Okay, great. Welcome back, everyone. Thank you so much for joining. My name is Toshiya Hari. I cover the Semiconductor and Cap Equipment Space at Goldman Sachs. We're very pleased, very excited to host the team from Marvell this morning. We have Willem Meintjes, CFO, and Ashish Saran, Senior VP of Investor Relations. As always, I will kick off with a list of questions, but I will certainly open it up to the crowd, so please, have your questions ready. Gentlemen, good morning, and thank you for-
Good morning.
Supporting the Conference.
Good morning.
Before we dive into what you guys had to report and guide last week, Willem, I just wanted to kick off with sort of a icebreaker.
Sure
Type of question. It's been, you know, several months since your appointment as CFO. You were obviously, the Chief Accounting Officer, and treasurer since 2018, so I'm sure there's, good continuity in terms of how you operate and run the business. Curious what you've been focused on as CFO, since your appointment, what have been the key learnings so far, if any, and what changes have you introduced to the company?
Sure. Yeah, thanks, Toshiya. Just as a reminder, I've been with Marvell about seven years. I joined pretty much a couple weeks within when Matt joined, have been part of, you know, all of our strategic work that we've done, you know, pivoting the company to data infrastructure, all the M&A and integration that we did. So, you know, I really sit up and work with the team on all of our sort of internal processes. You know, I think what's changed really in the last couple of months, quarters, is just how dynamic the macro has been.
Obviously, we went into a very sort of significant inventory correction, and then the sort of inflection on AI, and just having to move really quickly to stay ahead of those. you know, in terms of just changes, you know, we've been through this upcycle last couple of years with supply chain and really investing in our working capital. We're on the other side of that now, and so have been very focused on free cash flow and just really optimizing our working capital, optimizing inventory, our DSO, DPO, all those, and so really focused on that.
Got it. Any surprises or key learnings? Again, you've been with the company for a long time.
Yeah
Wouldn't think so, but-
Yeah
Curious.
No, really, no surprises, Toshiya. Yeah.
That would be a problem?
Yeah.
Right. Right. Yeah. Okay, I guess, shifting to what you guys reported last week and I guess guided to, solid results, very solid outlook. I think many of us were positively surprised by the resiliency in your outlook. I guess maybe give you the opportunity to talk about some of the key highlights in the quarter, perhaps what surprised you? In terms of the outlook, to the extent there were any, you know, end markets or product lines that improved in terms of the outlook, maybe touch on those.
Sure. Maybe just starting with the outlook, you know, clearly on data center, we guided flat, with the cloud actually resuming growth, you know, a nice 10% growth. Really seeing that inflection and, you know, a lot of strength coming from connectivity product on driven by AI. Saw an inflection on storage, right? You know, for Q1, we were sub $100 million, and, you know, if you go back pre-pandemic, that was really a $200 million, you know, run rate business. That's really inflected back up to, let's call it $100 million here in, you know, for Q2.
Starting to see some good growth on the, on the data center cloud side. you know, when we look at the rest of the, the business, we did indicate enterprise softening in Q1. That's taken a little bit longer to start, but certainly we're seeing that inventory correction in Q2, and we expect that sort of to continue, you know, just to be soft through the rest of the year. Wired's still very strong. sorry, wireless. you know, we expect that to continue to be strong through Q3, and then we should sort of take a step down. you know, automotive keeps on chugging along, you know, nice year-over-year growth.
Yeah, I think, you know, definitely good to have the bottom in on Q1. You know, we have a lot of visibility into our orders, you know, driving sort of the AI upside on the, on the back half, which gives us confidence on sort of accelerated growth going into the second half.
Got it. I guess on that point, on AI, I think this was the first time you disclosed AI revenue and sort of framed it for us. Just to level set, everyone here and on the webcast, what is included in AI? I n terms of products? I think you sized it, I think last year was $200 million. This year, more than doubling, so $400+ this year. What's included in that? How did you define what's included and what's not included? To the extent there's overlap with cloud-optimized silicon, which was the previous and still the current definition that you guys speak to what are kind of? Yeah.
Sure, sure. A couple pieces there. Let me unpack that. You know, we decided to go through a very robust process to classify what's AI, and really, you know, looking at it in 2 main buckets. The one is on the networking side, and then other one on custom cloud-optimized. When we looked at networking, you know, won't go through everything, but sort of 2 of the main things is really on the optical interconnect. That's the PAM4 800G today, going to 1.6T and really today, the primary application for that product is in these AI clusters.
