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

Welcome back. I'm Blayne Curtis. Next up, we have Marvell from the company, Jean Hu, CFO; as well as Ashish Saran, Investor Relations. Given, Jean, that you just reported, I thought maybe a good place to start is recap a high level, and then I'll definitely drill into some of the pieces, but maybe a couple of minutes on just what you saw when you reported.

Yes. First of all, thank you for having us. I'll do a quick recap of our fiscal Q3 earnings result, which we literally announced the last Thursday. So we actually delivered record revenue of $1.54 billion, growing 27% year-over-year. Also, we delivered 37% operating margin, really profitable business. The earnings per share growing 33%, much faster than the top line revenue growth. Overall, our team executed really well despite a really challenging supply chain environment. And also during the quarter, right, we saw end market dynamics. In our storage area, some of our customers really changed their behavior very quickly. In the early part of the quarter, they were still trying to get more supply and ask for more components. But when the quarter progress, toward the end of the quarter, literally, it's like oh, we really want to push out. We probably have inventory, and we really want to reschedule. So our philosophy has always been to try to manage in an orderly fashion to make sure, yes, if there's inventory, let's go through the digestion, take the medicine. So when you look at our Q4 guidance, we actually guided the revenue to grow 4% year-over-year, but it's a 9% sequential decline. And also, it's below we are generally forecast because of the inventory correction. And the large correction is in data center. And the majority of them is actually all in storage in HDD area. We also got some headwinds in China. Enterprise networking, we do see the demand slowing down in China. It's also reflected in our Q4 guidance. But if you step back, look at our other end markets, carrier 5G continue to be really strong. Automotive, we continue to see the strength of adoption of our technology. And for the fiscal year, fiscal '23, we actually expect revenue to grow close to more than 30%, operating margin of 36%. And all our secular growth drivers, they're all impact, right? 5G, automotive, cloud optimized solutions, 800 Gb and 400ZR adoption. All those things are really strong. So overall, we are very well positioned and also we are operating at a scale as a company. So we feel pretty good about the long-term prospect of the company.

I mean that is the question I got most after the print is everybody needs to assess your Analyst Day, you talked about 15% to 20% top line growth. But when people have Analyst Days during this pandemic, you have to reassess everything. And I think the question is really, is that still the right growth rate? Obviously, data center needs to have more growth than that for you, the whole company to grow, and that's obviously where you're seeing some of the correction. So kind of assessing what's going on and kind of what's your answer to are you still -- is that still the right category for the company?

Yes. I think if you stand back and think about when they came up with a 15% to 20% kind of -- you have to really dissect it by the big end markets. And most of that growth, we were assuming is going to come out of automotive, which is clearly a new product entry for us, which is rapidly growing well above that number, quite frankly. Our assessment on cloud CapEx was, yes, I think it needs to be somewhere in that 50 to 200, which is below what we've seen in the 30s and 40s, we just assumed a more conservative number. And I think that in the long run still seems to make sense to us, right? And the other component, of course, is 5G. On top of that, we were expecting and now still expecting significant share gains and new product cycles. And those are actually holding up extremely well. So I think we stand back and look at it. We grew a lot faster, right, over this calendar year than our long-term target. If you average over the next couple of years, even with what you would perceive a correction here, we're still going to be somewhere in that range, quite frankly, right? And as we come out of it, especially, I think it's going to be a tale of 2 halves. First half is where you're going to see some of this inventory correction work out. I think you start getting in the second half is what I suspect you'll start to see the acceleration. So overall, I think we feel very good about where this business can go from here.

What happens overall, in terms of inventory really at the customer? I mean, I think a lot of management teams have said, look, they're still asking for expedites and calling me every other day. It must be lean, right? But you talked like particularly in storage, it happened quick, and they ended up -- they're harassing you and all of a sudden, they have too much. So how do you reassess the level of inventory at your customers now that you've seen what's happened in a couple end markets?

