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

All right. Great. We'd like to get started. Good afternoon, everyone. My name is Toshiya Hari. I cover the semiconductor space at Goldman Sachs. Very excited to have the team from Credo with us this afternoon. We've got Bill Brennan, President and CEO; and we've got Dan Fleming, CFO, from the company. I have a list of questions. We'll keep it pretty open and interactive. But before we do any of that, I was hoping to kick it off to Bill and if you can help with the intro.

Yes, thanks. I appreciate you hosting. I appreciate everybody attending. I appreciate the interest in learning about our story. So I'll just frame the Credo business at a high level. The -- we're a pure-play high-speed connectivity company. The world is going faster. Connectivity is a key piece of the puzzle. It's a very difficult space, communicating over copper wires or optical fibers at the speeds that are being pushed in the market today. We started with the Ethernet market, which is a standard that is really ubiquitous in the data center world, especially in the switching and routing layers. We are the only company in the market that is addressing every Ethernet connection within the data center from active electrical cables to optical DSPs to line card PHYs to even to selling IP to the companies doing the larger chips that we're not doing and also, we build chiplets. We're a leader in that space. So our approach is that we don't tend towards one technology. We're looking to cover the entire space. And we can talk more about that, but I will mention that the rest of the world is going to go faster as well. We've got developments for USB even. So the other end of the spectrum, consumer is going to be going at speeds that are very, very difficult from a technology standpoint. And those are the problems we're attracted to. PCIe and CXL is also going to go very high speed in the future, and we'll intersect that market when we can deliver the same kind of advantages that we're delivering in the Ethernet space, which is really power efficiency and cost effectiveness. That's what ultimately our technology boils down to as a differentiator.

Got it. That's super helpful. I definitely want to spend more time on longer-term market dynamics, your competitive differentiation and some of the financials. But in terms of the near-term outlook, you just had your earnings call, I forgot the exact date. It was fairly recent. I doubt much has changed. But at this conference, clearly, a lot of focus on the macro and implications to cloud spend and so on and so forth. So in terms of what you're hearing from customers, seeing in the marketplace, any deviation from what you guys communicated on the call?

No real deviation. I know it's become a hot topic, and I can give you perspective, my perspective. The 4 U.S. hyperscalers, if you add the spend, it totals about $150 billion for -- that is kind of forecasted for this year. And so, China is also a big conversation. But if you look at Alibaba and Tencent and you look at their total spend that's forecasted, it's really a fraction, it's 1/10 of what the U.S. top 4 are expected to spend. There is -- we have seen some softness in China. And I think that's because that business is driven by that domestic market in China. We're not seeing weakness. I'm not seeing weakness at a macro level with our customers in the U.S. And even if we did, I think we're in a stage for our company where we're ramping in new technologies. So we're not even close to a point of saturation where we would be affected by the macro trends. So we're ramping all of our programs or new technology. And so, if our customers are looking at spending a dollar, they'll spend a dollar on the new technology because they're going to get more productivity out of the next generation. So we find ourselves in a good position. We will reach a point where we're affected by the macro level, but we're not there yet. And I'll say that we're not immune to schedule delays because the new technologies that our customers are trying to ramp, the entire system solution has to come together, other suppliers, they have to execute on that technology. So we'll wait. The bus will take off when everybody is on it. And that's the only deviation that I would see that would affect us. Nothing related to the macro.

Got it. And, I guess, Bill, just given what you described, i.e., you guys being uniquely positioned and being nascent, right, in terms of that technology curve, if you will. Maybe asking questions around overall cloud hyperscale CapEx might not be the most critical question, but again, I'll ask it anyway because everyone is asking the question. Based on the numbers that our colleagues forecast for overall hyperscale CapEx, this year is up around 25%. Next year is actually down 5%, 6%. To the extent you do have visibility on that overall spending profile, does that sound about right? Or is that too cautious? Any comments -- again, your business is different, but...

Yes. So it sounds about right for this year. In fact, we're hearing even more strength, possibly than the 25%. I think the cadence that folks that have been in this industry a long time is grow and then stay flat, grow and then stay flat year-over-year. So that wouldn't be surprising to me. So that's kind of a long line with expectations.