You know, it's not being rolled out widely in it, so it's a fairly straightforward classification. So we do it in our system by part, by customer. Then something like DCI, the Fortnite ZR product, we look at, you know, we sort of had a run rate business there, and then that just upsided very significantly. So we had a conversation with the customer saying, "You know, what's the end application?" It was very clear that it was being driven by the AI, rollout to their different data centers. So able to very cleanly sort of classify what percentage of that is going to AI.
In our system, we've tagged these things and have gone through many reviews and actually had our audit team review our process and all of our rule sets just to get, like, external validation on it. That we had a very clear. Today, I can just log in, and I can see that breakdown at any time of all of our revenue, of all of our forecasts. It's very systematic. It's not just sort of. The key point we make is there's some areas where, you know, we know that there's some impact from AI, but it's really, you know, early, and it's not very. So, for example, storage, we just decided not to include in this bucket. I think that was the first part of your question. Remind me, what was the second?
I guess, well, I guess you touched on this, but to the extent there is overlap with optimized silicon, because I think it led to some confusion.
Yeah. Absolutely, yeah. We're not giving exact numbers, but I think broadly, the way we framed it is, originally, when we, you know, went back to our Investor Day, we scoped the size of the cloud-optimized opportunity as, like, 400 going to 800. When we looked at the lifetime revenue driving that revenue, we said originally, when we scoped it, about 20% of the lifetime revenue had AI application. That's grown so significantly that now 50% or more than 50% of it is actually, from a lifetime revenue standpoint, related to AI. So that's a sort of a way to think about, you know, when you, when you look at cloud-optimized, what's that overlap? Mm-hmm. To be clear, on the cloud-optimized side, it doesn't include the optical interconnect. It's the custom product.
Yeah, maybe, Toshiya, I can add just a little bit more.
Yeah.
On the $200 million we said last year, it's almost entirely optical. It more than doubling this year is also mostly optical, right?
Mm-hmm.
These custom opportunities start to go into volume production next year.
Mm-hmm.
Optical will grow again next year. We did say, even after we more than double this year, we're gonna more than double again next year. Right. That will have a combination of the optical product continuing to grow with 1.6T in particular, and layering on top of that would be our custom silicon products, which go into volume production. We've got a few programs. Some are already taped out, one is gonna tape out fairly soon, sample by the end of the year. That's how I would think about it.
Got it. Got it. I guess, and that was sort of my next question, but how should we think about your visibility there? The one sort of, you know, question or pushback that we get is, "Well, three months ago, you guys had lowered your opportunity for cloud-optimized silicon this year from $400 million- $200 million. What gives you the confidence to guide out to calendar 2024, or fiscal 2024?
Let me kick it off. I think first is, just to be very clear, you know, when we sized these opportunities at $400 million, that was about three years back. I think we tried to put them a little too accurately, quite frankly. I think our confidence on that has not changed at all, just to be very clear, right? Those are going into production volume this year. It got pushed out maybe a quarter or two, right? I don't think the magnitude really changes.
What's changed is that the overall magnitude's become a lot bigger because of the AI inflection, which Willem talked about. I think in terms of our AI revenue confidence, I would say for this year, you should imagine those orders are already booked at this point. Customers are chasing also parts at this point, as you know, as is very clear from us and our peers, you've seen a very significant growth in AI demand. I think for this year, you should assume those orders are not just in, we're in the process of basically scheduling this as soon as we can. I think as we look at next year, we certainly feel like, you know, on optics, the trend will continue. We get an ASP kicker because we go from 800G to 1.6T, it's a 2x the bandwidth.
You know, we get a pretty nice premium for that. On top of that, you know, we've got multiple programs, as I just mentioned, which are very close to have either taped out and already in customers' hands for qualification or will be within the next few months. I think it's a combination of both those things which, you know, give us the ability to talk now about very significant revenue growth in the near future.
Got it. Maybe this is a little too granular, but in terms of contribution to growth next year, is it roughly 50-50 between optical and custom, or is it skewed one way or the other, or?
Nice try, Toshiya. I would say you should imagine it's both, but certainly from a percentage basis, right? You've got these custom silicon going from almost nothing this year to a meaningful number next year. I would say you should expect strong growth from both the product lines.
I would just add, just more broadly, right? When we look at next year, I think there's a couple of additional drivers. You know, when we look at storage, we don't expect that to come back all the way this year, you know, that should be a nice contributor next year as well in data center, as that sort of comes back.
