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

Hello. Hi. I'm Tim Arcuri. I'm the semiconductor and semi equipment analyst here at UBS. And for this session, we're very, very happy to have Marvell. Fresh off with a great quarter last week. We have Matt Murphy, the CEO; and we have Ashish Saran, the VP of Investor Relations. And so just very, very happy to have both of you. So thank you, first of all.

Thanks, Tim.

Yes. Thanks, Tim. Thanks for having us.

So I guess, I mean, first, Matt, the obvious question, great, great quarter last week. The stock was up huge. Maybe the big question that I got, and I sort of want you to sort of riff on what the drivers are for the growth that you're seeing. But the big question I got on Friday was, well, how sustainable is this? And when you look beyond fiscal '23, can they grow at 15% to 20% CAGR even in '24 or '25? So there's -- I mean it's the inevitable question when you get a year where you're growing about 30% and you have tough comps. And so maybe you can talk about all of that.

Sure. Yes, do you want -- I'll start off, Tim. So it's great to see everybody. And yes, it was a great set of numbers and outlook. We're super excited. I was telling Tim before we started. The employees of the company are beyond motivated and going to continue to really drive hard with all this opportunity in front of us. I think it's a first-class question to have, Tim, which is, "Hey, next year looks great. What about the year after," okay? So while we typically don't guide like years away, I think thematically, we laid it out pretty good at the Investor Day, which certainly had a longer-term 15% to 20% CAGR. We're obviously in a period of even stronger growth than that. But it certainly continues beyond calendar '23 for a number of reasons. The first is, if you just think of that basket of the high-growth initiatives that were a part of the cloud, 5G and then automotive, right, those 3 segments. We had said at the Investor Day that, that we believe we could grow that at twice the rate of the market, okay? So -- which was 20%, so you can kind of take it and say 40% a year plus. Next year, it will be stronger than that. That's what's helping the 30% year-over-year for the company. But that doesn't abate just because you get through one more year. In fact, if you look at the cloud, we have the incremental design wins that we've talked about, Tim, that were, one, call it, earlier this year in the custom arena. A number of those, which we had given a bogey of $400 million of incremental revenue from those in calendar -- in our fiscal '24, and then that actually double into $800 million in fiscal '25. So you've got the cloud momentum. So think about the base cloud is already growing, and then you've got these incremental wins on top of it that we never had before. These are brand-new sockets for Marvell. 5G will continue to grow as content gains start to layer in with one of our customers, which is Nokia, but also as these other geographies ex China continue to deploy, that will be really good for us given our presence with everybody except Huawei. India will be -- India hasn't even deployed yet as an example. So that's coming in front of us. And we have all the ORAN/vRAN content gain, which was really with sort of the cloud hyperscale type of companies. So that's all still in front of us as well on 5G. And then auto, I mean, this business has been doubling basically on an annual basis. We were really pleased to call it out not only clicking over the $100 million run rate but kind of pushing the $140 million run rate. So -- and that's got very strong momentum. And we outlined kind of a path between here and $500 million a year of revenue, call it, 5 years from now or something. So there's a line in between. So if you just sort of assume that, that happens and we can debate it or not and we can dive into different questions, then the rest of the business doesn't have to grow that crazy. Really what's left is you've got enterprise, which certainly we don't think we can probably keep up 50%, 60% a year year-over-year growth, but it's not going to end. That's going to continue to grow at a really healthy clip for the next few years based on content gains, new design wins and share shift. And then finally, you've got a consumer business, which for the Investor Day, Tim, we really modeled it very conservatively. But look, it's up, what, Ashish, double digits? No, it's up 20% year-over-year in Q3. It will be up again double digits in Q4, and consumer as a segment will be up next year because the PlayStation SSD win. And so that's got some legs in front of it. I wouldn't model as a long-term grower but not a shrinker. So you kind of have -- all the markets are basically performing really well. And I think this final point, we kind of managed to get rid of either reduce investment or even divest pieces of the business that were not growing or shrinking. And so we don't have many of those anymore. And from an end market perspective, you can see it. Every end market was up in Q3. All those markets will be up in Q4, and they'll all be up in fiscal 2023 next year.