Yes, that's a great question. Frankly, all the semiconductor companies are super focused on what you just asked. I think for us, the great thing about Marvell is if you look at most of our product lines, we have a direct relationship with the customer. Either the customer design or we are the only supplier to the customer, that helps us to identify the end market applications, the demand and how we know if they have inventory or not. It's better if you look at 5G, right? We work directly with the OEM customers. So we have understanding about 5G adoption. That helps a lot. They may still have inventory, but at least we have more visibility. The areas like storage, the challenge is really our customers are storage OEMs, like WDC, data, Toshiba. And then they sell to either cloud or they sell to channels to the cloud. So there are multiple layers between us and end market demand. The visibility is really limited. The other thing I would say is if you look at our different end markets, like storage, it was not that supply-constrained. So they -- largely when they demand a lot, they keep asking us, we gave them. And then when they realized they have inventory when they tell us, of course, it's quite late because we're at the end of the food chain. But most of our customers actually, we have a better understanding and we have some visibility, especially with new product ramps.

Yes, I think, Blayne, I mean, to Jean's point, where storage was really not that supply, these are relatively smaller, simpler parts, right? When the demand went up, we were able to supply. But we've been dealing with supply constraints on most of our other projects, essentially, right? So in some ways, I think the inability of the entire industry to catch up to what customers are really asking for may have turned out to be a silver lining, right? Where you really never got to that full demand they were looking for versus there are certain area storage in particular Yes, we did actually catch up and we went behind. So I think that's the other reason why I think you're not seeing the same level of over-inventory, I would say.

Moving to the data center. From a high level, you talked about what cloud CapEx may be. What's your view on cloud CapEx next year? I think Matt talked about it coming down like what's your best guess to where it is.

Yes, I think, look, the reality is, I think we're all waiting to see what actually happens, right? I think our planning assumption is, yes, there's going to be another deceleration. I think we've seen last year, the numbers are somewhere in the 30% to 40% growth. Clearly, this year, the number looks to be more like somewhere, on average, somewhere in the mid-teens. Are they in forecast, hey, maybe it's mid-single digit, maybe you want to plan for flat, right? I mean clearly slowing down, whether it's flat, it's up 5%, I think it's too early to know. But you do have to plan. I think the planning assumption certainly -- I mean you've seen the public activity from a number of cloud customers where they are having to reassess and make some changes. So I think you have to be a little practical going into it. But then I have to also think from a long-term perspective, fundamentally investing in bandwidth growth is critical to their long-term business success, right? So I think, yes, you've got to deal with some macro conditions in the short run. But when we are looking at the design win activity, the design win funnel, right, the requests for us continue to invest in their next products, it remains extremely high. So I think in the long run, I don't think anything changes. But you do have to work through, call it, a few quarters.

There's obviously a couple of pieces within data centers, so maybe let's get the storage piece out of the way. Like how big is storage of data center? And kind of how do we think about what the prior run rate is, where is it going to be in January and could it get back to them?

Yes. Storage, if we go back before we acquired Inphi and start reporting by end market, we actually said storage is around $1.4 billion annualized run rate. That was like before COVID and everything. So if you look at this year, the early part of this year, the first half, definitely, the run rate was much higher. But now Q3 and Q4, especially the run rate is much lower. Our $1.4 billion storage annualized revenue, we also said about more than 60%, it's largely in data center. So it's very much data center when you look at Q3 to Q4 sequential decline, a larger portion of that is from HDD, from storage side. There are some SSD and the fiber channel, too. I think it will take a few quarters for the inventory digestion. But overall, we do think eventually Q4 next year, it should more go back to this $1.4 billion run rate. And within our strategy, it's very different today. It's -- the majority of them are really nearline. We know the demand in cloud everywhere for the nearline drive will continue to be there. And also our flash controllers, the demand for flash acceleration continue to be there. So we do feel pretty good about the long-term prospect of storage.