Got it. Okay. Maybe a follow-up on the optical business. You obviously have a very dominant position here through Inphi. I guess you talked about this a little bit, the 800G transition, where are you? Where are your customers at in that transition? The 1.6T, which you announced, I think several months ago, a couple of months ago, what's been the early feedback from?
I think this is another example of us being significantly ahead. I mean, this is a market which Inphi and now Marvell have been typically at least 2+ years ahead of pretty much everybody else. We introduce our 800G part almost 1.5 years, 2 years back. It was shipping in production volume last year, right? Where versus alternatives are still in the process of qualifying as we speak. I think because of AI, as I mentioned on our call, the rate of growth is significantly higher, not just in terms of units, but in terms of adoption of new technology. Typically, AI is running at, call it 18-24 months from one transition to the other. We were the first and the only, as far as I know, who've announced a 1.60 product.
That was at OFC earlier this year. Basically, it's already in customers' hands as we speak, and it's going to be in volume production next year. I think we'll be far ahead of almost everybody else in the market. I think AI, in particular, is one of the reasons why we feel this is going to be one of the bigger growth drivers for us going forward.
Got it. Yeah, maybe it's a good place just to talk about, you know, this optical application in AI and sort of what we're seeing, and maybe, Rajesh, you can expand a little bit on from an architecture standpoint and why it's such a critical part of.
I think, you know, one of the questions we were asked last quarter, right, on AI is, look, as some of the spending shifts towards AI at the cost probably of standard cloud infrastructure, is that good or bad for Marvell? I think this optical example is a perfect one of looking at the total optical density of connections within an AI network is significantly higher. Just to give you an idea, you know, in a standard cloud infrastructure, a server typically runs 50 to 200 Gb per second , which doesn't even need an optical connection. That's a copper-based connection going to the server top of rack, right? Versus in an AI system, you basically have each individual accelerator running multiple terabits, and the total system is driving somewhere between 20 and 30 terabits. That's multiple optical connections, essentially.
We see that the bandwidth will continue to grow, the number of connections will grow, and the speed will increase. That's one perfect example, and the other one we talked about, certainly, which we can touch later, is we see a much bigger compute TAM available to Marvell in AI versus standard cloud infrastructure, and we can certainly talk about that later today.
Right. Right. I guess before we go there, just on linear drive, quite a bit of buzz at OFC. I think in terms of incoming, it's calmed down from immediately after OFC. How should we frame sort of the credible threat from linear drive?
Yeah, look, we, this is, I mean, the reality is this market is moving so quickly. You know, at this point in time, I think people are not waiting for an older technology, which is an analog technology, to be proven. I think linear drive is a very interesting idea, in terms of, you know, lowering total bill of material, right? From a practical perspective, tuning every single port in a cloud application in the field is just not very feasible, and that technology is still in the demo stage at 800 gigs, which we've been shipping in volume, and the reality is, by the time they can probably qualify it, the customer has moved on to 1.60.
I think there's just a massive gap in terms of timing, and there's a number of practical limitations, which is why you really haven't heard much more beyond. It's a very interesting idea, and I'm sure there will be some niche application, there always is, but we don't see that as something our customers are asking us for.
Got it. Okay. In terms of Innovium and the opportunity in cloud-optimized switch silicon, it's been, I forget how long since the acquisition, but how has the integration gone? What's been the customer feedback? I think you had thrown out, like, a 12-month forward estimate in terms of revenue contribution when you made the acquisition, but how have things progressed, and what's the outlook?
That grew, so it's a little bit more than a year ago, right?
Right.
Last year, it grew really nicely. You know, I think we had sort of $150 million. It didn't get all the way there, but it grew really nicely. We expected, again, to grow this year. You know, we're really focused on our next 51.2T product. Working very closely with our customers on that. I think it's. We announced the product.
Yeah.
Again, not want to keep saying AI, but it's probably not a bad idea. I think certainly as you go towards these applications, that's going to be one of the key drivers, right? As you see much higher interconnect speeds, and if you want to be able to go to 1.6T on the optics side, the most efficient way to take advantage from a switching layer is to go to a 51.2T switch. Products announced, we should basically be sampling soon, and, you know, it's a product which we look forward to ramping, you know, in the next few years.
Got it. I guess the competitive differentiation relative to some of the incumbents in switch silicon, anything you can speak to?
Yeah, I think this is a market where, quite frankly, you know, there's been one key player for a very long time. There's been a number of startups, including Innovium, which have been trying to go after that particular socket. Like every other large market, customers want choices.