Yes. I mean just to add a little bit. I think, Tim, fundamentally, I think what we're exposed to is the digitization of the economy of the world. You've got the whole excitement on virtual environments. I think that's what -- if you stand back and think about the moves which Marvell has made has really given us all the key assets, which all tie to data, data infrastructure. And I'd argue that's probably the most secular long-term growth driver by just fundamentally. So while, obviously, there's near-term demand growth coming out of the pandemic, I think the pandemic, to some extent, and we said this before, this is accelerating things which are already happening, right? And I think you see a lot more of that. I think this thing really has legs, certainly feels like that from our driver seat.

Great. Great. I guess, Matt, one other question that I was getting and I know that you've guided to is are your efforts to sort of scrub the backlog and to really determine the degree to which customers are placing orders because lead times are long and maybe customers are buffering product to basically have it on hand when they can get another component. So do you have any sense of this? And maybe can you walk through the logistics of sort of how you do that?

Sure. Yes, we're very focused on it, Tim. I guess I would say, obviously, but we are. And as you know, I've been around the block in this industry and seen these cycles come and go and had gotten bitten a fair number of times in the process. So I think there's a few things that we do. And I would frame this by saying, we are fortunate at Marvell because we have a very proprietary sticky footprint, okay? These are typically, I mean, almost exclusively sole-source type of arrangements we have. Whether it's the SoC that powers the entire system or even if it's a complex mixed-signal product, like a PHY as an example, you might think that you could swap it out, but you actually can't. It's just -- the recalls are very, very time-consuming and difficult. So I think we have that benefit going for us. And so we do a couple of things. One is we have very -- we built up a strong capability relative to our ability to understand the end markets that we sell into. So we've actually built a very strong data analytics capability. And I'll give you an example. I do a meeting every quarter. It's just one example, but you can extrapolate this, where Ashish is in this meeting, too. We actually go through in the 5G market. Every base station being sold, where are they going, which telecom operators, what share is going to which vendor, even the block diagram of the digital unit, the RU, by account, how many channel cards? How many OCTEON's infusions per card. What's the average number of cards per chassis? What's the average number of our use per system by geography and then kind of mapping in where share rolling off, where are we coming in. And we basically net all this out to a -- back to a Gartner trend-focused level sort of assessment. So we don't really quite take the bate of, "Hey, I'm upsiding. I need more product." Okay. Give us the context for that, right? So our -- so you can imagine we do that across our enterprise business, our automotive business. I mean we know how many cars we're going to be in next year, by which manufacturer. We know kind of the configurations in the SKU. So that's one backdrop we used, Tim. The second is that the backlog is almost exclusively secured as noncancelable and nonreturnable. And we have -- in general, we've been getting almost kind of 1 year kind of like orders. So the book-to-bill has been off the charts for the last few quarters and continues to be very strong as that backlog has been laid in. So that's one thing as investors, so how do you know? It's -- well, we have orders and people have committed to take them. And then we bounce it up against our market-based forecast. And we don't -- and then I guess the final thing I would say for the investors in Marvell relative to others -- so I guess the bad news story has been for the last 6 or 9 months, hey, we were kind of behind relative to some of the other suppliers because some people had better inventory, and our products are kind of more complex. So whatever the reasons are, right, you could say, "Well, we had to make some changes in our operations team to get more strategic." So we were a little bit behind relative to others. The good news -- so that constrained us. The good news story is we're probably more constrained than others. So we haven't, in my view, overshipped into the end market, if you will. I mean we're like hand to mouth, right? And I mean one other final story. I remember I had a CEO call me last quarter asking for more allocation on a particular product. And he said, look -- and his team was asking for all these numbers. And so "Look, I'm just going to break it down for you. I have demand for 25,000 of these boxes a quarter," whatever this box was. "I can get 15,000 units a quarter worth of product for that box. And you're giving me 4,000. So can you get me 10,000?" So -- but if somebody is shipping the 25,000, you follow me, and he needed 15,000 and we only -- now of course, we ended up closing the gap and getting basically, call it, 15,000. So what I'm saying somebody else is shipping the extra 10,000, and this is the worry everybody has. And I'm not claiming that we're totally immune because you don't know until you're done, but we have been highly constrained. And all these allocations are being run at a very high level in the company because you're having to suss through all the issues. And it's so precious, the material, that it gets a fair level of scrutiny.

