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

Good afternoon, everyone, and welcome to day 2 at the Stifel 2021 Virtual Cross Sector Insights Conference and the Marvell fireside chat session. My name is Tore Svanberg, and I'm a senior analyst covering analog, connectivity and processor semiconductors at Stifel. It is my great pleasure to introduce Marvell management this afternoon. With us from the company, we have Matt Murphy, who is the company's President and Chief Executive Officer; and we also have Ashish Saran, who is Vice President of Investor Relations. The format for this particular session is a fireside chat, but we do invite you to ask questions. And you can type them into the chat function, and I will address them as we go along. With that, let's just jump straight into it. Matt and Ashish, thank you so much for joining us again at Stifel CSI this year, and welcome.

Yes. Thank you, Tore. Thanks for having us.

Absolutely. So I always start with a few warm-up questions, Matt. So we'll start with those. The first one is, what is it that makes Marvell such a unique investment for investors, especially when you consider there's so many other good large caps sitting companies out there?

Yes. I think the biggest differentiator is our focus. And what's interesting, Tore, is, and you followed the industry for a long time, the consolidation wave actually has driven a lot of conglomerate type of companies that are very, very impressive financial models, diversified benefits of scale. And I think that's been a good thing for the industry. What's gotten a little bit lost in some cases is the ability to really kind of pick your end markets, your segments and find companies that you want to invest in as a proxy. So you still sort of have that maybe in the handset world, right? If you look at one of the pure-play RF guys, if you want to play iPhones, you would buy Skyworks, although they're going down their diversification path and so forth. I think our uniqueness is if you want to bet on this, which I think is kind of an unquestionable secular trend of the growth and importance of data in the world, in data monetization, data as the new economy, right, the infrastructure of the world actually, as much as the physical roads and physical infrastructure is important, the digital infrastructure, you could argue is even more important. We are unique and that we are solely focused on that mission. And so if you believe and you want to invest in this infrastructure play, which we are really a proxy for across multi-markets, right, that could be -- and we'll get into these, but it could be in the wireless communications market. It could be in the cloud data center. It could be in the way that enterprises are going to transform post COVID in terms of how they architect their networks and even adjacently into automotive. I can't think of any other companies that have that level of focus. Everybody else seems like they've got another business. Well, we have this one business, but then we have this other thing. We're also in flat panel TVs, or I don't know, we're also in software, we're also in PCs or something. We -- our exposure there to the sort of noninfrastructure, like consumer-oriented part of the business, has -- is much different than it was. It's radically different than it was 5 years ago. And so I think if you want to bet on 5G, you want to bet on cloud and you want to bet on next-generation automotive as the -- some of the key growth drivers in the world, I think we're the single best investment you could make as a proxy for those, given not only that we're in those markets, but we're actually in them in a meaningful way in the case of cloud and wireless. And in the case of automotive, I think if you want to make a bet on the next big trend, right, which I would call the third big trend in automotive in the last 30 years, which would be the connectivity in the vehicle and the data-centric nature of that, you would want to bet on Marvell.

Great. No, I love the answer. Second warm-up question is, how has the world changed for Marvell over the pandemic? And I usually ask companies to pick maybe 2 or 3. And remember, I'm sure there's a lot of them, right? But maybe the way your customers behave, maybe inventory management, maybe the culture itself at Marvell. What are some of the 2 really transformational changes you have experienced at Marvell because of the pandemic?

