Marvell Technology, Inc. (MRVL) Earnings Call Transcript & Summary

Thanks for joining. I'm Blayne Curtis. Very happy to have with us for our last but definitely not least, presentation, Marvell. From the company, Matt Murphy, President and CEO; as well as Ashish Saran from Investor Relations.

So Matt, I mean, you've led the company 5.5 years. I want to leave the first one fairly open-ended. You just had your Analyst Day. It's pretty dramatic, the shift from consumer. People think you've been a HD company, next now you're 84% enterprise. Maybe just talk about that transition? How much of this is complete? You've been very acquisitive over that period. You've been there. And really, what's the vision of the company for the next 5 years?

Yes. Thanks, Blayne. Great to be here. If I reflect back, I mean, we basically set a mission in front of us, really 5.5 years ago, and it was within a few months of me joining. And that mission statement, which is kind of simple. It totally resonates today. I think back then, you could kind of envision it, but now it's like -- it's the thing, which is we want to be the leading company that makes semiconductors that move data, store data, process data and secure data. Do that faster, more reliably than anyone else, be the best in the world and just do that, figuring that, that would probably be, in our view, the biggest end market growth driver of semiconductor demand, at the highest level for the sort of 10 years, let's call it, 2016 to 2026 or something. And I think that was a pretty good bet. The issue was kind of the hand we were dealt wasn't that hot, if you remember. We had a whole bunch of businesses. The only one that was #1 was hard disk drive controllers, where our claim to fame was being #1 market share in 2.5-inch drive selling into laptops, which was a market going downhill, on a slope steeper than the site of Mount Kilimanjaro, okay? So that -- and then we had a sort of legacy enterprise business and not a lot else. So I think the -- we kind of remade the company off a fairly small base, but the organic effort in storage, in automotive and in networking, both the switches and PHYs for enterprises has been -- were enormously successful. And then basically, what we did to your point was we divested or shut down almost all of the consumer effort and did aggressive M&A, right, and pretty, I think, reputable, respectable capital returns as well in terms of how we manage the balance sheet. And as a result, we've really stitched together and kind of merged into this platform assets from Cavium, assets from Inphi, from Avera, Innovium and Aquantia. And now when we look forward, that mix has shifted to, like you said, 84%, it will keep kind of going up as we go forward here, really an enterprise data center, carrier type of end market-focused data infrastructure with the remaining consumer. And last point, I would say, even the consumer we have left it's a grower. I mean, it grew 20% year-over-year in Q3. We guided it up to be up, I think, double digits year-over-year in Q4, and it'll grow next year. And some of that has to do with which you are kind of early, Blayne, as Barclays and you personally and sort of figuring out some of the opportunities we had in consumer SSDs, which was in the PlayStation. But that's still in front. So even the consumer piece is not bad, but the real focus of the company is in the other 84%.

I want to ask you is in -- through this whole process, you've kind of graduated, obviously, much larger cap and people think about you differently. There's enough drivers. When you're smaller, people say, "Oh, what's this socket, how much it's going to be?" Now you're breaking out by end market, which I think is helpful. It gives people a broader perspective like large cap companies do. But I still get the question, you'll say, all right, but what's the ongoing growth rate? Yes, they said the 3-year CAGR is 15% to 20%, but what are they going to grow over 5 to 10 years? And I'm curious if you think about -- there's 2 functions, right? I mean, you've been growing your SAM. I think Analyst Day, you guided $10 billion, and you're still only representing 1/4 of it. So maybe just start with those 2 points. How fast is the SAM growing? And do you think there's opportunities to continue to gain more share and capture more of that SAM?

Yes, I'll take part of it, and then I'll have Ashish answer. So maybe Ashish, you can frame up some of the growth rate stuff but I'll just lead off on the size. So if you look to the 2017 Investor Day, which was the first one Marvell had ever done, it was an $8 billion SAM. Okay. Now it's a $30 billion SAM. And the TAM that we articulated in the 2019 Investor Day was about $100 billion TAM with $20-ish billion we go after. So now it's -- we're basically, every year, to be honest, we're taking -- we're addressing a bigger and bigger portion of the TAM. So the SAM is getting bigger, and it's actually accelerating. And the growth rate of that SAM that we're going after that $20 billion to $30 billion is 13% a year. It's just kind of the market that we're in. And that includes, by the way, some of the lower growing stuff like as the consumer base. So it's fair. But I think just kind of our mix that we're in, what we call the sweet spot of the semiconductor industry, I think we've kind of nailed our opportunity relative to what we think is one of the highest growth rate of scale, right, relative to the SAM. And I anticipate as we go forward, and we fire up another Analyst Day sometime in the future, I'm pretty sure that SAM's going to be bigger because we just keep doing things organically and inorganically to address this now $100 billion-plus TAM by driving adjacencies. But Ashish, maybe you want to frame up some of the -- that we're seeing relative to that opportunity.