Yeah.
I think, you know, the very simple way to think about this market is, you know, now you actually essentially have someone with a product which is as good, you know, the same low latency architecture, high bandwidth. That's essentially what we offer. We've got a very key large cloud customer who we've been shipping in volume to our existing 12.8T product. You should imagine at 51.2T, there's significant interest from the rest of the cloud ecosystem, so we're very excited to get this product out in the marketplace.
Right. Maybe transitioning to the custom compute business, very exciting area. Maybe talk a little bit about the design win funnel in the business. Are the big projects mostly DPUs for offload acceleration security, or are you likely to participate in sort of the CPU. O n the CPU space as well?
Yeah, you know, we've been fairly consistent on, like, we can't comment on too specifically on the application, right? Certainly there's a mix there, and it includes compute, right? You know, when we look at the design win funnel that's ahead of us, you know, it's extremely strong. You know, probably, I think the number is, like, 50% plus that's AI related. Yeah, clearly a huge opportunity. As you sort of alluded it, to it, right, when you look at this architecture shift from sort of a CPU-centric with some fraction of acceleration, you know, when we see that move towards where acceleration is really the core, and then there's some CPU component, that creates a massive opportunity for us.
The way we see it is that that pie is growing very significantly. You know, we don't look at it in terms of, "Hey, we're competing with NVIDIA," or anything like that. I think there's so much opportunity in that market and really being able to complement, you know, and having, you know, offerings that allow, you know, some of our customers to really optimize what they're doing. We see that trend continuing. Yeah, very excited about that opportunity.
Yeah, I think what I'd add is, you know, we entered this market, call it about 3 years back, post the Avera acquisition, which was the, you know, the old ASIC arm of IBM.
Right
Once upon a time, you know, have done lots of very large chips for customers. We've taken that capability, we've accelerated going to our five-nanometer platform. We've added all the IP, which we have from the other parts of the business. If you stand back and think about, you know, the three main things you would want in a, in a cloud data center, which is compute, networking, and storage, we basically can combine all those three capabilities with some very strong in-house security abilities as well.
I think we can provide that as a standard IP, which we do to certain customers, we have the ability to basically build your custom part, and you can pick and choose different parts of the IP blocks we provide and combine them with your own, call it, internal secret sauce, for lack of a better word. I think that's one of the big differentiators we have in the market, and we can do this today at five-nanometer. We're going to 3 nanometers going forward. We have now, you know, world-class, you know, industry-leading services capabilities. I think we basically are a one-stop shop, right? That's what some of the large cloud customers like, that they can come to us, to one single place, and get a variety of different IP products in a custom package tailored to their specific application.
Got it. Yeah. I guess that was my next question. That differentiates you from the rest of the pack in terms of the ASIC and IP providers.
Yeah, we can always do a plain vanilla ASIC, which was the old Avera business.
Yeah.
We've done that very successfully, but I think we have a number of cases. In fact, if you remember, you know, a few years back, you know, 5G was one of the new things we talked about, and it's been very successful. We've grown revenue very rapidly, and one of the keys to success there was our ability to provide essentially a custom baseband tailored to 2 large customers' unique applications, but a lot of the underlying IP was essentially provided by Marvell on the core processing side. It's the same formula now applied to a much bigger, much faster-growing market.
I would add, though, Toshiya, that when you say the pack, there's really two people that's doing leading-edge ASIC at this point.
Right.
Yeah.
Yeah, good point.
Maybe one last question on data center, and then I'll certainly open it up to the crowd. On your storage business, and you gave great context in your prior comments, but I think pre-pandemic, you were doing $200 million-ish- in storage revenue within data center. I think at the peak, you were closer to $300?
Yeah.
A lot higher.
It's a lot higher.
A lot higher.
Yeah.
During the pandemic now you're sub $100, and I think you talked about ultimately getting back to $200-ish.
Yeah.
Is that sort of the steady state, if there is such a thing in this business?
Yeah, I think, you know, when you look at it, Yeah, you know, exiting this year, we don't think we get all the way back to 200, maybe, you know, three-quarters of the way. There's some decent upside next year. We do think it gets all the way back. I think the piece that it's very early, and it's hard to know exactly what the impact is of AI, right? I mean, the, you know, feasibly, you know, there's a scenario where it starts creating a lot more data.
Sure.
You know, what that looks like, you know, when you intersect the sort of density in increases, sort of hard to tell. you know, sort of baseline, I would expect it to get back there, and then sort of a low, you know, low to mid-single-digit grower.