Yes. Well, I think the first is, it's had a huge tailwind on our business, actually. I think, and that's because of the fundamental pull forward of digitization everywhere. And certainly, that was felt in cloud computing. And even as enterprises try to rethink how they're going to go about managing security and work from home and all these trends, it's only been good for us. It's only been good for Inphi also, by the way. I think what's a great sign as well is even in the 5G area, although there was literally human limitations, right, of putting base stations in the ground, right? You saw handsets flying off the shelves, right, with people basically voting with their wallets that this is going to be an important new technology. And now you're starting to see the catch up this year. So I think the first one would be that the effect of the pandemic to drive digitization and to drive digital communications in a much more meaningful way was a huge benefit to us, and that's going to just continue, right? Actually, that only pulled it up. It doesn't stop. We don't -- we're not writing some PC boom where everybody -- all of a sudden in the last 15 years, everybody decided to buy a computer in 1 year, right? I mean -- and I'm sure that market maybe has reasons why it will do better or not. But fundamentally, these are like x unit growth kind of markets that have suddenly gone berserk. We're not in those. We're in all the things that need to get built out. So that's been a transformational thing on our company from the outside. From the inside out, I think a couple of things. I think one is, we do -- I think one of the biggest challenges I had when I became CEO, which was actually about 5 years ago in July, okay? So I'm almost at 5 years. I think the single -- one of the most things I'm most proud of what we've accomplished is actually the cultural transformation in this company. In fact, when I took the job, I had a very close friend of mine, a peer and person I worked with for a long time and he said, "You can't fix it. It can't be fixed." And I said, "You're really betting against me on that one as well." If anybody I know could do it, I mean, I'm serious. And it was really -- because it was a different company that was very relevant for a different time. And for us to really participate in the scale of what's required and the kind of company you need, whether -- and by the way, the company was also mired in a whole bunch of issues, right? The question -- the integrity of the company, the ethics of the company were questioned, the team work -- some pretty fundamental issues, right? And to be honest, you can't get anything done, okay, if you don't have the right people, you don't have the right team. So I think putting in place the core values of the company, making aggressive changes where people weren't ready to really align with a much different approach. We turned the ship very fast, and the employees were willing to make the turn, actually amazingly. And as a result, what happened, Tore, is during the pandemic, we have built up such a bank of trust over the prior several years that we -- I think, if anything, because of that and because of the sentiment inside the company, the engagement of the employees, we outperformed. I mean we were taping out complex state-of-the-art 1,000 I/O processors on advanced node. We did the whole 5-nanometer platform development IP in the pandemic. Can you imagine? And the team did great. I have a lot of trust in them. That's why when we even look at like going back to work, if you were saying, "Oh, are you going to do a 180," and now -- and a lot of people -- I know they have different opinions about, okay, it's over, so everybody just get back in there. My view is we have -- we could trust the team. And so we can do a hybrid thing. We can -- so anyway, that's a huge -- I mean it never gets talked about, Tore, and I actually rarely get asked about it, but I just thought it's important to know. We've built a culture that's built to last in this place where people really buy in, and it's all around things that you would expect around respecting people, around executing our projects, around helping other people achieve their goals and being extremely rigorous and thorough and data-driven in everything we do. And it's a different deal than the old place. So I think that was very profound. And then what came with that, which would conclude the answer would be just all of the great execution that we had in terms of getting fundamental process technology in place, executing projects. And they're executing projects within days of commitment, like we're supposed to tape out December 20, and we just tape out December 22. I mean you go back in history, whether it's Marvell or Cavium or Aquantia, we have all the data. We've measured all of this. Everybody was 9 months late on everything. We're within striking distance on almost all our chips, plus or minus weeks at the worst, on average. So I think that's been a huge testament to why we've been able to put ourselves in this position.

That's a great perspective. So moving on to the business. And obviously, you guys reported on Monday, so I'm not going to ask you about the quarter. But the one thing that was pretty striking to me, something that you mentioned on the call is the opportunity in cloud. Cloud is becoming the biggest opportunity for Marvell. Obviously, with Inphi, you're doubling down on that. And by the way, I can't wait for you guys to report the new segments. It's going to be exciting next quarter. But yes, maybe we can start there. So why did you make that statement and especially now?