And Tim, I mean, just to add just a little bit more comfort around next year. I think even when we -- and I know it sounds crazy. But even if you look at our 30%-plus kind of target revenue growth we discussed at the call and bounce that up against kind of the bottoms-up forecast, right, from customers and sales, we don't actually still eat into any of that, right? So even next year, the growth isn't because, oh, because you've got base growth and now you're finally eating into your delinquency. That's why you're growing at higher than market. No, it's because we still don't close the gap, right? So that is a discussion almost for another day. There just isn't capacity out there.

That's super, super helpful background. I guess on the delinquency side. I know, Matt, you had talked about that last quarter. And then I asked about it on this call thinking that, well, maybe delinquencies got better because the supply situation got better. But they really didn't get better because the bookings just took up the slack. So delinquencies didn't really improve. So I guess sussing through what Ashish just said that even throughout fiscal '23, you don't think that delinquencies will materially improve, and that's already assumed in your up 30% guidance.

Right. 100% correct. The outlook we gave does not comprehend any improvement on the delinquency, which is that we'll exit -- we'll -- grew every quarter this year, and we'll grow again in Q4. And so we're not -- we're looking at next year, and that's what Ashish was saying. We've got end-market demand that's actually above kind of what we're saying we think we can do. And then that -- so that's not even captured. And then there's the sort of moving massive delinquency that keeps growing every quarter, unfulfilled backlogs and other term people use that just isn't shipping. So that's -- even if some things, let's say, don't materialize, it's sort of like you just reallocate. So that's what gave us -- some of them say, why did you give some -- like I guess people said, "Why did you lean in so much?" And our view was, to be honest, we felt like we owed it to our investors to contextualize the current kind of growth rates we're seeing and the fact that they are going to continue next year and that actually, there's even legs beyond that. And there's enough moving pieces that we feel very comfortable about where we're heading. We're going to drive -- we're driving hard, obviously. I mean, I'm showing that delinquency number personally to all of my suppliers, right, and saying, "Look, this is way you should bet on us, too. Not only do we have a nice forecast for you next year. I've got a backlog I can go ship right now. Like just allocate it to me. I've got the orders. I can show it to you. And this is how much is yours." And when suppliers see us it's astonishing. Okay. Now let me go work on. Because it's -- forecast and backlog is a very different situation for the supply base, obviously.

Yes. Well, that's amazing. One more question, Matt, that I was thinking about -- I haven't been asked about it too much, but I was thinking about it. If you look at the incremental revenue that's dropping to the model, year-over-year, you're dropping $550 million, give or take, to the model. And OpEx is up only about $100 million. So it's -- so the incremental OpEx is only about 20% of revenue, where the model is 27% longer term and you're running higher than that right now from an absolute company perspective. So the incremental drop-through from an operating margin perspective is huge. And so there was some -- I don't know -- I don't want to say concern, but there were some questions about, okay, well, how long can Marvell continue to drop 20% of incremental revenue through on the outline? And at some point, does the investment have to try and back up?

Yes. Well, I mean, a couple of things. One is, I think we've been waiting for this for a long time. So to me, some of this is a little bit of catch-up, right? We definitely overinvested when our former President lid off the trade tweet war and blew up 2019 for a lot of people. That was a rough year. We were down. We kept investing. So some of this is a little bit of catch-up for us. But I think it's a valid one. Look, I think, Tim, we did guide OpEx up kind of organically high single digits, right, which is a meaningful number relative to where we've been. We expect the growth certainly is going to get a couple of years in front of us. So we'll probably keep scaling the expenses up. I guess the way to think about it is nothing's changed with us or me, meaning we're investing for the long term, okay? So I think if there's a concern, got it. These guys are going to start milking the cow now. To be honest, the revenue is moving so fast and the drop-through, you can actually hire fast enough. So some of this on the OpEx side is also a bit constrained candidly. I mean I'm running a meeting every 2 weeks on recruiting. There's a record number of open recs in the company. Our opportunity pipeline continues to be very, very strong. So I think -- I don't know if I'm helping in this answer and, Ashish, you can jump in. But the way we're thinking about it is we leaned in on OpEx next year. There's still a phenomenal drop-through. But if you kind of play out to '24, and I'm talking about fiscal now, fiscal '24-'25, where that long-term outlook still looks really good, we'll continue to kind of drive both. We can get revenue and margin growth, and we can continue to invest for the future. But I think you should -- we really -- we're not updating the margin model. At some point, we said 38% to 40% is kind of the long-term range. So we'll be putting the investment in to keep that. You know what I mean? I'd rather drive top line, to be honest with you, and build a big gigantic successful company with very good profitability. But it's -- the growth story is really unique that we've kind of got now, and we want to continue that.