Okay. Got it. And then you sort of addressed this in your prior comments. But in terms of what you're seeing in China, hyperscale spending there, you spoke to softness. We've been getting a lot of questions around the recent restrictions on high-end data center GPUs. And to the extent that impacts deployment plans so on and so forth. Anything you're seeing there?

It's not going to really impact our business. We're deeply focused on the 4 in the U.S. We also focus on smaller players in China, but our business is not something that would be considered significant where we would have to talk about adjustments.

Got it. This one, again, perhaps a little macro in nature and not directly driving your business. But a lot of investors are focused on next-generation server processors being deployed, hopefully, over the next 12, 18 months as both Intel and AMD introduced their respective new product lines. Is that something that could catalyze an acceleration in growth and perhaps help your business? Or are you completely detached from that dynamic?

We're absolutely attached to that dynamic. And so, when you look at a mix chip, and you talked about Sapphire Rapids is a popular discussion right now. And so, that's Gen 5 PCIe, those lanes of SerDes are running at 32 gigabits per second. And NIC will ultimately take lanes of PCIe and then put out lanes of Ethernet, and we connect to those lanes. So the 32 gig lanes will turn into 100 gig lanes for us. So that will accelerate the move to 100 gig connections between the NIC and the top of rack switch. That's very good for our business. If we look at Ice Lake the generation before, that will represent the conversion from 25 gig lanes that NICs are at today to 50 gig lanes. So we're very much attached to it. So it's quite meaningful. I'm not sure if I'm as bullish as others are, that Sapphire Rapids is going to ramp in a big way next year, but that's in the sense we get great visibility with the customers we're engaged with. And our view of the hyperscalers is that each one of them represents a different market. They're all marching to a different drumbeat. And so some may ramp it sooner, some may ramp it later.

Got it. Diving a little bit deeper into your individual products, high-wire AECs are clearly the driver for your business for the next couple of years and beyond that. I guess, for the audience, what are high-wire AECs? Why are they so important? And maybe talk to your differentiation there as the leader and the pioneer of that market?

Yes. So when we think about AECs, we think about Credo pioneering a multibillion-dollar market. That's very -- that is going to be very quick to ramp. So going back several years ago, we had this -- we had a vision that passive copper cables were not going to be sufficient at some point. In connectivity solutions, the speed of the connection or the speed the data rate on copper, especially because copper is a lossy material, it really impacts how far the signal could travel without being recoverable, but on the other end of the wire by the receiver. And so, we knew that a 25 gig that DAC would be sufficient. But we had a feeling that at 50 gig significant signal integrity challenges were going to happen and that many companies were going to be able to navigate those signal integrity challenges. So they would have to consider another solution and the other solution that people talk about are active optical cables, AOCs. So we had the vision of saying, let's extend the life of copper by embedding DSP chips in each end of the cable. And if you open up an AEC, you'll see our chip is the main component. If you open up an AOC, you're going to see several electrical chips, several optical chips. And so you can see making the conversion from electrical to optical and then back to electrical. There's a cost to it. It's expensive and it's high power. Keeping everything in the electrical domain means that we're able to deliver a solution that's equivalent at a standards perspective, actually better bid air rates. We expect a 10-year reliability where the optical is nowhere near that. Our designs are very rugged physically in a sense that you're not going to break the glass. You're not going to break our cable if you're routing several in a tight RAC deployment. But we're really half the power, and we're half the cost. And so it is compelling by nature that it's a solution that is being well accepted in the industry. And so this reality is that I don't think if Credo played the role that we played that the market would be where it is today. So I think this is probably the top highlight to take away is that, we pioneered a multibillion-dollar market that's in front of us during the next 5 to 10 years.

And in terms of customer adoption, you've talked about Microsoft publicly as your leading customer there. You have a follow-on customer. And, I guess, my question is, how should we think about adoption beyond that? What are some of the catalysts that could drive further adoption on a broader basis?