Yes. Well, this is part of a longer-term strategy we've had, right? I mean if you even go back, I think I was quoted after our first restructuring, when I took over as CEO at the end of 2016, and there was an article on -- some CNBC, it said, "CEO Murphy states Marvell is pivoting to the cloud." Okay, now we had a long road ahead of us story at that juncture, okay. If you look at how much cloud revenue we had at that time, it was very -- it was de minimis, okay? But that was my fundamental belief, whether it was in storage, right, and driving from consumer-based storage controllers to data center and enterprise and hyperscaler, custom controllers, we did all that, right? And we could talk about storage later, but I think you've seen the results. Same in our networking. My first meeting with the networking team and others was what's our strategy to move out of SMB and SOHO and at these higher-end core and aggregation switching and so forth. And then as we bought Cavium, they had a lot of enterprise business, but they had aspirations in the cloud with their DPUs. So every one of these acquisitions we've done, I've just moved where they were in terms of the wins that they had. And even Avera, by the way. They had very good 5G, and they had very good kind of enterprise wireline things. All of these businesses story, we've moved them to where they're in the design win zone now of all these major opportunities. And I think that pivot that market-based approach is significant. So last year, we actually said at this time that cloud had exceeded 10% of the company revenue, which, in my experience, is always a milestone, but I think is noteworthy if you're trying to really grow your way into a market that's got legs behind it. And then we said basically exiting the year, we were at low teens, right, as a percentage of the total. So not very respectable, right, from a last year perspective, so call it $3 billion of Marvell revenue, low teens in cloud. Inphi, there was a huge bearing on our decision to do Inphi was basically they're well north of 50%, probably closer to a 6 handle in terms of their cloud concentration, right? And that revenue is growing like crazy. So that adds significant help. So I think when you add it all up, Tore, you're going to see a significant cloud -- broader data center revenue position, right, which would include all the enterprise on-prem as well as cloud. That bigger bucket will be very substantial with the cloud portion being the majority of it and growing at a much faster rate. And so that's the end markets coming together. And we've been able to -- as a result of that, I think at doing Inphi, we've been able to close a number of significant opportunities now that we were pivoting to and tracking and chasing. And now we've gotten design awards and wins that we're very, very excited about. They're very -- all these companies, obviously, you have to respect the confidentiality that they request in those contracts. So this current juncture that we're a little bit limited, but just to kind of size it, if you take that revenue we did last year in cloud as just a Marvell stand-alone, the incremental opportunities we've identified are in line or more just in that amount alone. So the incrementals were sort of what we did all of last year. And then you've got Inphi layering on top of it. And I just give one quick plug for Inphi, and you may come to it, but their revenue at $215 million that we guided for Q2 is up 23% year-over-year. But if you remember, they had a Huawei contribution in that base from a year ago. So if you strip that out, just big round numbers, it's about 30% growth. And you can imagine that within that, the cloud piece is growing faster than the total. So anyway, you get the sense, right, that this is a big opportunity. The TAM is over 2x what the 5G is. And so we should be very much viewed, in my opinion, as the cloud and 5G company that you want to bet on, that's got kind of the growth all in front of it. I'm not saying there's not other good cloud properties out there. There's phenomenal companies. There's so much innovation going on in the space. But I just think the next few years really is going to be an exciting time at Marvell, and we're excited to innovate along with our customers on that, with leveraging our resources to help build really what they want.

Right. And I mean when I think about cloud and I think about cloud over the next 5 to 10 years, I think about 2 really fundamental trends, right? One is disaggregation of the architecture itself within the data center. Because obviously, there's so much garbage processing that's happening today. So I mean disaggregation has to happen. The other thing, of course, is disaggregation of the data center itself, meaning more of a DCI-type architecture where these big mega data centers are prohibitively too expensive and going down that path isn't just going to work out anymore. So when you think about those 2 really important trends, how would you position Marvell within both of those? And obviously, including Inphi now in the discussion?