Yeah.
Yeah.
Yeah, we do expect there's gonna be growth in data center storage, right? Both on the hard disk drive side, Nearline remains the only way of mass storage, right, in terms of any workload you can think of in cloud applications. On top of that, we have a very strong and growing position on the SSD side.
Right.
You know, we pivoted our SSD business completely to data center about five years back, and we won a number of designs which are now ramping in production, and we have even more new design wins in next-generation PCIe Gen 5 and Gen 6, which we have not gone into production yet. You should expect that overall data center storage, once it re-normalizes post the inventory correction, is certainly a growth business. To Willem's point, we don't count on fast growth there, but we do expect it's gonna grow, you know, fairly continuously over time.
I guess the slight recovery you're forecasting of this business is primarily a function of you guys going from under shipping end demand to converging toward the end demand, right?
Yeah.
Your customers still sound relatively subdued in terms of?
That's exactly right. Yeah, even, like, the subdued tone that they have, we're still under shipping, you know, where they are very significantly, right? Exactly to your point, like, we just assume sort of getting back to shipping to what they see as end demand towards the end of this year.
Yeah. Great. Maybe I'll pause here and see if anyone has any questions. Okay, we'll keep going. Enterprise networking, I think you said, you had expected a correction in the quarter. It got pushed out, and now you're guiding this business down 10% or more than 10% sequentially. Where are we in the correction process? Do you have visibility into customer inventory? How should we think about-
Yeah.
Going forward?
Yeah, I mean, the demand is still actually fairly healthy. You know, I think, as we said last quarter, we were trying to get ahead of that inventory correction, and to me, it started, but we had some strong ramps in that custom ASIC product that offset it. Like, you know, we saw pools continue to be a little bit stronger than we had expected. This quarter, we do expect a step down, as you mentioned, but over 10%. Again, I think we're trying to get ahead of that inventory correction a little bit. But we do think that, you know, based on what we see from the end customer, it's probably soft through this year.
We don't really, you know, in our, in our internal model, you know, at full cost, a recovery by the end of the year. Sometime next year, you know, Now, you know, I think the key thing there to highlight is we've taken a ton of share, right? Our And, you know, when you look at how significant we've grown that business over the last couple of years, it's really driven by share gain.
Right.
We don't think we don't believe we're losing share, it's just more in inventory correction. Once that growth returns, you know, we'll be part of that. I don't know if you want to add anything?
No, I think you captured it. I mean, we're basically being cautious, I would say. I think we've guided for a pretty decent step down in Q2. You'll see a little bit moderation over the next few quarters.
Got it.
Unlike some of the other market stores in particular, I think this one has been managed even by our customers in a much more, you know, clean way. There's no real surprise here. I mean, we were actually thinking of bringing the numbers down in Q1. That's what we guided to. We just never got there because customers kept pulling. I think we're just trying to get this thing more normalized, so I don't think of it as a cautionary outlook, but we're not seeing anything which is very different than what we expected.
Got it. I guess the medium to long-term outlook for this business, I mean, to your point, you've gained share. I think there was good content growth with multi-gig. three years out, ±, how do you think about the sustainable growth rate in enterprise networking?
Yeah, we continue to invest in these products, right? I think we have very competitive product offerings. I think we'll continue to do really well. You know, I think it's a changed world in terms of enterprise. I think, you know, people were very skeptical of that, but we've seen that sort of continue to grow, right? I think, you know, there's all these sort of applications from a, from an AI standpoint that's very nascent right now. I do think, you know, sort of the connected workspace, you know, that whole trend, more data, I don't think any of that changes. I think, you know... I don't know if you want-
Yeah, I think the end market, once this inventory it's still a growing market, but it's more like a GDP ±, and then we've generally done better. We did a lot better last year. We grew, like, 50%-100% faster.
Right.
I don't think that's gonna be a tough one to replicate, but I think we'll certainly continue to grow above kind of what the end market does. You know, in terms of things like the next generation Wi-Fi is still being rolled out. That generally puts more stress on the network. You know, multi-gig adoption is still fairly low in terms of total, you know, buying within enterprises. There's still, probably another 3-4 years it's gonna run, which has some very nice ASP uplifts. Overall, I think we'll continue to outdo kind of what the end market does. It's just a couple of quarters to get through, essentially.
Got it. In 5G, you've grown that business really nicely over the past couple of years. Ashish, to your point earlier, good design wins, particularly that's a couple of your base station customers. You grew that business 25% sequentially last quarter, I think you're guiding that up this quarter as well. I know you went on to guide, I think October flattish or up, and then-
Up, yeah.