So then excluding storage within data center, I mean, when you think about the PC sales to the cloud, you have optical, so some switching. What trends are you seeing in that business?

Yes, I think a couple of things. So switching, as you know, is a relatively newer business for us, came in through from the acquisition of Innovium. That in Q4, that is probably still growing, by the way, right? This is one area which was actually -- remains fairly supply constrained. We have much bigger orders, right, as the customer transitions towards us. And as I look into next year, kind of somewhat irrespective of kind of the CapEx discussion we had, we expect pretty strong growth out of kind of that switch business. So switch trends look very, very good. I think when you look at optical, I think there is a couple of -- so there's a run rate business. You can think about 200, 400 Gb PAM4 which you've been shipping in volume to multiple customers. I think that probably follows more kind of the CapEx trajectory. But we also have newer products and 800 Gb PAM4, which are being used not for switch applications, but really for AI applications. And that remains a net investment area for multiple cloud customers. So that for us remains a big growth driver. And there was a new product we introduced all 400ZR, which does connection between data centers. That's a new product, and we expect that to keep ramping. So overall, I think when you look at our electro-optics portfolio, we should still expect growth next year despite some macro choppiness because of some of our new product cycles.

I ask about the cloud optimized silicon because you guys have talked about $400 million and then another $400 million on top of that over the next 2 years here. I think the question I get is it's still on track. Obviously, if we do see -- we don't know exactly what the environment is going to be, but it's probably slower than you thought about 6 months ago. Does that change any of the cadence? And when you know what products you've developed. But you can -- how well can you tie that to the actual ramp schedules in volumes?

Sure. Yes. I mean, this is something we look at very closely. And our customers look at very closely because these products are essentially going into mission-critical revenue-generating capabilities inside those companies, and these are replacing our solutions a lot more efficient than what they use today. So the pressure from customers to get these products out on time has actually been very high. We've already started ramping some of the products, which will, in aggregate, drive $400 million of revenue next year. In the last -- when we started talking about the sizing of this, we had sizes a little bit above that. We said it's going to be more like $400 million plus. And obviously, we have some error bars we keep inside some of that probably has come down a little bit. But in aggregate, we're very comfortable with the $400 million for next year. Like I said, customers are really pushing us to get these things out on time. And then some of them continue into the following year, plus we have some new products coming online. So overall, yes, this is going extremely well.

In terms of the visibility of these products, is this something that you eventually will be able to press release or talk about in more detail? Or is this kind of customer-specific stuff that

It's very customer specific.

Yes. It's very customer specific. I think one thing for sure to add to what Ashish said is those are some very complex solutions. So manufacturing cycle tends to be long. So we do have some visibility from register security, substrate securing, supply and then make sure it has the ramp. So that from that perspective, definitely, we see what Ashish mentioned, right? He said $400 million next year because a large portion of it, you have to have a capacity to provide the supplies.

Yes. I think, Blayne, in terms of, hey, when would you hear about it. I think what you'll hear with for our customers, but -- these are important enough things for our customers, right? I think as you hear their rhetoric and their tone in terms of how they think about semiconductor sourcing and strategy, I think that's when you're going to start to hear more about it, right? This is very important to how they look at their data center architectures going forward. But yes, to your point, these are very customer specific. So we can't really say much more about it, but I suspect you will start to hear more of this as time goes on.

And I wanted to circle back because I wanted to ask you this. On the optical side, just the competitive landscape. So when you look at like PAM4, I mean every time we go to OC, you think you're going to see something. And I think Broadcom showed some products, but it seems like you're still a pretty high share of that. And then -- so if you could just comment on what you're seeing in PAM4. And then ZR as well. I mean, I think it's early days, but in terms of that landscape where there's a half dozen companies kind of vying for it? What are you seeing?

