lane market that we're ramping into and really emerging as the clear leader. The 100-gig market is going to be a multiyear market, and we've never been in a better position with the products that we're bringing to market and the engagements that we're seeing at a customer level. We missed the first 2 movers. NVIDIA, we're not part of that architecture and Google as well. But the rest of the market is wide open, and we're here on time and engaging early. And that, I think, is going to play out over many, many years. And so that's where I see us gaining market share really over the next 2 or 3 years. The 1.6T market will follow. We took a path to be the lowest power in the industry. So we're going straight to 3-nanometer. Everything in the market right now is at 5-nanometer and just simply not competitive from a power standpoint as measured by the end customers, the hyperscalers. And so I feel good about our strategy in 1.60 to kind of ramp as that market takes off as well.

Okay. On an apples-to-apples basis, how cost competitive are you vis-a-vis the incumbent? Rough numbers.

Yes. So, now understand that you're talking about what we've talked about is the N-1 strategy, where if our competitors are in a more advanced process geometry and we're able from an architectural standpoint to build competitive solutions in something 1 process or 2 more mature. Our cost, we're able to manage also the size of the design. Our cost base is we could be as much as 50% lower from a cost standpoint in that case. For the 1.6T market, we've listened to the call to action for power. And so we're actually first to go to 3-nanometer. And you can say that I guess, the whole N-1 story may be shifting. So ultimately, we're responding to the market demand for lower power, but we'll be in 3-nanometer. And probably our competition, if they want to match us on power, we'll need to go to 2-nanometer just because the architectural advantages we're bringing. But that's probably not going to be a long-term part of the conversation. .

Right. Okay. Got it. Maybe I'll pause here. I've got a bunch still and see if anyone has any questions.

I had a question around the 5- to 7-meter rack-to-rack opportunity for AECs. You brought up that architecturally, some designs don't have a top of rack switch, which would be the classic [ end rack ] AEC use case and have a kind of row of switches adjacent to the racks. Is that something that opens back up into the AEC market as you get into the 5- to 7-meter length where even if it was a design that has kind of lease switches adjacent to the rack as we densify those networks, AEC becomes addressable for that architecture as well? .

Yes. So I'd say that there's -- when we first imagined the market, we definitely targeted 7 meters. And the idea was that AECs could span 2 or 3 rack connections. What ultimately developed was the market was really for in-rack from Nick to tour and also switch racks where you're connecting switch to sit, switch to switch. So for the last few years, we didn't see much demand for 5 meters or 7 meters. . But as clusters are being built and there's progress being made on power sourcing and also cooling, it allows for the densification, some of the things that we've seen in the last 3 or 4 months are super impressive. If you look at the density, and it's enabled by, again, power sourcing and cooling. If you look at the Pareto of failures for a cluster and you have link laps for AECs and all of the link flaps that you're having are on optical. It's natural to say, how can we architect in a more dense way so that we can get -- have the switches to be closer to the appliances. And so that's where this 5 to 7-meter market is kind of now emerging. We're working with a lead customer that's literally rearchitecting so that they can have AEC connections between the racks of appliances and the switches. I will say that our approach to the market is maybe a little bit different than a typical IC supplier. We're really a connectivity company. And so when I talk about this effect, I'm naturally thinking about what can we do with our optical DSPs to help address the problem in the market that's being seen. So it's not as if Yes, absolutely, we'll deliver this capability of the market with AECs, but I'm really looking beyond that to say what's the next innovation that we can deliver to help address the problems that our customers are having. I see -- yes, there's an opportunity for AECs, but I into see beyond that to say how do you deliver a zero flap cluster .

Any other questions?

Yes. So for what you have mentioned about what has been seen in the last like 3 or 5 months? And what will be the -- for -- in terms of the revenue and the guidance, is there any change for the whole year guidance or something? .

You mean for our fiscal year?

Fiscal year.

Yes. I would say that it's -- for the fiscal year, so -- we're 4 months into the fiscal year now. So 8 months to go. And really, the ideas that we're talking about here will play out over a longer period of time. So I would say that anything that's kind of near term is really in flight. There's always puts and takes when you look at what's happening within a given fiscal year. But everything we're talking about is really FY '26, '27 and really beyond. .

Maybe bring Dan into the conversation. Gross margin -- product gross margins in the quarter, really impressive. Curious what drove the sequential and year-over-year increase. I think the guide for the quarter implies a step down Q-over-Q. What's driving that? Is that conservatism? Or is there something going on? .