Yes. No, I think we're big believers, obviously, architecturally on the disaggregation that's happening. And that's a very broad catch word now because there's literally -- you could say there's a dynamic that's around disaggregating compute and storage and networking, right? That's one dynamic which creates change and requires different solutions. You could also say just kind of monolithically, if you went back 10 years, everything was done on x86 and an industry standard server and people just bought those, and that was sort of what you did. And you fast forward, and that's where we're seeing so much activity, particularly on the custom side, is that these large companies have concluded because they're running their own workloads and they have their own demands on them that it's sort of like everyone, there was a period of time where it was, hey, software is eating the world, and you really don't sort of need this hardware anymore. It's just totally not true. I mean you basically need both, right? But you really need -- so if you think about sort of offload, right, or processing that's done in a very specialized manner off of the kind of core compute, that's -- and you see it now, right? You see cloud companies coming out with big announcements of their own ships, right, that have their own specialized things. And those are the kinds of things we're good at partnering on with people, right, to help them do that. We're not -- and by the way, we're happy to not be the front person, right? We're happy to work with them. They put their stamp on it. We work with their engineering team. We -- we're fine to be odds. We're the guy behind the guy. That's okay. We have a lot of capability to go drive all that. And it's created -- and by the way, even the notion that the core CPU itself needs to be x86, kind of our fundamental view is that, that market, in aggregate, is probably kind of flat and there's a share battle, AMD, Intel, it grows, but it doesn't sort of have that -- I think all the legs on the processing are going to be in very specialized types of hardware and accelerators to get the job done, and it's certainly not happening in FPGAs. I mean that's just not -- that's the way to get it sort of prototyped, as it's been for the -- since sort of the dawn of man, both and -- but once it goes to high volume, it needs to be a lot more specialized. So that's creating a lot of opportunities. And then to your point about all those chips now need high-speed optics to connect all the servers together, so that's the intra data center where it's really -- I think Inphi is a very unique asset there, really only probably one company that can actually compete sort of with the scale and technology cadence. And that's okay being 1 of 2. And no problem. We're used to that, but it's very complex stuff. As you know, you've covered analog forever. But then also the idea that you can now effectively virtualize multiple data centers through DCI and have 2 look like 1 or multiples look like 1. It's a very powerful concept. And Microsoft was a pioneer here within Inphi on COLORZ. That was a very public announcement that, I think, drove the industry. That was really 1 customer, 1 particular technology, which was 100-gig based on a PAM-based DSP and modulation technique 400ZR and ZR+, which is the 400-gigabit standard, which is now ramping into production this year at a much faster rate than we thought both this year and next year. That's going to be multiple OEMs. It's broadly adopted. And it's based on a coherent DSP, which again, Inphi had through its acquisition of ClariPhy. So it's really unique what we ended up with, with the Inphi asset, both connecting data centers, inside data centers and then the sort of core IPs around really complex advanced digital DSPs coupled with state-of-the-art rocket science-type broadband analog chips. And even we have our own silicon photonics. We're high volume in silicon photonics. One of the few companies in the world. I can say we're shipping and it yields and it works. It's not a PowerPoint.

Yes. And that's the whole thing, right? I mean if you think about your acquisitions and what you have today as you continue to target cloud, I mean you have ASICs, you have custom processors. You have network processors. You have obviously all the storage. And now you have a PAM4 DSP. You've got a coherent DSP. You've got high-speed analog. You've got photonics. So do you have everything now? Or is there anything else that you think you need to add to your arsenal?

Well, look, we're -- we've been -- and I appreciate the comments. And certainly, the M&A has been a big part of our story. But remember, the organic investments that this company has made, the ones that I decided to bet on have performed really well, okay? So I'm not saying these are cloud necessarily. Well, some of them are. If you look at our storage business, right, that's certainly been a bright spot. It's growing double digits now in aggregate. Why? Because we pivoted all in the data center. And then you can look at the success in our Ethernet business for enterprise and campus and kind of higher-level switching on PHYs. That's done extremely well, too. So we have a number of investments and things that we look at, Tore, because as to your point about disaggregation, there's also all kinds of pull-through trends that are happening where there's things that we can go off and go do organically, right? Whether that's leveraging optics, whether that's leveraging high-speed interconnect, there's just a whole bunch of new things. So we feel if we just step back, right, we believe we have one of the best portfolios you can find, spanning all of those different technologies you mentioned in processing, in networking, in security, by the way, and in storage. And we feel really good about it. And the other thing I'd say is these are not just who we have the product. I mean we just went through OFC, right? There's a lot of companies that put out some press release and say, "Hey, we made one." And I learned this a long time ago at Maxim, just because you make a part, it doesn't mean you sell a part. You make a part, you could do a press release, you can talk about it. But can you actually sell it? And so I feel really good about our competitiveness. This company is in fighting shape, okay? All of our products now have leadership product definition, leadership technology, great support, plus just a great sort of go-to-market customer engagement model that I think is customers find refreshing. And so it's all those things that had to come together, like all the spinning plates over the last few years. But I think for the cloud, we have -- we should be a very good supplier into that market and be viewed that way.

Right. Yes. And I feel like I could sit here and talk to you all day long about the cloud market. But it's just the time I got to move on. And I do have a few specific topics I wanted to ask you about. The next one is ORAN. And you mentioned it a few times on earnings calls. Obviously, it's still very early days. But I do still think there's some big question marks out there as far as how 5G infrastructure is going to evolve from a standards perspective. Is ORAN really the response longer term, especially when you start thinking about like how all these different geographies now becoming very protective of their networks and so on and so forth?