Yes. So I think there's a couple of key things which are relevant to your long-term question, Blayne. So first is in that 13%, that basically assumes the current implementation of cloud, basically the current Internet essentially, right? And the same thing on 5G. It's primarily connecting handsets, like the PlayStations. And auto, in this particular, it was given the window of that SAM was really around Auto Ethernet, right? So when you think and extend that time period a little bit longer and back to the sustainability of 13% or probably even faster, right? I think obviously, you see today, there's a big discussion on the virtual environments, which are probably going to require a lot more compute storage, networking, security, which will probably end up pushing that cloud growth rate significantly higher, right? Again, we don't know how much higher, but I think the writings on the wall is probably going to be a lot bigger than this today. And as an extension, right, 5G, which has been kind of looking for a killer use case beyond just connecting handsets. I mean, again, these virtual environments are going to be huge, right? There's also use cases extending 5G into enterprise networks, right, to enable private 5G applications. I think those things are probably going to push up the numbers we came out with not that far back, right? And automotive, I'm sure we'll have the discussion later today. I think this was just the Auto Ethernet opportunity, which we said is about $1 billion market, which is already a pretty impressive market. But the auto compute opportunity a little bit further out in time, is going to be magnitudes high, right? We did some initial sizing it of a $5 billion market, but my guess is that number is going to end up sliding higher over time. So I think the sustainability of the market is certainly there. I think the growth rates are probably going to be at or probably even higher out in time and I think we've got a hold position in terms of all the critical IP required to pull off access to this particular market.

So I did want to ask you, you brought up the concept of cloud optimized silicon. I think that was one of your lead slides actually, Matt. And within that, I mean, ASICs have been around for a long time. There's always a high cost to develop the ASIC and historically, there's a trade-off time, flexibility and such. Now you have the market kind of moving kind of on one side, more to chiplet architectures and things can be more configurable. And I'm assuming the cloud volumes are just so much bigger as well, then maybe there's room for optimization. So maybe I just answered my own question, but I'm kind of curious as you -- what are you seeing that you made this as important-enough topic that you brought up so early on the Analyst Day? And maybe you talk about the opportunities across there's -- I think you mentioned GPUs but also compute, which is not an area you're in right now. So when we talk about expanding your addressable market that would be huge.

Sure. Yes, there's a lot in that question, which is a really good one. Yes, we think this is probably the biggest growth driver for us from a market opportunity. And we outlined, as you said at the Investor Day, this transformation from, hey, the world is going to run on software. Everything is going to be on x86 s. It's all going to be homogenous and life's good. Now that obviously didn't happen. We're now in a world of heterogeneous and FPGAs get thrown around and GPUs and other types of sort of specialty function products brought in. And our view is simply, hey, this is just an extension of that, where the volumes you're talking about, the scale of the workloads that are being addressed and the kind of the cost benefit ROI that do-it-yourself cloud optimized makes total sense for everybody involved. Say ourselves if we're involved in our customer. But it's -- but it will be very clear. It is not your traditional ASIC business, okay? I think this is a misnomer. There was 7, 10 ASIC companies 10, 20 years ago, everybody had low margins. You priced it on the price per silicon micro acre, pennies of SAM, right? What never was a real great business. When we say cloud optimized, there's all kinds of business models playing. I mean, there is still a pure ASIC aspect to it where the customer does the full design, we would do the back end, the packaging, the layout. But increasingly, in those -- even those engagements, it is a -- very much a technical deep-dive cooperation on architecture, chiplet, how do you interface the memories, what IPs can we bring to the table that maybe the customer doesn't have. So even if it's an ASIC, it's not quite bad. And then from there, there's a range of business models. I mean, to be honest, apart from cloud optimized silicon, we have a number of projects right now, which are built to spec. So if the customer's chip. They defined it, but we're doing the design. There's no IP from them. But it will have their stamp on it, but they literally -- they don't have enough resource to do it. Or we have 90% of what's needed if you follow. And then we'll just go develop the last 10. So yes, and it's taking form, not just in -- so it's taking form certainly in accelerators, in compute, but also versions of this in storage for acceleration in security as well as in networking. And even if you think about the Inphi optics area, the asset that we bought, every one of those fundamental designs with the hyperscalers is really a custom architecture. I mean, it's not a one-size-fits-all, here's one PAM chip and everybody buys the same chip. I think you know this, different configurations of DSPs, drivers, TIAS, which laser cyclo do you pair with, how do you stitch it all together. So that's kind of the umbrella we talk about when we say cloud optimized silicon. It's across a number of product categories and business models, not "Hey, it's a vanilla ASIC, so it's low margin." So how are you going to -- it's like a completely different paradigm. And that's why we called it out because the cloud titans are really going to drive tremendous volumes in terms of silicon need for a long time in the future.