Yes, I think, I mean we made a commitment, hey, look, we set a target -- and our long-term targets are not like 5-year targets. We tend to think in kind of more reasonable time frames. So hey, we need to get to 38% to 40%. We updated that. I think we're making good progress. Still we get there, revenue growth will be far higher than [ OpEx cost ] for a number of reasons. Also, remember, we have a very good platform order, which will really get the leverage and we can certainly talk about that. A lot of our revenue growth is selling more of the similar stuff to similar customers, right? A huge lever in OCTEON, a huge lever in Inphi, right, and there's a number of examples. Once we get to 38% to 40%, right, and you're still growing top line in that somewhere in the 10% to 20%-plus, there'll be a pretty fun discussion at that point, right? And as Matt said, I think it gives us the ability -- there is -- just back to my comment earlier on digitization of the world, this trend is just going to be massive, right? I think we have all the assets, and now we have the firepower, right, to go after a lot of opportunities. So I think you should expect us to remain very focused on investment and a lot of organic investment, quite frankly, because we've already done all the hard work of kind of doing all the surgery we had to do, right? I think a lot of that's done. So a lot of that now is really about -- we have all the capabilities. We've got customers who want access to it. It's -- the design end funnel is just I haven't seen anything like this for a long time. So yes, I'm pretty excited.

Yes. And as you said, you talked about the custom stuff, and you made the decision on Thunder to go custom. And can you just talk about sort of what makes Marvell unique? And this is a question that I get a lot, so I assume you do, too. But what makes Marvell unique? If you're a big cloud customer, and you -- and of course, they want to drive more customization, all of them do, what makes Marvell a unique partner? Maybe you can just riff on that a little bit.

Sure. I mean -- by the way, I would argue this is not just a hyperscale story. I mean I think the supply is very well in automotive. And candidly, Tim, I think this has been part of our playbook in the wireless area as well. But I think a couple of things. I think one is speaking to the platform kind of efficiency that Ashish mentioned. We were very purposeful, okay, to basically take organic Marvell plus all these assets we acquired, everything on 5-nanometer technology, right, as a platform. So our SerDes, our PCIe IO, our ARM CPU subsystem, our interconnect, I mean, everything is very efficiently done, okay, on a single platform. And that breadth of IP from switching to storage to Ethernet to CPUs; we even have our own AI/ML block that came from Cavium that we actually turned into an IT, not a chip. I mean we've got probably the richest array of IP that's not off the shelf. It's ours. And the power of the model is we actually use it for our standard products as well. So one is there's sort of this unparalleled IP set on a common process technology platform. The second is, I think, how we've organized the company and the go-to-market. It's a One Marvell approach. So feedback we get a lot candidly from some of the big customers is, this is really refreshing because a lot of -- one big account told me just 2 weeks ago, he's #2 guy is this large company, basically, we have to talk to 7 different VPs at these other companies. There's like different business units, and they report to these different GMs. And basically, we go in and Raghib Hussain, who's my President and runs our products and technologies. We have a single point of contact. We resource all the -- marshal all the resources of the company. It's like one clean interface to the customer. And then we bring the best of Marvell in. Oh, you want our security architects. Oh, you want our storage guys to come in. So we're thought leaders in those areas, Tim. And then I think the other one is the technology, it's the team and then it's also the business model. It's actually not trivial to -- well, one, be able to execute a range of business models from full custom kind of vanilla ASIC to, look, we'll build the chip, but then you give us a spec and then we'll go execute it with our team, right, to, oh, you want a metal mask option derivative. Oh, you want to swap out the DSPs. And we just have this company that's able to pivot and do that and just our larger -- the larger companies are just not quite set up to do it. And I think the other thing is we're very focused. We're the only pure-play data infrastructure company. So like my time, Raghib's time, management's time, it's not spent on, oh, we're also in the TV business. Oh, we're in this -- we have a system business. Oh, we have a PC -- but there's 40 accounts, and we just go super deep. And so I think it's a combination of all of those that makes it attractive. And then it's a trust -- it's a very trust-oriented company as well. So at some point, they have to -- they're putting their crown jewels in your hands. And so again, I think our -- kind of our customer model, our customer engagement, our whole philosophy is suited pretty well. Ashish, did you want to add anything or...