Yes. I think -- well, it's got a lot of traction, that's for sure, both on sort of what you would call open-source industry standard base stations, right? That's sort of one model? Can you effectively -- the dream would be to white box it, right? That's what -- now the reality is, obviously, is you follow telecom for a long time. These things do take time, but it's got real leg story. And in some ways, I think people view that as the answer to protectionism, that, "Hey, I don't want to get locked into a certain vendor, from a certain country that might have some issue with my country." And then -- so there's -- I mean, that's actually the way under the Trump administration, that administration was going was like, "Hey, why don't we just go drive ORAN faster, right, and make it -- and then we can kind of control all the hops in between and all the software and get the telcos to join in." But it's got legs. There's also the aspect of it, which is under the bigger catchall of ORAN, which is vRAN, which is, I think, another trend where you've seen a number of hyperscale companies actually come out and make pretty big announcements, right, with telcos or emerging players like DISH. And so I think the notion that you can -- that Layer 1 baseband processing is going to move into the cloud, that's a very real trend as well. And I think that's a very disruptive trend if some of the cloud titans can execute because then all of a sudden, you've got this massive infrastructure on the CU side and you can augment -- by the way, the same deal, you don't want to run all that on x86. There's a reason why people in macro cells use custom base bands, right, because that's how you get the job done the most efficiently. So I think that's a major disruption that is sort of worth paying attention to. It's how does -- it's not just about open source, it's also about driving kind of cloud scale capability into the RAN. And I think there's an opportunity for probably the cloud titans to figure that out to help the telcos probably offset some OpEx, right, make their networks more efficient or even enable new entrants. So it's an exciting space. We got our bases covered. I think that's what you should know. We've got these really deep partnerships with the big companies that you know about, and we value those immensely, but we're not going to also get trapped in the innovator's dilemma and go, oh, no, what's going to happen. So we've also put together a very competitive suite of ORAN offerings that are more open, right, for others to go buy. And we're selling those in the form of a card, right? We sell chips too, but think of it as an accelerator card, you could plug right into a server. And obviously, it's not that simple, but that's the route to enable this type of baseband processing.

Yes. Tore, I think the takeaway is we've basically taken our existing portfolio, we made it ORAN and vRAN ready. And the big advantage for our existing customers is because, we have these -- all of these is going to exist. There will be people who want classic RAN, but they also want ORAN and they also want vRAN. By using our solutions, right, it's the same software stack we invested in. You don't have to create now a second product line, right? If you're a Samsung or Nokia saying, okay, I need to have an ORAN player, I got to go reinvest all of that. We've already done all that work for you, right? And we're also the only kind of merchant baseband provider. So if you're a regional player or some of the start-ups, as Matt mentioned, we'll kind of get into the space. Your only choice is to start with an FPGA, which is highly inefficient, right? So now you actually have a dedicated 5G baseband processor, which is available, and it's already ORAN and vRAN ready. I think that's really the big takeaway is that we can address all of these different go-to-markets.

Right. Right. Okay. So I know we're running out of time, but I wanted to address automotive, Matt, because that's the 1 segment that investors don't focus a lot on with Marvell. And obviously, you have some great Ethernet technology there. But you also have a lot of other IP and a lot of other know-how, right? So not to steal the thunder of your next Analyst Day or whatever, but I mean what else or what more can you be doing in automotive given this strong IP arsenal that you had?