And you're clearly bringing a big IP portfolio to those customers. On the process technology, packaging and those Marvell and even Cavium were kind of N minus one. You did a huge job to kind of get up to 5 nanometer. I would say it was the first 5 nanometer where I think the message was you're very happy where you're at, right? And there's still lots of 5 to go. You also talked about 3 at the Analyst Day. So can you -- can you with 3 close that gap even further in terms of the timing of having the SerDes and the process development done, kind of in line with where the state of the art is?

Yes. Maybe, I guess, to clarify on 5-nanometer, we were -- we haven't lost a -- we never -- we didn't lose a single socket to time to market. So just to be clear, whatever you can argue who take out when, we were very competitive on that -- on that node, as an example, in terms of announcing and marketing it and having working SerDes and silicon. And then that was the basis really for driving this enormous pipeline of new designs that we've closed, not just in the 5G area, but as we called out earlier this year, a tremendous amount of 5-nanometer designs in cloud optimized silicon for that end market, which, in aggregate adds basically $400 million of incremental revenue we never had before in fiscal '24, doubling to $800 million. So I don't think you can achieve that kind of a game-changing revenue scale opportunity if you're late. I just don't buy it. I think we were there. We were competitive, and we'll be there on 3. That's the simple answer.

Got you. I want ask the -- you mentioned Inphi, I think that's been off to a very strong start since you bought them. Can you talk about this opportunity? They -- you gave some color since the first couple of quarters, but it seems like that little opportunity just continues to accelerate. And then when people see big CapEx numbers on hyperscalers, kind of figure a lot of that is translating into optical. So can you talk about that pipeline for you, where you're seeing the strength? And now that it's part of Marvell?

Yes. No, that -- we're thrilled with that acquisition. What a great team, we were able to merge into the company. You and many of the investors know them well. Ford really built a first-class premier asset, top to bottom. I mean, just from every way that company was managed. And if you think about when we started talking to them and you can look at the proxy. But I think Street at that time had them probably in the mid- to high 700s for calendar '21. And by the time we announced and made our own kind of estimate, we were pretty close to Street, which is about $800 million, if you think about it for this calendar year. And we've given enough indications sort of on breaking out their numbers and others, it's like high 800s. Now it's like 10% over kind of big round numbers, right? So the asset over performed in the first year, which is fantastic. Synergies have been on track. Team integration has gone very well. And like Cavium, right, their opportunity pipeline has expanded since we've closed it. One of the reasons is it's kind of the same thing as Cavium, and that they were really good at what they did, really trusted, good engineering team. But having them as part of the bigger platform, access to all of our goodies, access to the ability to add resources, pull in schedules, that's actually increasing the design win funnel and accelerating it from where it was. So it's sort of had a year 1 tremendous performance, and we're accretive, right, the first quarter out, which was a big sort of question mark. That's behind us now. And now it's going to be accretive to earnings. It was accretive to revenue growth this year in terms of percentage, and it's set up for a very nice year next year and beyond, not only because of the transitioning from NRZ to PAM, which again, I think Barclays did a really good job of calling this trend. I mean I remember a few years ago, when I left Maxim, and I was just reading your notes, and it was like PAM is happening. And then you guys called it pretty early, okay? And then Inphi ended up being the winner. But that's still in front of us, right? There's really 2 of the 4 of kind of meaningfully ramped, and there's still 2 of the remainders, plus you got China. And then you got the 400ZR ramped as well, which is going to be much larger than just kind of what was more of a one-off pioneering effort on COLORZ, one, right, with Microsoft. So a lot of exciting things. And I mean, I'll just say, even today, we announced the 50-gig front-end PAM4 chipset for fronthaul for 5G, right? As that market moves from 25-gig NRZ front end to 50-gig PAM, that's another industry transition that happen. So that ChIP-Atlas is out now and very, very competitive, right? So again, those are the kinds of things we're able to sort of pull together and get out even faster under the Marvell growth.