Yes, I think the reality is something we realized before we made the access of Inphi because we obviously had internal capabilities with mixed signal physical layer technology. I think what we found out is it's a very difficult product to do. Even with all the investment, even if you have the underlying expertise. So at this point, we remain in a very strong leading position, and that's my expectation going forward. We, in fact, just yesterday or today morning, in fact, just announced our second generation of 800 Gb PAM for second-generation, while most of our competition is still kind of launching the first generation. This is 5 nanometers, right. It's being deployed in volume and AI clusters. So all I can say in PAM is we would certainly expect to keep a very strong leading position here, right? And as you can imagine, if you couldn't quite do it at 200 and 400, it only gets more difficult as you go to 800 Gb, right? And of course, as you know, there's a next generation of 1.6T flag coming along. So bottom line PAM, I think, will remain in a very commanding position. On the ZR side, it's fairly similar, right? This is a fairly complex module using silicon photonics and all our co-head and DSP technology. There is one other key player in this market, clearly, right? Cisco through that acquisition. So I think both of us will have a pretty good position in this market, right? I think beyond us, I think it's going to be a bit of a challenge, right? You need access to all the optic capability, the core packaging capability. I think it's a pretty optical matter, right? I mean as you know, Inphi invented this technology, working with a key partner in the cloud space. So I would say -- and like I mentioned, 400ZR for us is going to be a growth driver for next year. So I think we feel very good about our position here.

Can you tell you when you acquired Innovium, you talked about $150 million of revenue. Can you talk about did you achieve that? And how do we think about that stream? And then I guess, I mean, there's a road map at some point where you would kind of have a 5-nanometer chip that just kind of just talk about the timing of kind of how do we think about what you have today in terms of run rate? And does that grow from here? Does it take a pause, until you get the new one? Like how should we think about cloud data switching?

Sure. That's a great question. I mean the first is we've been extremely happy with the Innovium acquisition. The team is integrated, right? The business is doing extremely well. This was one of and still is one of our bigger supply constraints. So we are on track to deliver, call it, close to $150 million. So we feel pretty good about that. We're not quite done through the year, but that's kind of the plan. And then one of -- maybe the only positive outcomes of the industry slowing down is capacity becomes more available. So certainly, as you look at next year, you should expect strong growth out of that business, right? Maybe you already have the orders, we have the customer, we just need more substrate capacity. So I think that future looks very good. This is all on the 12.8T generation. To your point, there is an upgrade path coming right? I think the key point there is it's -- while the switch availability is one part of the equation, probably a more important part of the equation is you need the high-speed optics, which is 1.6 T. And that's the road map we are driving. As you know, it's all again PAM-based solutions. And from a -- even from a spend perspective, I think, as you know, the average spending on optics is about 10x higher than it's on the switch. So it's really the optics cadence and availability, which drives this next-generation upgrade. So there's more news to come in the, call it, in the near future. So we'll kind of leave it at. But you should expect to see significant activity on that front on both switch as well as some 1.6T optics.

Maybe moving to enterprise. So that was the other area that you saw some weakness you called out China. Can you talk about what types of products or this? What did you see the change and think about that business next year? .

Yes. We basically said looking into our Q4, which we gave guidance, the China revenue in enterprise networking is going to be really much lower, significantly drop. And as an overall company, the China revenue actually is already below 10% in Q4. When we look at it, it's -- the major thing with the China side is the early part of the year, there's a supply constraint. They continue to struggle for supplies. But as we end up getting more supplies, of course, the demand is coming down significantly because of COVID. So once we get into Q3, we do see continued reset from Chinese customers, they want to push out orders. They may still need it in the future. But for the short term, they really try to adjust their inventory position, try to manage their demand situation locally. So for us, our China revenue right now has derisked significantly. If you look at the enterprise market, it's probably low teens percentage of enterprise market revenue is China. So it's quite small in Q4. And going forward, our assumption is we don't know if the demand is going to come back or not. Our assumption is, hey, at this level. Let's just assume that's the level we're going to go -- move forward. So that's the China revenue part.