Yes. No, great question. And I think for the long-term investors, the way to think about it is we've got this 5G wave happening now. And I think we've been saying that and we've been generally executing to what we said. And I think it's -- we feel confident -- it's well understood, that's going to go well, right? We're in the right designs. 5G is actually happening. So that's great. We talked about cloud, right? That's kind of on the heels of this. And I'm not saying -- by the way, 5G is going to be a 10-year cycle, okay? So this isn't -- but you sort of wonder, well, what's next, right? So at some point, 5G will get to its sort of peak run rate. That's still years in front of us, okay, before that happens. But then you've got this cloud kicking in now and then what I would call a step-up once these new designs are in production in the '23 to '25 time frame. And of course, Inphi will grow through that in our existing things. So they say, okay, well, then what's after that, right? And I say, if you look, I may give you high-level numbers. The largest projected end market for semiconductors, based on some analyst research, you could do your own, by 2040, okay, we have a 20-year horizon, it's going to be automotive, most likely. Units aren't going to change a ton, but the content, as you know, it's completely becoming digital. And the way I think about it is sort of epochs, right? You have this sort of wave of the electronics and embody, okay, and sort of transmission control and traditional auto. That was NXP, ST, Infineon, in their former instances, Motorola, Siemens, Philips, et cetera. There's this wave in the 2000s, which I was a part of at Maxim, which was really the infotainment revolution and all the companies with the consumer-based IP, whether that was NVIDIA, right, with Tegra or it was Maxim with all of our power management display and other companies. That was a booming time, right, for the automotive industry and suppliers who hit that inflection point. The third big wave of this thing, Tore, is going to be the data-centric opportunity. That's the next big 10-year kind of plus run. Not that the others won't still be needed to do what you do. And that's, I think, where we come in, right? Because every conversation we're having with automotive is around networking, which we're already in. We have design wins with 24 different OEMs. That business, we had said, was going to exit this year at about $100 million run rate. It was approximately 0 5 years ago, and we invested ahead of the curve. It will probably get there earlier in terms of the ramp that it's going through. That's pretty much as predicted, although the SAR and all the other things that are happening is giving it more of a lift, but that's a new $100 million type business that will be realized this year on a run rate basis, and there's line of sight to do $200 million and then beyond that. But then the exciting thing is, we're in a lot of discussions now with our 5-nanometer platform around custom chips, compute, different architectures for how you think about doing electric vehicles versus traditional combustion engine cars where you're stuck with 100 ECUs and you've got all these legacy complexities. The EV guys can do a clean sheet. And so their in-car networks look a lot like a data center. And disaggregation is a theme. And how do you actually -- or centralization, right? In the case of storage, there's a whole architecture we're proposing to centralize storage, right, using Ethernet, security, right? Our HSM that we sell into all the cloud and leading sort of companies for key management, that's going to be a critical thing, as you connect every car up to 5G, you're going to want to have some kind of automotive security. So storage, networking, computing, processing, those are all of our key IPs. And we are very much going to be on the TSMC N5A flow, which is an automotive grade flow that we're right at the front end of. So there's more to come on that, and this is a longer-term journey. I know exactly how to message it, size it, wait. I'm super patient. By the way, I invested in this business in automotive, my final statement. 2016, losing money, Board said shut it down, what is this automotive thing. It's horrible margins. Nobody -- actually, I was the only person the company has supported it, me and like the product line guy. And I just -- I knew from my experience that Ethernet was a real trend and that having network. So I kept it alive. We bolstered it with Aquantia, by the way. And now we've got this business that's got technology leadership, market share leadership in terms of the high-end stuff we're doing for the future. And so we feel really good. We can now -- because we're in some of the OEMs, we can talk to everybody now. And they're big enough to understand. Hey, remember back in 2004 when all the infotainment guys, we brought in all these new suppliers, and we had to go train them all how not to do automotive. That's what they're looking at for the data-centric ones. The difference is I had the experience and we had the experience to put in place all the necessary measures. Tore, we're shipping at 1 DPPM right now in our automotive chips or less out the shoot. Tens of millions of units already gone through the system. I mean, no -- like 1 fail -- it's really good. The team has done an amazing job. So we're well set up to culturally to really participate. This isn't going to be some learning curve crazy thing, what do we do? How do you do automotive? We put in place the people, the systems. So I think we're probably the best positioned company to take advantage of that.

Fascinating. Very good. Well, with that, we've run out of time. Matt and Ashish, thank you so much for doing this. Like I said, I could sit here and talk to you all day long, Matt, because your insights are amazing. So thank you again. And for everyone...

Well, we have to grab a beer now that we're...

We have to do that. Right. Exactly right.

I think the Dutch Goose is gone now, maybe, but we can meet in Menlo Park somewhere. I think the Goose is there. It's the Oasis that's gone. Anyway, it's great to see everybody. Tore, thanks for the time. Appreciate it.

Thank you, everyone. Have a good afternoon. Take care. Bye-bye. Thanks. Bye.