Lots of things to cover, and I do want to get to 5G, but maybe first, I want to talk on switching. It's really 2 questions. So one, you break out the segments, I think it was surprising to see enterprise grow. And I know you've been talking about the switching portfolio growing double digits for some time, but the market was down. I guess people are a little skeptical. But that business has actually seen a pretty meaningful acceleration in the last 2 quarters. So maybe just talk about, obviously, the refresh in the product portfolio, but then what are you seeing kind of for the mid-part of this year for that business to kind of rebound? Is that just people come back to the office, and it's just overall demand improving? Or are you seeing some acceleration of the share gains you were seeing before?

Yes. I think it's both. I mean I would say enterprise was one of the shining stars. Actually, if you look at our last earnings report. I mean, I think everything was up and everything looked good, but I'm incredibly proud of that team that really has its roots in original classic Marvell, both on the PHY side as well as the switching side, we did acquire Aquantia in 2019, and that's gone very well in terms of combining and having a much more beefed up PHY team. And if you remember, they were very strong in multi-gig. And so then you fast forward to like what's happening? Well, what's happening is companies are opening back up, hybrid -- I mean, Marvell is going through this right now, okay? We just made a bunch of campus upgrades. We're going to multi-gig, we're going to WiFi 6, right? We're getting everybody ready to come in the office. We continue to buy more storage, networking and compute, like on the -- this would be on the enterprise data center side. But still, there's a spend going on for sure. So that's coming back. I think based on the quality and the performance of our products and our sustained investment in enterprise for the last 5 years, we've also had a number of customers move over to us, right, in terms of their longer-term solutions. So there's been some share movement as well that has been a tailwind for us. And look, Blayne, you don't put up 50% year-over-year growth by accident. It didn't happen last week. This has been part of the plan. We've been trying to tell investors this enterprise thing is going to be better, even better than the teams we're talking about. And now it's showing up, not only in Q3, the Q4 year-over-year will also be very strong, almost all of it organic, by the way, and we anticipate enterprise will continue to be a strong growth driver for us on top of this run rate we're at in fiscal '23, next year. So there's more to come in enterprise. And again, it's a combination of new product ramps, content shift, especially as we talk about multi-gig and then also some of the share movement that we've seen to occur.

I want to ask you on the other side of the spectrum, why was this the right time to move in the data center? Because on the enterprise side, I think Cisco was kind of maybe going to open up some more merchant and you clearly were refreshing your portfolio that hadn't been refreshed for a while. So there's huge low-lying fruit, and then you were going to move that up to stack, which you've done. The data center is tricky. There's a lot of investments, leading edge, speeds and feeds. What made you think are this novum and this is the time that I need to move into that market?

Yes, it's a great question because, as you know, I've been -- was asked about that repeatedly for the prior 5 years before we bought it, right? Hey, when are you going to do a...

We're never happy, so we'll...

Keep coming, right? And then even when we acquired Cavium, we shut down XPliant. And again, that was more because, candidly, we didn't feel like that road map that the team, the product was -- it wasn't going to be able to compete. We're -- nothing else at Marvell, we're pragmatists, and we're realists. So we tend to look at everything, not just, okay, is it a technically good product, but what's the incumbency level? What's the customer sentiment? Like what's it really going to take? So what changed is really when we acquired Inphi, Blayne, and we really moved meaningfully into the heart of the networking inside the data center. And as you know, the interoperability and, I mean, literally, the chips directly interface with each other. The high end switch ships with the optical BSP. So there's a product synergy there, if you will, in terms of having that solution. That request from our customers never abated. They were always asking us to go do something there primarily because mostly for diversification reasons. But I think for us, 2 key things. One, it was the right time because of Inphi. Second, Innovium had gotten themselves to a point where they had strong traction at one large Tier 1 and then also a lot of traction with the others, but not as far along. And our view that, over time, copackaged optics, which is effectively moving some of the optical interconnecting components inside the switch ASIC would be a meaningfully important future trend. We really felt we needed to be there. And we have internal plans to do it ourselves, by the way. I mean, I've got a very good team, right? The thinking that's executed on its enterprise success, more than capable, okay? But it's a time issue. It's a cost issue. And Innovium's team excellent reputation in the market, excellent reputation, but small start up, can we bet on them or not. And so basically, we brought in the entire team got the leader, the teams come over. And so now we've actually added that. So there wasn't a lot of need -- we didn't need to do a lot of cost synergy on this one. We were able to actually lift the whole team in. So now we get several hundred people that have been doing this for a living for the last 7 years. And so we can now get it into the Marvell product development flywheel, and that's why it's a good time for us.