Yes. I think overall, we look at Enterprise Networking. I guess the other comment I'd make is, I mean, we've done very well in this market, right, with our new products, right? So clearly, we've seen that revenue improvement. Our content has gone up pretty dramatically. Multi-gig fiber is one example, particularly on the switch side. So I think -- in fact, I would say the bulk of our year-on-year growth, which has been very impressive, you'd say it's averaging probably in the 20% to 30% year-on-year if you go back a number of quarters, has done higher in some quarters. I think a lot of that really is that content game we were able to drive with our new products. So as we look at next year, which is what a lot of investors are looking for, I think we still have some content gain in front of us, right? And we also have share gains. We have a number of -- we've won some new custom ASICs in this space, some new programs which have just started ramping. So while it's kind of difficult to know what the end market is going to do and you could even say, hey, let's assume the macro correction. The end markets does step down even outside of China. I think we're pretty well placed to offset a pretty significant portion, if not all of it, with what we see in terms of our new programs and content gains. So we're not quite sure exactly where it pans out. But overall, I think we're very well positioned to even better through if there is a, call it, a reduction in demand even in the U.S. market.

And I think the other part of it is you had, I mean, it was up at one point, I think 70% year-over-year. Dipped down a bit. But I'm assuming the U.S. part of it is still going quite strong. You look downstream, you see a decent amount of inventory at those customers' balance sheets. And I think they're still short certain semis and waiting for them to shift to their backlog, right? But then within it is what you just mentioned you have content, you have share gains. I mean you have better visibility on this. I mean is there a way to kind of parse out when you look at that 50% year-over-year growth or higher?

Yes. I think at the end of the day, right, this is -- as an end market level, I mean, if I just go snap all the way back to a couple of years back to holistically looking at the end market, our assumption typically is, look, this is a GDP business, right, enterprise networking on, in general, right? And then we expect to outgrow that. But our means like 2x, 3x, right? So I think we did have a phase of growing much faster, which makes sense because we have won the designs about 2 to 3 years back, but we didn't see the benefits because our customers have not started shipping the new platforms yet. That only really started next year. So we're in this phase of, call it, outsized growth because now those design wins are showing up completely. And it was also the first year of meaningful multi-gig, right? If I had to fast-forward letting go of macro for a second, I'd go back and saying, yes, our assumption is assume a low single-digit growth market and we'll be, call it, 1 to 2x higher than that. I think that's still a good reasonable assumption.

I do cover the other 2 end markets and maybe the end markets are in a little better shape. I mean auto seems like the one that's holding up. You're obviously a content gainer. So I think you crossed $200 million run rate in autos. You talked about it being a $1 billion product. Can you just walk us through that trajectory to get to $1 billion and kind of what you're seeing in terms of adoption of your product and I'm assuming regardless of what autos, whether it is a correction or not, you still should have a growth trajectory in the semi regardless?

Yes. I mean, this has been -- actually has been an amazing business. I mean this was basically 0 revenue essentially a few years back, right? And we hit $100 million milestone about a year back. We just crossed $200 million. I think what's really -- and this is all ethernet content essentially going to cars. And what's fascinating is maybe a year back, it was a bit of a discussion within the car OEMs. If -- which parts of our network do we switch, which platforms, I think now it's moved to a when. There's no longer a discussion. It is going to be the backbone of these cars. And equally exciting, it's not just the EVs and hybrids, which drove the initial set of adoption. What's actually much more meaningful is there's going to be ICE cars that we like it or not for a very long time, okay? And those are now switching over to using Ethernet, right? So some of our newer design wins are really much bigger in scale. They're platform-wide. So that gives us very high confidence that, hey, we sized this at about $1 billion Ethernet market, we'll get to at least half of that. And I think that would be in line of sight, meaning a few years, literally went 0, 100, 200, so you can kind of project that type of cadence for a number of years. But it doesn't stop there, quite frankly. I mean, I think the $1 billion is still maybe 1/4 of the total connectivity in cars and that will inevitably become bigger, right? So I think we're on a very good line of sight in the next few years to get to our initial target, call it, $0.5 billion of revenue. But we'll keep going from that point onwards.