Yes. I think, Tim, back to uniqueness, right? I think if you think about a world where data is the most important piece of currency information, pick your poison, what do you need, if you think from a technology perspective? You need compute, you need storage, you need networking, high-speed connectivity, security; plus, you've got this flexible business model. And you have to be really good at all of them. I think this is one of the unique things we've done both organically -- organically storage and networking. We've established that capability internally. We picked up Cavium, which gave us compute and gave a security, right? And then we picked up Inphi. These are all leading technology assets in this space by far, right? And now you have the ability, right, especially if it is Venn diagram of saying these are my capabilities. I don't think you'll find any other company at the intersection except Marvell. And on top of that, as Matt mentioned, you throw on, "Oh, by the way, I'm going to give you access to this in a flexible manner on the latest process generation." I think it's a completely unique capability. I don't think it's replicable anytime soon by anyone. We've already picked up all the key assets. We've got this unique business model. It's successful.

Yes. I mean just to that point, and then we'll just move on. But I think if you look, Tim, and I know at different times, maybe we took a little heat for some of the valuations, maybe we paid for things. But we had a -- I had a -- I very much had a philosophy from day 1. We're buying the suite assets. We're buying the best assets. We're not buying the junk assets, and then we're going to try to fix the junk assets so that we can say we have a piece. So to Ashish's point, Cavium was the absolute best ARM-based microprocessor company in the world. Inphi was the best electro-optics company. I would argue the Avera team, capability-wise, the business wasn't all that great because they had kind of been hooked into legacy older nodes and they were -- but the actual IBM team that we took on, the heritage there, the 25 years, outstanding team and asset capability. So that Venn diagram actually looks pretty good. Then the Marvell storage classic, that was the #1 storage franchise bar none in the industry. And even our enterprise and kind of PHY teams with the Aquantia add, by the way, on the PHYs, they were the best stand-alone PHY asset available. They pioneered multi-gig, right? So one reason we gained a lot of share is they -- so I think that's -- yes, it's a really important point. It's not just that we have the pieces. Each individual piece of technology and team is really impressive and at the leading edge of performance. And so it adds up to a really competitive platform.

So Matt, I guess, hearing all of that, the inevitable question is that, well, how do you decide? There's so many opportunities coming at you on the custom side. I mean at some point, you have to be very strategic about where you allocate resources and -- so that you're not allocating resources to a project that's never going to get off the ground. And that's maybe your other custom ASIC peers, they won't even work with you unless you're going to be a $1 billion product. So you're sort of -- you're not up in those -- in that scale, but you've got so many opportunities. How do you think about where you allocate resources and where you don't? Is there a conversation with the customer where you hear the tone that they're using and say, "You know what, this guy is not really serious about what they're actually talking about." How do you think about that? Because that's where you could get OpEx creeping, not get any revenue for it.