So talk a little bit about the Microsoft effort. We've -- you could find a video online where we introduced the product together. We did a live stream talking about the technology. That's Microsoft's approach is a very open source. And there's actually an innovation in the cable. We're not just replacing a DAC or an alternative to AOC. We actually delivered the most intelligent cable that's ever been delivered in the data center. Our cable has the ability to sense when a TOR port is failing, and we have the ability to redirect the traffic to a redundant TOR. And so, that is a huge efficiency benefit that Microsoft has been trying to solve for, for many years. We were the one they collaborated with. And so, there's a level of innovation that we're doing that further separates us competitively. I'll say that the second customer is also a unique design. It's not a standard connection. There's actually 4 connectors to the cable. They're not as open with the applications. I can't really speak about it. And so, you've got that thread of innovation that should be considered, but more directly answering your question, as the world goes faster, this will be a product category that is adopted. So 100 gig lanes, we believe that DACs will be obsolete when the world moves to 100 gig lanes. And we can kind of talk about the timing of that, but that is coming within the near future. So that really is a true inflection point when we expect that the adoption rate of server to TOR connections and switch rack connections will adopt the AEC category.

So it's fair to assume that you're engaged with other customers, other hyperscalers, other data center operators?

Yes. We've shipped cables to more than 30 customers, not all -- not everyone is a big customer. But yes, it is something that's well-known in the industry now. And yes, we've had activity with pretty much all the hyperscalers.

And then in terms of the competitive landscape, there's one other company in particular that has publicly talked about their efforts and their aspirations in the market. Can you talk a little bit about your differentiation in AEC?

You're referring to Marvell.

Correct.

Yes. I have a soft spot for Marvell because I spent 11 years there. So I have a lot of friends that are there, and I respect the new management team a lot. So I think I was really very pleased to have them come to the market. I think it's a reflection of the customer base asking them to invest in this area. It's a reflection of the customer base accepting the product category and asking supply chain partners of theirs to invest. So Marvell is surely a capable company. I would say from a differentiation standpoint, there's kind of many levels to it. Our model at this point is to take full ownership of the cable design. We are the single throat to choke with each one of the end customers that we're selling to. We've started the process thinking we could just enable copper cable companies, but we've changed our model and we realized we weren't going to be able to develop the market as fast. I would say that another differentiation that we're going to have long term is, look, the reason we win head-to-head competitions where there is head-to-head competition, we win because of power and we win because of our cost efficiency as well. And that's something that's going to continue for the foreseeable future even as we move to 5-nanometer to compete with us on power competitors will need to be in 3-nanometer. And that's what we refer to that as an N-1 process advantage. So even at the SerDes design level, this is primary decision-making on which process we use and the architecture that we build. We have a strong commitment to being different than the competition in a better way, lower power, lower cost. And so, when we talk about in the future, there will be 1.6 terabit per second AECs. And if I built that in 5-nanometer, my power would be on the order of 10 to 12 watts per cable length. You wouldn't expect Marvell to even come close to that, they'd be more like 20 unless they went to 3-nanometer. So this is a real world example of why our core SerDes technology ultimately delivers significant advantages to our customers in deployments.

So that N-1 competitive advantage you talked about is very much sustainable. It's fundamental.

Yes, it's proven in 5-nanometer.

Got it. Earlier, Bill, you talked a little bit about the risks around projects being pushed and pulled on timing. Again, just given that this is more of a tech adoption dynamic as opposed to simply replacing something. It feels like your business should be very resilient even in the face of macro headwinds. But, I guess, going back to that risk around projects being delayed, what typically drives that? Is it kitting issues? Is it...

It's typically not supply chain issues. It's not that they can't source something. It's typically on the technology execution. There's so much that has to be proven and qualified that sometimes certain aspects take longer. And so, ideally, everybody stays on schedule, but we live in a world that this is not easy technology that our customers are trying to deploy. So it's hard to predict. But it's not -- I don't see supply chain issues causing delays.

Got it. Okay. I guess, last question on AECs. There were some supply side issues, dynamics, headwinds a little while ago. I think you're taking steps to alleviate diversify your supply chain going forward. Can you talk a little bit about those efforts? And at what point do you feel like you'll be comfortable from a setup standpoint?