Up, and then January down quite-
Yeah
Quite significantly. Very transparent. That's super helpful. How should we think about the medium term in this business?
Yeah.
It's very lumpy.
That's right.
I guess 5G, you're kind of approaching the mid to late innings from a deployment standpoint.
Yeah.
How should we think about this business?
Yeah. You know, when you look back at, like, 8, 9, 10 quarters, you know, maybe there's, like, 1 exception. We've basically had quarter-over-quarter growth over this whole period, so very strong for wireless business because, as you mentioned, it is, you know, typically very lumpy. You know, 25% growth in Q1 and then, sort of continued growth for the next couple of quarters here. We do then expect quite a significant step down and sort of long down. You look at the medium term, like, to your point, you know, I think there's some pause and so, you know, it'll probably be weak for a couple of quarters.
You know, long term, I think there's a bunch of, you know, rollout that's still ongoing, combined with we have another design win that's really only ramping second half of next year with one of our customers. That's ahead of us. You know, once you get through this normalization period, we do expect growth to continue, over the call it medium term.
Okay. Beyond India, which geos are you guys sort of focused on as you look out?
I think Europe would be the next one, I would say. I mean, that, you know, there has been some deployment for sure, but it's still been on the slow side. Some of it's macro impacted, for sure. I think also remember, even when you look at India or even the U.S. as an example, while, you know, there's certainly a, a big phase of initial building in the first year, and there's typically a pause, you typically can't upgrade an entire large country's network in one year. Typically, it takes about 3-4 years, right? Initially, you do what's called population coverage, where everybody gets a 5G signal.
Over time, you have to go actually add capacity, right? If you live in any major city, you know exactly what I'm talking about, where your throughput isn't actually as good as you think, your battery doesn't last as long. The reason, it keeps hunting for an active open channel, so you need more capacity. I would say, once you're past this first stage of deployment, which is, "Hey, I want to make sure all my subscribers get a 5G signal," then you start adding capacity, which typically is in year 3 and year 4, right? You're going to start to see that as well. I think it's a combination of additional geographic capacity coverage, you know, coupled with some additional sockets, as Willem talked about. We still expect, you know, decent growth out of this 5G business.
I mean, we said it hit $600 million kind of run rate ending last year. We just grew 25% in Q1. You can see this business has gone from, you know, five, six years back, was basically sub $100 million. It's done extremely well for us. We still see pretty nice growth in front of us here.
Got it. Perfect. Transitioning to automotive, again, it's been a really nice grower for you guys. It's still low single-digit % of your portfolio, not necessarily a big business, again, in terms of contribution to growth, it's been a good one. Maybe talk about the medium to long-term visibility you have in this business. Auto is an end market where investors are increasingly concerned because it's been so strong for so long.
Yeah.
Should we be worried about SAR? Yeah or because you're tied to, you know, secular growth, i.e., Yeah Ethernet adoption, you're good?
Yeah. No, it's a great question. You know, because we don't have a legacy business there, like, all of our ramps are sort of product into new platforms, basically. Mm-hmm. A lot of it is driven by, you know, both the EVs but also internal combustion, so it's not just EV. Specifically, that dynamic of, you know, new model years taking up Ethernet, and it's not replacing anything for us, it's just a new ramp. We're less exposed to sort of the unit story on, okay, you know, units might be up or down.
We're sort of agnostic as we sort of ramp into these new platforms. You know, we've continued to see very strong design win traction there. I mean, as you know, like, the automotive cycles are very long, like, for that to translate. You know, the what we're seeing today is sort of work that's been done over multiple years. When we look out ahead, you know, we've got a very significant design win funnel that we still. We expect multiple years of growth here to continue. Certainly, you know, we expect growth this year, we expect growth next year, yeah, it's a very good business.
Yeah, I think what's changed is maybe just kind of pivoting your question a little bit. If I look at the design wins we won about three to four years back, those tend to be singular in nature, meaning it was sort of new technology for car OEMs, especially traditional OEMs, where they would adopt you in one particular model type. What's changed is the design wins, which we have won in the last year to two years, have now been across multiple product lines, because now the technology is proven. Now it's not a question of, you know, if, it's more a question of when: When do I adopt Ethernet? It's pretty much a de facto standard at this point.