Right. Well, one thing I think that lays over all this question you had that helps -- but I mean what's important is the capital allocation decision making, which I'll get to. But pretty much all of our projects now come with very substantial NRE, Tim, which is for the investor community's nonrecurring engineering. And some companies treat it different ways. We treat it as an offset, accounting-wise, to R&D operating expense. And so that's a co-funding model that's in place. So a lot of the risk, Tim, gets taken out when you actually have the NRE because when you're -- like Thunder had no NRE. Thunder was on our dime. We were burning a ton of money on it. We were sampling Thunder chips. We were talking about Thunder. Nobody was paying for Thunder. You know what I mean? And at some point, you're like, what the -- and then nobody wanted Thunder. Nobody wants a stand-alone ARM 64-bit CPU for server. They don't want -- they want their own part. So that was one example where we pivoted. But -- so that helps. But the way we do it, I think, is kind of in 2 ways, right? Certainly -- and one thing I think we've done a good job of from day 1, certainly after I arrived, was really instituting a very rigorous and thorough review, I would call it, all of the projects that we at the time we're working on or when we buy an asset, we actually have a really good kind of way to evaluate the -- both the financial and the technical and kind of commercial all-in-one aspects of the investment. And we've been pretty quick to kill things. Like we bought Cavium, we killed XPliant day 1, right? We just said not going anywhere, no traction, team wasn't the same as -- whatever the reasons, get out. So we're pretty disciplined about it. But where we are now, because I think, to your point, we have -- even the customers have realized this, that the capability that we bring to the table is actually scarce. It's not like you can go bid these chips out to 4 guys and see who quotes the lowest price per silicon micro acre. It's like, "Okay. Can I get the right team from Marvell? Can I get them to do my schedule? Can I get the commitment from the top?" And we're very focused from a customer standpoint on the top leading companies. So think of it as we're all in the top 4 hyperscalers. And what I mean by all in is we actually -- the aggregate kind of returns from those investments are very good. But candidly, some of the projects, they might be lower in ROI. And some of them may be outsized. You know what I mean? So there's a customer lens, which is -- think of it as the leading companies in the 5G and kind of cloud area, where we want to be the go-to partner, okay? Automotive, we're -- there's a lot more of them. And so we're really trying to figure out who are the thought leaders there, both start-up kind of companies as well as the established ones where we can really do not only the Ethernet kind of stuff but beyond. And then I think for other more established kind of markets and customers, we tend to just really make sure the economics are -- it's got to work. So the hurdle rate just kind of keeps moving up every year, candidly. The returns we're getting are better. Our relative market share in every product line we have has gone up dramatically. And so -- yes. So -- and we haven't run out of engineering capacity just yet. I mean, we've been kind of hiring to keep up, but we're -- we kind of almost keep hitting this limit. But it's enough to where we have to turn some things down, and that's okay. I mean I learned this from Jack Gifford a long time ago. You better have a -- if you have 10 ships you want to do, you have better have another 15 that you're not going to do. And it's a really healthy thing to force yourself. So it's pretty good.

And Tim, what we've seen is -- I mean there is enough visibility customers give us. I mean that's a nice order part, a new semi custom. You get a pretty good idea of where this thing is going in. And a lot of these are going into, what I would call, very key revenue-generating properties right within some of these customers. A lot of them are existing properties where they're shifting from something they've been doing for a number of years, and now they want to go to semi custom. There's a track record of usage and capability. So it's not as much of a crapshoot as it looks like. Hey, it's not like customer shows up a little bit on it. You don't know where this thing is going. You actually have a very good idea of here's how many servers they've historically deployed. Here's what the growth rate looks like. Here's the attach rate. That stuff's pretty well understood, like we've had great experience in the base station market and the switching market. We have a very good understanding. So I think that prioritization already happens that these are like -- these are very important to the customer. If it's important to the customer, it probably makes sense that it's going to go to production. It's going to drive all the revenue. I think that's the other lens we put on it is...

I mean I'll give you one other kind of interesting example. This one just came up. I don't even know if you know about this Ashish, but there was an example just recently where -- again, because Tim, as you know, we have this full custom ASIC capability, right, the kind of the true ASIC. And so we just had an example about a week ago where we actually converted it. So what happened is there's a company that's been doing their own ASIC, their own team for a long, long time. And through one of the acquisitions we have done, we have more of a merchant product, if you will, but that we're willing to kind of do spins on if somebody wants it. And we were kind of looking at the customer saying, "We're not so sure this internal team is going to be able to execute. Like do we really want this ASIC or not? Because are they going to be able to actually do it?" And then we kind of went in with the merchant the part we already had and said, "Look, we'll tweak that for you." So to be honest, we dodged a huge bullet. We won. We didn't take on a project that might have spent a few years and not being -- been successful. And literally turn into like a metal mask of a chip we already have. That's, I mean, just an example of the power of the platform, right? And while working together as one team, -- because in theory, our ASIC group lost that design, like they didn't win that bid. But our team, sales, [ BU ], kind of got everybody converted over and then that one engineering team -- do you follow the example I'm giving you?

Yes.

Super efficient.

Yes. Yes.

Because -- fine, you would have done the development and offset the R&D expense, but you're waiting years versus we know this other chip works. The yields are great. It's already got other customers on it. The interfaces are proven. And that helped get the customer converted over. It's a win for them, too. Why are they going to go burn 2 or 3 years making this ASIC? We can get to them this year. So that's the kind of power of the model, just to try to use real examples here because sometimes it's like theoretical, and then people scratch their heads. But this is like a real one we just won.