Yes. So let me first say the headline should have been last week. From a product perspective, the strong execution we had -- if you exclude product engineering services from the way you look at revenue, product revenue was up 32% sequentially. And so that increasing scale gave us some natural leverage to increase that gross margin. The product gross margin expanded by over 900 basis points in the quarter from Q4 to Q1. So we had a lot of favorable tailwinds is what I'd say. So it's the scale, which I highlighted on the call, but there are other things as well, kind of the usual suspects. We have favorable product mix. favorable manufacturing variances. We can't always assume we'd have those going forward. So there is some natural conservatism built into our guide. Overall, we said IP revenue will grow modestly from Q1, which implies that of that 11% sequential increase that we projected, you'd also have product growth as well. And we guided kind of flat overall gross margin line. So we should be in the same ZIP code. There's nothing specific driving it 1 way or the other. But longer term, we're within eyesight of that long-term gross margin model that we had laid out 3 years ago, which is 63% to 65%. We were nearly 63% with a predominant revenue mix being product last quarter. So you'll continue to see that trend evolve over the next few quarters and into next year. .

Got it. Okay. Maybe one on R&D and how to think about leverage -- OpEx leverage going forward. I think you guys have talked about OpEx growing at roughly half the rate of revenue, which is great. You're driving leverage. At the same time, Bill, just talking to you, it feels like you've got massive opportunity ahead of you. You do -- maybe you do not have competition, direct competition today, but you've got large competitors who are kind of coming your way. How are you able to deliver that kind of leverage how do you balance sort of investing in growth versus delivering margins and returns for your investors today? .

Sure. It seems like we're doing a lot. Yes, we are. For us, the core is the SerDes technology. That's the core of everything that we do. We think that if you're going to be in the connectivity space, that's got to be the initial building block. From there, everything we're doing is a leverage from that core. So it may seem like we're in different areas, but it's really almost like the application-specific implementation of this core technology. So we are growing from an R&D standpoint. And the question is, can you grow your top line revenue faster than your expenses are growing. And I think we're in that stage. So we're very much heavily investing for sure. and we're going to continue that from the standpoint of the foreseeable outlook. .

Got it. Okay. Another question that I often get from investors is on M&A. You frankly as a buyer and a seller. I think to your point, you've got a really competitive SerDes technology. I don't think many companies globally have that. So I often get the question, why wouldn't companies A, B and C would be interested in Credo? The other question I have is kind of the other way of looking at it is, okay, you guys are doing really well today, but you do have large competition. Over the long run does Credo need to become bigger. So not bigger for the sake of being bigger, but do they need more IP? Do they need more parts to be successful long term? How would you sort of respond to those questions .

Yes, I appreciate the question. .

I do get some.

I'm just kidding.

Don't shoot the messenger. .

I think that -- I think the NVIDIA Mellanox combination has kind of turned a light on to the entire industry about the importance of connectivity. I think it's center stage, spotlight on the things that we do. AI cluster is not a cluster if it's just GPUs. The connectivity was really the key piece. So does that become table stakes to compete with them, right? And so that's -- we've seen a lot of conversations around how do certain people in the industry respond to the lead that NVIDIA has. It's an interesting conversation, maybe more academic for us at this point, but we don't really think too much about that. On the other end of the spectrum, of course, we do have a currency now. We do have the ability to consider certain adds to our resource base or certain ads from a technology standpoint. I think we're eyes wide open as it relates to different pieces that could complement what we're doing. There's a lot of opportunity in the future as we look at 1.6T and even beyond. And I think the winners are going to bring creative solutions to the market that are going to be system level. And of course, we're thinking through that. it feels good to have the currency that we've got right now. .

Okay. Great. Maybe in the last minute, anything that we haven't touched that is topical, you want to highlight I know we're keeping you busy with meetings all day.....

I appreciate that. .

Of course, it's all you. Anything we collectively miss about the Credo story or what you guys are working on, what you're excited about? .

I think you've done a great job. I appreciate the fact we've got more time today than we typically have. But I think generally, include me in the category of being very bullish about AI long term. There's a lot of conversation, are we in a bubble? And bubbles are measured, what -- if there's less CapEx next quarter than this quarter. That's not the category. I mean, I'm very bullish about the application. I'm very bullish about the activity that we're seeing. I'm bullish about the potential expansion of the customer base beyond the 5 U.S. hyperscalers and that being a long-term trend. I think this is a technology that that's really going to democratize the end customer space. So I really couldn't be more excited about the market pursuing .

Great place to end. Thank you so much. I appreciate it. Thank you all.

Thanks.