Right. So just to add color to the issue that we faced in April. Kunshan is a very popular place to build the world's technology products. Companies like Tesla were impacted. So Shanghai and Kunshan very near each other. They went into a shutdown in April, literally factory shut down. So we found ourselves unable to build the AECs that we had demand for. And so, there's a near-term plan to -- if we are impacted again in the future, we've built an inventory. We've gone far ahead of customer demand. And you see that reflected in our financials. That was something that we told you we were going to do and we did it, and we're going to continue to do that in the short term to make sure that we've got a significant buffer to make sure that we're not impacted by any kind of shutdown. But it also -- it kind of accelerated the need for geographic diversity. And so, we've got 2 suppliers, and we're in conversations with them about bringing up an additional line. We're not transferring any equipment. We're going to bring up additional lines to support our growth, but doing it in other low-cost manufacturing locations like Malaysia, Vietnam. So -- and I would say if you want a time frame on that, it's probably within the calendar '23 window. It's not something that's going to happen in 3 months, not something that's going to happen near term.

Right. Got it. And then, I guess, one other supply chain question. Just on your foundry strategy. I know you guys are very close to one important foundry supplier, very successful, I forget if you've been public with...

We've been public.

Okay. So TSMC.

Yes.

Surprise, surprise.

Yes. Credo is very happy with TSMC, the relationship that we've got.

So there's a lot of industry level debate around localization and things like that. As you think about your foundry strategy, I guess, you kind of answered the question, but you're happy with TSM (sic) [ TSMC ]. Any thoughts around diversifying away from TSMC? Still relatively small-scale business. So near term, probably not. But as you think longer term, how do you think about your foundry setup?

So I think with our supply chain partners today, TSMC, our packaging partners are cable assembly partners. We're very comfortable with where we are. There's a lot of discussion about diversification. And we're in a point with our company, we've really got to focus on execution of the next generation, duplication, my engineers will shoot me if I have them duplicating instead of innovating. And so, right now, we're full steam ahead on innovation, and we're super comfortable with our supply chain.

The optical DSP business, obviously, feeds into high-wire AECs. But as you think about that business as sort of a stand-alone chip business, if you will, today, still nascent, young in its life cycle. You've talked about engaged -- being engaged with optical transceiver companies, quite a few of them actually. You have a very large competitor, again, in Marvell, [ classic ] Inphi. Again, talk about your differentiation there and talk about your aspirations from a market share perspective, if you will, if you think about it that way?

Sure. So we talked about this multibillion dollar market that Credo pioneered AECs. It is us and it's the rest of the world, right? And so, if I make a parallel to what Inphi accomplished in their optical business, they were the pioneer of 50-gig lanes and 100-gig lanes. They were the first. We're part of the rest of the world. We think that we're a very strong challenger. We played the role of disruptor in the market. We've got products, optical DSPs, laser drivers and soon in the future, TIAs that will cover any design that's happening, any advanced design that's happening for all 50-gig lane on the fiber side and all 100-gig per lane on the fiber side. That includes 200 and 400 with 50-gig lanes, it includes single lane 50 for the 5G space, single lane 64-gig fiber channel. And for the 100-gig per lane, we've got 100, 400, 800. So we're very, very well positioned. And we are penetrating 50-gig per lane designs, and we've got plenty of activity on 100-gig per lane designs. One of the things that is a reality in that space is that, the hyperscalers are the ultimate decision makers, how they spend their resources with qualification really matters. And so, we've been successful with 2 of our customers, 2 hyperscaler customers, getting them to specify our DSPs. It's referred to as a joint development model or JDM. And so, we've had success with that. We spend probably more efforts now with hyperscalers directly than we previously had been because if you go win and design with an optical transceiver company, you have to -- you're kind of betting that they'll be able to get qualified by the hyperscaler. And so, getting involved in that decision-making is really the strategy. So we see that as a $1 billion market. It's not quite the size of AECs, but we see it long term being our second largest product category, where today, line card is our second because that was our first product category that we ramped. So long term, we see AEC, optical and line card. From a market share standpoint, we're going to grind. I mean, we're going to be a persistent disruptor. And so, my belief is that, we'll have double-digit market share within the foreseeable future.

Okay. Got it. Helpful. We have about 10, 15 minutes left. I wanted to pause here and make sure we didn't have any questions from the audience.

Yes, can you comment on whether you have any strategies in the automobile industry for the SerDes, et cetera?