Appreciate that. Awesome. I guess I wanted to ask about Innovium because that's sort of the next new piece of the IP in your bag of tricks. And I think the one thing on Innovium that you hear when you talk to prospective customers is that they say that they've tried to use them in the past, but they just couldn't execute to the road map. And I certainly understand that they have a big customer that you're ramping with. But beyond that customer, the feedback was often that the customer wanted to use them, but they just couldn't execute to the road map. So is this a similar -- I mean, is it the exact same playbook with Cavium? Inphi might not fit that, but Cavium will look now they're part of Marvell and boom?

Totally. I think it's the same analogy. Too small, start-up, don't have enough resources, what happens to them? So it takes us time. This is -- but I think it's a very, very similar type of situation where -- I mean I just was in a review this morning, reviewing kind of the whole combined integrated switch team, which used to be 200 guys in Innovium now, right? But previously -- and that was like, I don't know, 700, 800 people combined with the Marvell team, right? And again, we're already proven on the 5-nanometer platform. Our SerDes is in place. There's a whole kind of how do you get them on to our platform as fast as possible. And this -- there's -- we did a press release recently, but one of the big things you'd hear, Tim, when you talk to these customers is about support for things like SONiC and SAI, which are -- and SAI, which are the 2 open source -- the initiatives around open source network operating systems. We basically quadrupled the number of people working on that when we acquired it, meaning if you looked at -- I don't know, maybe we had 10 and they had 30 and now we have 40 or something. I was like, "My God, I have 40." Maybe it's the other way around. Anyway, my -- the point is you end up -- as you combine these, we've got significant scale now, which we didn't have before. And same happened with Cavium, by the way, right? They just couldn't take on some of these projects. And the big customers literally said, we can't -- "You're just too small." So yes, I think it's the same thing. It's going to take us some time. We have to get in there. But we're really happy with the team. We're happy with the leadership. The collaboration is going great. They're already heads down trying to pull in the road map.

And then, Matt, I just wanted to -- you touched on auto a bit. But the next sort of compute opportunity is a auto compute opportunity. And you did mention -- in the call, you mentioned Ford. You mentioned EW. And I think you mentioned it not because to say to your customers, but more to say, "Look, they've made public announcements that they want to do custom silicon." And it doesn't like these cars have an engine that they can differentiate on the engine. They have to differentiate on being compute. And so can you just talk about that because that seems to be a really -- I mean albeit longer term, but a potentially really, really big opportunity.

Yes. No, I think it's a huge opportunity. And yes, just to be super clear and -- so nobody read that wrong, and I was trying to be as clear as I could in the prepared remarks in the call. Yes, I was just citing what I had heard them say or we had heard them say or read publicly, right, both Jim Farley as well as the CEO of EW. And I think even the CFO of one of them was being interviewed, and they're talking about this, right, in the context of not just -- what's interesting is some are talking about it in the context of the chip shortage, okay, we need to control our destiny. But the companies we're talking to, they see this as, to your point, the key differentiator. Because when they go to ADAS, okay, and you start looking at these off-the-shelf offerings, right, from the larger companies that might provide the maps and they might provide the modules. And by the way, you got to give us your data, if you use us and you got to pay a subscription fee. And so some of the -- I'd say all the car OEMs are wondering, well, if I see this key piece of technology long term to somebody else, then how am I going to differentiate because all the people I used to have, to your point, working on engines and carburetors and transmission, they're all -- they don't need those people anymore. It's going to be about apps. It's going to be about connectivity. It's going to be about autonomous and electricity and the whole 9 yards. So the chips are going to be incredibly important and the compute kind of -- the networking architecture and the compute sort of decisions are the 2 most probably important decisions these companies are going to make. So I do think it opens up a large opportunity. It's not a winner take all, okay? It's a little bit like the hyperscale deals. They're going to buy x86s for a lot of things. They're going to do custom for certain things. They're going to -- it's like automotive. We're not making any claim, but there's a significant amount of activity. And our model is the unique one because it's really for the companies that truly want to control their own destiny in the future and work directly with semiconductor companies and develop their own and have the scale or the funding, if it's one of the -- because look at the capital sort of funding that's gone in some of these start-up companies, it's huge, right? So they can put some serious wood behind differentiating the internals, and they really care about it. So it is a longer-term opportunity. There's no question. But these things -- look, we're getting older, Tim. Time passes faster than you think, right? 5 years ago, I became CEO of Marvell. What happened in 5 years? I mean I think about 5 years from now, we'll be shipping automotive compute. And hopefully, between now and then, it's not just one account, one big win with a bunch of pipeline, it's like, hey, this is a real business. And I think that's where it can go. And that's really one of the things -- and the reason we talk about this now is in our business, as Ashish said, it was like this data infrastructure business, you got to think in 5-year kind of increments. You can't just be saying, "Well, I'm going to get it -- oh, what's the next iPhone cycle, I got to get it on. I got to do my annual cadence. This is like how do you get in these platforms that last 5, 7, 10 years? And then how do you -- like how do we get in model year '25 right now, right? And how do we sprint to go do that? Those are conversations we're having. So yes, but it's very cool, and it's real. And we're probably -- if you just think about it, just last point, I mean, we're kind of the unique combo of having all the platform assets to do it, the compute, the networking, the IOs, the interconnect and having the auto-grade capability built. I mean we're shipping at 0.7 DPPM right now on our Ethernet parts. Our quality is like gold standard in the semiconductor industry with tens and tens of millions of units shipped. So we have a lot of credibility there. And then it's like how do you find someone has all that and is willing to do the semi-custom thing and it's just -- it's pretty unique. So I think we're kind of creating a market is the way I'm thinking about it.