The volume of design wins is significantly higher, and at this point, we have pretty much 8 out of the top 10 OEMs from a unit perspective worldwide. You can think of any major manufacturing region, you know, in Asia, in Europe, in the US, where you can pretty much tick off all these key customers. To your visibility, you might ask, that's the beauty of automotive is once you win something, it's a 7-year design cycle. There's multiple years of spares after that, typically, and once the technology... Think of this as the nervous system of the car now, right? This is not a piece part. Your entire electronic ecosystem is being designed around Ethernet as the backbone. That's a very, very sticky set of sockets.
That's why I think, to your point, it's a small part of revenue, but it's reuse of existing technology, Ethernet, which we've always had on the enterprise side, so it's a great leverage for us, and it's a very sticky product, and over time, it can lead to even more opportunities, not just on the Ethernet side, but in other areas as cars become more and more automated.
To that point, Ashish, maybe talk about the compute opportunity within automotive. Sure. What's sort of the customer engagement?
Yeah, it's exactly the same model. I mean, we started down this custom model essentially on the compute side in 5G, right? We talked about 2 key customers. They're ramping in high volume. We took the exact same idea, now it's playing out on the cloud side, right? We've discussed a cloud-optimized platform, where essentially we're solving a very specific compute need the customer has, which they cannot satisfy effectively, either from cost or power or utility, from a merchant product. If you fast-forward and think about what a car 5 to 10 years from now looks like, compute's a very big part of the differentiation, and that's a big part of the key IP. Having a vertical stack, meaning hardware, software, firmware, integrated and done by one single person, is a big differentiator.
We've seen this play out in other ecosystems like, you know, a phone is a perfect example of that. You can imagine, if you're a car OEM, you want to find a semi partner who can do these custom capabilities without you having to invest, right? Thousands of engineers in building basically one or two chips a year, right?
That just doesn't make a lot of sense. That's the opportunity, is what we see it. It is early days, I would completely agree with you. I think this is a discovery period, where people are trying out different things. There's going to be a big merchant opportunity, which will always be there, but some of the OEMs will also find a way to saying, "Hey, we want to own our own IP, but we don't want to necessarily create a mini Marvell internally to do it." We see that as an opportunity. We'll keep you updated. It is early days, but it can be quite exciting going forward.
Got it. I guess, given everything we've talked about in terms of, you know, revenue dynamics and your outlook, curious the long-term CAGR that you had presented at your 2021 Investor Day of 15%-20%, is that still the right range to think about and assume? I think there was a moment in time where I think the market collectively was a little bit skeptical.
Yeah.
At this point, what are your thoughts on?
Yeah, absolutely. you know, I think when you, when you look at, you know, the overall, sort of driver for that, it's obviously very much leveraged to data center and cloud. you know, so when you go back a couple of quarters, I think there was discussion around moderation and sort of the overall CapEx in that space. and that's clearly now pivoted to, you know, sort of continued growth, you know, pretty strong growth. And then, you know, Ashish touched on the question earlier, but it's really, you know, is this sort of momentum of shifting dollars to AI going to impact your overall growth rate? Our view is, it's actually really good for us. you know, I think the content that's being driven, you know, directly is much higher.
Over time, you know, we see the pull-through on board and networking being very significant. We touched on the storage, and so, you know, we still look at that sort of 15%-20% as a very reasonable target from a, from a growth rate. I don't know if you want to add anything.
No, I think you hit it, basically. This is the whole focus on shifting our data infrastructure. That's what, you know, drives the economy in different applications. We actually did better than that, quite frankly. If you go back and look at our performance over the last couple of years, right, you could almost argue we overshot it. Even if you normalize it, I'd say we pretty much grew in that pattern, and this was even before, you know, the meat of some of these big, chunky sockets on cloud-optimized actually hit our revenue. In fact, almost none of that actually has, right?
That was a big part of that long-term growth, and that's what's in front of us for the next few years. Automotive was very nascent, and now it's becoming more meaningful. It's in hundreds of millions of revenue. I would actually argue that, you know, we probably had a more difficult task in the last couple of years. Now, given that some of these sockets, which we've already won, are very imminent to ramping, are all in front of us. Yeah, we feel very confident about that kind of growth rate going forward.
We haven't touched on it yet today, but, you know, there's other opportunities out there, like, you know, CXL, that we do think is very promising. You know, we're investing there, and, you know, at some point, you know, over the medium long term, that does turn into, like, a very nice opportunity. I think this, the whole AI sort of trend is accelerating that, where you could really see, you know, having the memory be able to be shared as would be a big benefit to that architecturally.