Good question. We go to markets where the speeds are the fastest. So we entered the markets we're in at 25 gigabits per second. And so, we saw this exponential increase that was driven by Ethernet from 25-gig per lane to 50-gig per lane to 100-gig per lane and eventually 200-gig. But that represents is a technology challenge, a rising technology hurdle that only a handful of companies can navigate. We're going to the USB space because the future there is a super difficult problem to solve. We're going to the PCIe space because they're moving to 64-gig PAM4, Gen 6. So that's where we're going to intersect that market. Automotive, the lanes of communication in automotive, it's not faster than for us yet. So we're not going to be able to deliver the efficiency that we do compared to our competitors in other markets. So I would say that I'm not going to be one of the CEOs that shows automotive as the flashing new thing in the future. It will come when it comes when those connections go faster.

[indiscernible] that all the different hyperscalers are driven by different trends. And I know you're doing a Microsoft and obviously, the 25- to 50-gig or the 25-gig copper solution is an interim solution. It sounds like it positioning yourself to different products [indiscernible] upgrade speed. But can you talk a little bit about what different hyperscalers are driving? And then also sort of talk about the [indiscernible].

Sure. So I'll give you a couple of points to consider as an example. I'd like to say I'm relatively new to this market. I've only been doing it in 9 years. Okay. So prior to that, I spent like more than 20 years in hard disk drive IC world. So as I came into this market, the forecasters, the world was telling me that everybody is moving to 400-gig modules. The 400-gig is on the horizon, the whole market is going to move. It didn't really play out. So Amazon went and Amazon went because the benefit of having 400-gig in a small form factor allowed them to build the kind of switch rack. If they didn't have 400-gig, they would have multiple racks stacked up, they wouldn't be able to accomplish their objectives. So form factor drove them. Google really pioneered the 200-gig market. And so, they were the first to go to 50-gig lanes. Facebook decided their next generation, they were going to use 200 and then Alibaba did and then Tencent did, right? So the forecasters were not right. Of course, 400-gig will be popular in the future as people convert. When we talk about 800-gig, the whole world is not going to go to 800-gig. So one of the big guys in Seattle is pretty actively talking about moving from their 400-gig deployment and jumping to 1.6 T because that form factor gets them to the bandwidth, the density within one rack. And so, that's an example of how all of these teams are kind of marching to a different drumbeat. We could continue that conversation. That's why we tend to look at every hyperscaler as an independent market. And I think that's going to -- that's the best way to serve these customers.

Can you talk about the hyperscalers and those architectural choices [indiscernible] Microsoft and that for rack and redundancy to get there? What does that architectural placements to customer give you in terms of visibility on the growth, the durability of that [indiscernible]?

Yes. So I think the challenge is how can you be invited to collaborate with their customer. This is not something that is the right of being in the business. This is something you earn over time. And with the Microsoft case, we didn't start with this cable that enabled their redundancy goals. We started by building a switch rack with them. It was a pilot type of thing, proof of concept. They really got familiar with our products. And that credibility that we gained opened the door to collaboration for this other problem that they were trying to solve. Now it's a function of us trying to stay connected with their architecture and design teams so that we're getting visibility into what we should build next so that we've got our AEC aligned with their road map. And that's really the goal with every customer. That's the way you build a significant position.

Maybe bring Dan into the conversation, a couple of financial questions. Gross margins, given the transition from an IP-centric business to a product business and sort of AEC ramp at your largest customer and your second customer and so on and so forth. Lots of moving parts. How should we think about sort of the puts and takes from a gross margin perspective over the next, call it, 4 to 8 quarters?

Yes. You've got historically, a large portion of our revenue, as you alluded to, was IP, which is near 100% gross margin, obviously. That has been declining. If we look just at Q4 to Q1 of fiscal '23, the IP contribution in the revenue mix was 30% going down to 22%. Meanwhile, we have this underlying current where product revenue is ramping at a pretty sizable clip. So if you just isolate on product margins itself, kind of ignore the IP revenue, what you see, the underlying current is, we've been benefiting from scale. So as we've achieved larger scale in our product shipments, our margin has expanded over 11 percentage points from Q1 '22 to Q1 of '23. So over the next 4 to 8 quarters, really throughout the remainder of FY '23, we'll continue to have some margin expansion due to increasing scale. I'm not sure it will always be linear, but it will be a trend that's persistent. Beyond that, as we enter FY '24, there are some revenue mix changes that we believe will be favorable to us from a margin perspective. And that 52% product margin will continue to trend up towards 60%, let's say, over the next...