I know we run out of time, and I said I would get 5G, but I want to ask on auto because I think it's very interesting. And we took Aquantia public. I remember the year -- they give away basically a year of profitability because they said this auto is a big opportunity we need to invest. And with no revenue. We're just -- well, we have to do this because we have leadership. And now obviously, you've bought them and you're benefiting from this great ramp. Has anything changed in terms of the way auto, I'm just thinking about Ethernet because it was always a great story, but autos move, OEMs move slow. And it was going to start with like the core backbone and kind of eventually move out to some of the nodes. Has the market changed? Because you definitely are seeing a strong ramp for your Auto Ethernet.

Oh, for sure. And just to be super fair to my own team, there is very little contribution from Aquantia today. That's all in front of us. Remember, their focus was really multi-gig plane, which is where we're going to see a product transition in the future. It's a little bit like the enterprise. It took a long time to go from Gig E to multi-gig. It won't take that long in automotive. So this $140 million run rate we're on. This is team Marvell. Okay, this is team Marvell. There is no Aquantia in this. They're very little. The Aquantia success is really on the enterprise. Now what that gave us, though, was a very capable team that has invested there. They had a very good architecture. It was incremental R&D resources. But we definitely had to go take that sort of Aquantia architecture and take their really, really talented folks, but actually productize the product in the Marvell automotive flow. Okay. So multi-gigs in front of us. And the reason why the adoption is happening, it is happening much faster. It is happening much faster. There's a proliferation of sensors occurring in the car. The auto OEMS know basically #1 strategic decision to make in the future is their compute. What's -- how are they going to do? The second, what they'll tell you, almost as important is their network architecture. Because if they scrip the network architecture, then they can actually leverage the compute assets and the sensors. So this is really important stuff, and there's been a -- and then as sort of the newer cars have been selling better, to be honest with you, as OEMs have kind of reprioritized. We've just seen the adoption take off even faster. So we said it was on this much higher run rate. But if you look out into next year, automotive is going to be a very, very strong grower for us, all on the back of content gain and new product ramps, not really needing a big star kind of change to make that.

And I want to ask you maybe the final question, just the broader TAM within autos. Ashish mentioned the auto optimized compute wasn't something on my radar. I just finished up the fireside with Cristiano at Qualcomm, and there is a change. I think these cars are much more technical than they were in the past, and you might need different suppliers to enable all that. It's a bigger opportunity from our...

This change, Blayne, will be the equivalent of -- if you went back to 2002 or 2003. And you said, well, is Maxim or TI or NVIDIA or name all the infotainment players, right, that sort of did really well on autos, right, call it in the 2005 to 2015, a 10-year run, right? It was a boom for people, infotainment. Who are not the traditional suppliers and the auto OEMs and I was at Maxim at the time, right? And we had Jack said automotive is a 4-letter word. And if you design in, I will banish you. And then he sort of got religion and said, "Oh my God, they actually want our stock, right, because of the electronics pool." So the same thing, but on a much larger scale, Blayne, is going to occur -- is occurring now and will occur over the next 10 years, which is the data-centric the compute, the networking, those types of companies with that IP that's where the biggest semi growth is going to be bar none, period in automotive over the next 10 years. And so companies like the Qualcomm, like Marvell, like NVIDIA, like others, will be the ones that do really well, not in the infotainment, although they'll have some. I mean, that's really where Qualcomm sort of is today. And there's no winner-take-all either. I think there's going to be a proliferation of architectures, different ways people want to skin the cat. I think a whole bunch of companies are going to be able to win, win big, and it's the real trend for sure. And with our compute assets, think about OCTEON, Blayne. I mean it's actually -- it's a highly data path focused, optimized product for type of high-speed processing. And we are very aggressively investing in 5-nanometer automotive flow in 5a, which would have been automotivize. And we got a 5 road map we can bring in over time to 5, and we can integrate it all. And it's -- we have a long-term view here. It's pretty exciting.

By the way, Blayne. I think that's the other cool part. There's not even a theoretical exercise, right? This is -- we bought the first big design in under our belt, that chips in development. There's decamillions of NRE dollars already flowing our way. So this thing is already -- it's on its path now. Obviously, we are now talking to a bunch of additional customers. But yes, this thing is moving pretty quickly now.

Yes. Do you want to get a 5G in one at the annual, Blayne?

You got a minute. I was just going to sum it up and say, we didn't cover DPUs, we didn't cover 5G. We didn't cover storage, but I'm assuming you'd say all of those are good trends, too.

Afraid not so.

Absolutely. Look, let's end it on this. 5G guided up 30% sequential Q3 to Q4.

Very sorry. It's been fun to watch. I appreciate you joining us once again, definitely and to all listening. Thanks for joining the whole TMT conference.

Yes, great. Thanks, Blayne, for having us in the conference, everybody. Thanks.