And I guess maybe just one last thing from me, Matt. Are there pieces that you look at these markets and you say, "Well, you've done this now, it's been kind of [ rich ] and repeat now. This is what the fourth acquisition or -- that you're going to do the same exact thing with Innovium that you've done with Cavium. And Avera was a little bit different, but it's the same thing. And you now have the platform that you could put -- in theory, you could put more assets through this platform. Now I'm not saying we should go out and buy somebody else. I mean -- but I mean, are there verticals where you're saying, "Well, that's a big growth opportunity that we aren't really in." I mean is there something touching the consumer that you're not really in that -- or pretty much as you scan the landscape and you say, any -- I'm sort of happy with the IP portfolio I have, and we've got way more problems to solve than even have people. So the idea that there will be more M&A is out years away?

Yes. No, we're very happy with the portfolio we have. I mean this has been a labor of love to put this sucker together, took a lot of risk in the process, didn't quite -- obviously, we had a vision and a strategy, and it's -- we're very happy with where we ended up. Look, I think the level of consolidation in the industry has been pretty unprecedented. I mean just look to 2020, right? 4 of the best assets we've ever seen all got acquired or got announced, right? I mean Xilinx, historic property; Maxim; Inphi, ARM, although maybe that's coming back out again. Who knows? But my point is it really showed -- I think 2020 to me was kind of the seminal moment where even these large, established, really high quality, they all went. So I just don't think there's that much, and we don't need much. And you mentioned the consumer, we're not going to go into the consumer market. I think I brought a philosophy to Marvell, which was quite different than what had been going on, which was I just call it super simple, but we call it the power of focus. Do fewer things and do them really well, and you can get multiples bigger than just if you try to kind of spread tons of bats and just hope something sticks, right? So I think this -- we really like this focus on data infrastructure. I think we've hit the sweet spot. I think it's where the whole world is moving. I think it's the biggest sort of semi TAM that's going to get really big over time in not just 3 years, but 5 years, 10 years. And so we really do have the fundamental pieces. If we can find teams, acquihires or small technology companies or Innovium-like little properties, right, where you can park them in and the team can have its autonomy and like, sure, we'd be open to that. But the large-scale things just really don't exist. And so I think -- so it's a combination of those things. I will say if we found something, though, we're good at it. I think we've -- we're really happy we did the ones we did. We -- by leveraging ourselves to go with investment-grade debt, by the way. But we've kind of been in a leverage mode, buying things and integrating them. That's worked out really well for the company, for shareholders, for all the customers. But I'm being realistic, not as many properties. And we got all the pieces we need. So I'd like to go just drive tons of organic growth. That's -- organic growth is a lot of fun. It's easier. It's a lot easier.

Well, we are out of time, Matt. I can talk to you for much, much longer than this. So anyway, thank you so much for your time. Thank you, Ashish. Really do appreciate it.

No. Thanks, Tim. Thank you very much.

Yes. Thanks, Tim. Appreciate it. Good luck.

Thanks so much.