That's very clear. And then maybe just a question for both of you guys. But in terms of OpEx going forward based on, Bill, what you described in terms of the opportunity set, you've got plenty of growth opportunities across your product portfolio. I'm sure if your engineers could spend as much as they wanted to, it would be a lot of money. So far, you guys, despite being a relatively young company, at least public company, you've exhibited significant OpEx leverage. Should we expect that going forward? Or could there be a point in time where the rate of OpEx growth accelerates?

Yes, I wouldn't say it will accelerate right now. If you look at what's really occurring again, FY '23, we kind of reiterated our target for revenue at $200 million, which would be 88% plus year-over-year growth. Our rate of increasing OpEx year-over-year is less than half of that rate. So you're seeing that already going from Q4 to Q1, there was a significant expansion from that operating leverage, and we expect that to continue. Our long-term operating margin that we've stated is 30% to 35%. So that's something that we believe we're on the path to over the next few years.

Okay. And then, I guess, bringing it back to the business side. Bill, you talked a little bit about PCIe and USB as opportunity sets over the next perhaps a couple of years. Can you kind of expand on that -- on the timing of that inflection? How significant that opportunity could be for your business?

Sure. So I'll start with USB. So we're in development on the next-generation standard that will be introduced to the industry. We're -- our first contract is an IP contract, and we're engaged with one of the largest consumer companies in the world, the company that typically takes the first -- they're typically the first to ship the new faster speed standard. And so, that IP contract is something that's material, and we'll be earning that over some period of time. This will also help us enter the chip market because we'll build a retimer that's compatible. The company we license to, they build their own ICs. They don't sell ICs. So we're kind of a necessary piece of the puzzle for them in a sense that anything that connects to their USB ports, that's going to operate at this next-generation speed, we'll need to have something interoperable and available. I'll also say that the AEC market for USB is going to be substantial. Anything over 18 inches will need to have retimers built into the cables. I'll talk about PCIe. So we talked about Sapphire Rapids earlier. The next-generation is where we're going to intersect the Gen 6. We see that market -- we're going to build a retimer, and we see that market being substantial in a sense that -- in the storage part of that market, the CXL part of that market will drive big revenue opportunities for us. And when we talk about timing, everybody's got an opinion about when the next-generation is going to come. But we're definitely talking about something that's on the horizon '24-'25 time frame would be our expectations. But again, we'll follow the market.

Okay. Got it. I guess, in the last couple of minutes we have, I wanted to give you the opportunity to speak to anything that we may have missed or obviously, you're still a relatively young public company, so you're still sort of pitching your story to investors. But anything you feel like we collectively miss or underappreciate about your technology, about your competitive position, the markets you play in [indiscernible].

Sure. I would say that a key takeaway should be that as we look at the next 3 to 5 years. And by the way, we believe this is a decade-plus macro trend that we're enabling. The world is going faster. But if we look at the next 3 to 5 years in particular, this is not a technology execution play. This is not a customer adoption play in a sense of accepting us as a new supply chain partner. This is really a pure execution play. The technology is developed, it's available. The world -- the Ethernet world is going to move from 50-gig lanes to 100-gig lanes, all of the Ethernet markets that we're in, it's wind in our sales. As we look at the new markets that we're pursuing, the same advantages that we're building market share with in Ethernet, we're going to deliver the same in those new markets. I will say that, even though it's a bit getting into the details, our SerDes capability is the core competence for the company. This is primary for us and we are different and better than the other companies that are capable of designing these SerDes. What it means to our customers in the products that we deliver, it's fundamentally lower power, more energy efficiency, as well as more cost efficiency. And so, I'm pretty excited about the next 3 to 5 years because it's really let's go and execute. That -- those are the primary goals that our team is looking at.

Excellent. I think we're pretty much out of time. Thank you so much for the time.

Yes, thanks.

I appreciate it.

Really appreciate it.

Thank you.