The design win momentum in that business is much more than we thought. It's platform wins right now versus in the past days also, 1 model, 2 model. Right now, if there's a design win, typically it's across all different models. So that really supports a very longer-term ramp over this business.

I mean we have 8 out of the liberal top 10 OEMs by volume already won, essentially. So you can think of pretty much any name you can think of in Korea, Japan, Germany, U.S., you name it, it's part of that list.

I want to ask you, though, I mean, I think the Aquantia story was to get Multi-Gig and maybe 10-Gb in the car. But you're selling a lot of 1-Gb I'm assuming.

Yes. Today, it's gigabit primarily. But you make a great point, which is as these cars move towards these more zonal architectures, which simply means stead of having a lot of small ECUs scattered around the car, you're going to have much bigger, but more powerful cohorts, but that needs a lot more bandwidth, right? So you're absolutely right. The customers are now putting gigabit in cars today, but what they're really working with us on is multi-gigabit. And we even got a request for 10 Gb in cars.

10 Gb and 25 Gb.

Yes. We house out of 25 Gb, which kind of was amazing. But yes, it's very much going in the direction.

And what do you see in terms of opportunities beyond Ethernet? I mean I think you mentioned one ASIC, I think it was the outset, I forgot. I think there is a lot more advanced silicon going into cars and a few companies could do it. What's the longer-term opportunity beyond Ethernet?

That's a great question. I think the visibility we are getting because we are becoming the network, it gives us full visibility to the rest of the silicon architecture. They're already auto qualified. I think the opportunity is as you think long term, if you're a car OEM and you want to control your own destiny essentially, you want to own your own processor road map, which is your differentiation, but you don't want to necessarily build the whole thing yourself. You need a partner. I think that's the option. It's very similar to the cloud optimized silicon we are now doing, where we are helping our cloud partners develop optimized solutions. It's pretty much the same playbook, but in automotive. Now obviously, it will take a lot longer, right? For the next, call it, 4 to 5 years, Ethernet gives us a huge amount of revenue growth. But I think somewhere in that time frame is where I think you can expect us to look at additional opportunities pretty much on the compute side.

And just before we run out of time, the last end market is communications. What are you seeing in that market? I mean, I think you've obviously been a content gainer in there as well. I think people look at it saying 5G is going in its fifth year it's a little bit long in the tooth. How do you net those 2 together? And is it still a growth business?

Yes, 5G continues to grow. The strength is actually very strong. you say 5 years, but a lot of the regions adoption just get started, right? Next year, Europe, India. So if you look at our 5G business, not only we cross $600 million run rate a couple of quarters ago, we continue to see strong demand and the strength of the business. So going into next year, we definitely -- we are quite confident about the growth trajectory for 5G wireless. On the other side, wireline, we are quite cautious, right? It's a business that typically in a good time, it's a normal GDP kind of growth if economies kind of downturn, then typically it will decline. But we feel good, the wireless business growth is going to more than offset any wireline decline in the next year. And longer term, definitely, we have more content gain, right? Not the 5G adoption globally, but also we'll have new content coming in to play 5-nanometer or beyond.

You saw there was a press release with one of the key customers, right? So very, very pleased with that collaboration, right? It's across multiple chips now. So yes, to Jean's point, I think in the next year, we still benefit from the overall 5G adoption worldwide because we didn't have much content in 4G. So even if you can argue the CapEx is slowing down, it doesn't matter. You're still going to spend more on 5G than 4G So that's a natural kind of growth driver for us. And to Jean's point, we still have very significant content, which is going to play out over the next few number of years.

That's it. Thanks for joining.

Thank you.

Thanks, Blayne.