Got it. Yeah, that's helpful. I guess we have five minutes left. I want to pause here again to see if anyone has any questions. I'll keep going. Maybe on gross margins, you know, three months ago, we were a little surprised-
Sure.
With the guide. I think you guys characterized it as sort of the perfect storm.
Yes
Where mix just kind of flipped...
Yes
On itself. you are guiding to, I guess, 64% or the low end of your long-term range...
That's right.
Exiting the year.
That's right.
Maybe provide a bridge, if you will.
Yeah, sure.
between, you know, 60.5%.
Sure
Today, so.
Yeah, I mean, certainly, like, perfect storm is probably a good way to characterize it. You know, just some of the puts and takes. We really, you know, saw a lot of our sort of core, sort of higher than average, gross margin product, be impacted by the inventory correction in an outsized way. We saw sort of carrier and ASICs, which typically are below our corporate average, just sort of power through that cycle. The combination there really created a significant headwind from a gross margin standpoint. Now, you know, clearly, you know, we started taking very significant actions to manage our costs.
I think, you know, as we went through the supply chain, sort of shortages last year, you know, for the most part, you're a price taker, right? You have very little ability. But one thing that we were doing as we went through that was, you know, getting multiple sources for the back end. As we've now come through that, you know, we're in a much stronger position to go and negotiate our pricing. Really drive very hard intently on efficiency around, you know, any, you know, test times, you know, looking at the engineering, really sizing our operations team for this level of revenue.
All of those actions, you know, go effectively into inventory, and you have to, you know, go through a couple of quarters for that to flow through your P&L. We're very actively driving improvement. Setting aside that component, you know, if I bridge you from a mix standpoint but going forward, some of the main ones is really this, the wireless step down in Q4. I think that's a very significant tailwind for us. You see storage, you know, continue to recover. That's a nice tailwind. All the optical networking products, strong margin, and we see that growth accelerating. You know, consumer moderates from here into the back half.
You know, certainly, the majority of the story is mix, but then also, the actions that we're driving from a cost standpoint, will benefit us, not through, you know, not only through Q4, but, you know, we see that, you know, continuing through next year.
Got it. Okay. Maybe the last couple of minutes, I wanted to focus on capital allocation.
Sure.
You know, historically, you guys have been, or had been quite acquisitive, and very successfully so. On the call, you talked about using cash on your balance sheet to reduce debt, which is due, I guess, next month...
Yes.
very soon.
Yes.
How should we think about, I guess, M&A?
Yes
Going forward? Does that play a role in Marvell, or are you pretty comfortable with the IP blocks and all the, you know, tech that you have internally? Maybe talk about buybacks as well, so.
Sure. Yeah, a couple of questions there. On M&A, you know, we've got this history of doing M&A. You know, it's certainly one of our core competencies to sort of identify companies that's a strategic fit for us, and then integrating those and getting accretion. When we look at our portfolio today, we're actually very happy with what we have. I think we're very focused on organic growth. You know, always open to tuck- ins or if some opportunity comes. That's really not our focus right now. Frankly, you know, when you look at data infrastructure, you know, it's been so consolidated, there's really not a ton of opportunities out there.
You know, from a capital allocation standpoint, as you mentioned, paying down $500 million in our bonds in June, we're very comfortable with our leverage at that point. We have a term loan that's due April next year, we'll be looking to refinance that over the summer. That puts us in a position to really start doing buybacks. Now, our stated policy is to do at least 50% through buybacks and dividends. We'd love to do more than that and really limit the dilution, you know, through SBC dilution. Yeah, we're very excited to start our buyback program in the third quarter.
Then, you know, we'll be opportunistic if we can do more than the 50%.
Got it.
Yeah.
I guess in the last minute or so, any aspects of the business or the broader industry that we collectively underappreciate or overlook? I know Marvell is a well-covered stock, but anything that you feel like we're missing?
I mean, I think, as you said, we're pretty well covered. I think the story for us is very straightforward, right? I think, you know, the business is basically back on a growth path, right? Gross margin's improving. We've got a big tailwind essentially coming from the increased investment in AI. Storage business is going back up, basically bottomed out. Data center bottomed out in Q1. Basically, things are all on a growth path with accelerating growth in the back half. We feel very good about our position going into next year with multiple very key programs going into volume production. Overall, you know, couldn't be more happy with the setup we see in front of us.
Great. On that high note, we'll close the session.
Great.
Thank you so much.
Great.
Appreciate it.
Thanks, everyone.
Thank you.
Thank you.