Good afternoon, and welcome to Marvell Technology Inc. Third Quarter of Fiscal Year 2024 Earnings Conference Call. All participants will be in listen-only mode. [Operator Instructions] After today's presentation, there will be an opportunity to ask questions. Please note, this event is being recorded. I would now like to turn this conference over to Mr. Ashish Saran, Senior Vice President of Investor Relations. Please go ahead.

Thank you, and good afternoon, everyone. Welcome to Marvell's third fiscal quarter 2024 earnings call. Joining me today are Matt Murphy, Marvell's Chairman and CEO; and Willem Meintjes, our CFO. Let me remind everyone that certain comments made today include forward-looking statements which are subject to significant risks and uncertainties that could cause our actual results to differ materially from management's current expectations. Please review the cautionary statements and risk factors contained in our earnings press release, which we filed with the SEC today and posted on our website, as well as our most recent 10-K and 10-Q filings. We do not intend to update our forward-looking statements. During our call today, we will refer to certain non-GAAP financial measures. A reconciliation between our GAAP and non-GAAP financial measures is available in the Investor Relations section of our website. Let me now turn the call over to Matt for his comments on this quarter. Matt?

Thanks, Ashish, and good afternoon, everyone. For the third quarter of fiscal 2024, Marvell delivered revenue of $1.42 billion, growing 6% sequentially above the midpoint of guidance. In addition, on a non-GAAP basis, the Marvell team drove a sequential increase in gross margin, remain disciplined on operating expenses and delivered EPS of $0.41, above the midpoint of our guidance. We are pleased with our results in execution. In our data center end market, revenue for the third quarter was $556 million, well above our guidance, driven by stronger than forecasted AI revenue. We are also encouraged by revenue from cloud, returning to year-over-year growth. On a sequential basis, overall data center revenue grew 21% in the third quarter, while cloud grew well in excess of 30%. As expected revenue from the enterprise on-premise portion of our data center end market declined sequentially in the third quarter, reflecting weakening demand. Demand for data center storage also remains depressed, and industry expectations for recovery have continued to push out. In cloud, revenue from both AI and standard cloud infrastructure grew sequentially, with AI growing significantly faster. Growth was broad based led by our PAM4 optical products. Teralynx Ethernet switches, as well as our data center interconnect, or DCI products. Earlier today, we released a video highlighting our long standing collaboration with NVIDIA, where using Marvell's optical interconnect technology to enable the bandwidth scale and reliability required by Generative AI. Marvell has built a broad product portfolio which are customers are relying upon to power their accelerated computing infrastructure. We are benefiting from strong demand for 800-gig PAM electro-optic products, tightly correlated to the growth of in deployment of AI accelerators. In fact, we are seeing the overall attach rate of our PAM products to accelerators being higher than one to one in…

Thanks, Matt, and good afternoon, everyone. Let me start with a summary of our financial results for the third quarter of fiscal 2024. Revenue in the third quarter was $1.49 billion, exceeding the midpoint of our guidance, declining 8% year-over-year and growing 6% sequentially. Data center was our largest end market driving 39% of total revenue. The next largest was carrier infrastructure with 22%, followed by enterprise networking at 19%, consumer at 12% and auto industrial at 8%. GAAP gross margin was 38.9%. Non-GAAP gross margin was 60.6%, growing 30 basis points sequentially, driven by higher revenue and cost improvements. Moving on to operating expenses. GAAP operating expenses were $698 million, including stock-based compensation, amortization of acquired intangible assets, restructuring costs and acquisition-related costs. Non-GAAP operating expenses were $437 million, in line with our guidance. GAAP operating margin was negative 10.3%. Non-GAAP operating margin was 29.8%. For the third quarter, GAAP loss per diluted share was $0.19. Non-GAAP income per diluted share was $0.41, $0.01 above the midpoint of guidance. Now turning to our cash flow and balance sheet. Cash flow from operations in the third quarter was $503 million, which grew by $391 million sequentially. This significant growth was driven by our relative improvement in DSO, lower inventory along with better profitability. Our inventory at the end of the third quarter was $942 million decreasing by $74 million from the prior quarter. Our DSO was 78 days, reducing by 4 days from the prior quarter. Our CapEx was $54 million. We returned $52 million to shareholders through cash dividends and we repurchased $50 million of our stock during the third quarter. Our total debt was $4.19 billion. Our gross debt-to-EBITDA ratio was 2.21 times and net debt-to-EBITDA ratio was 1.83 times. During the quarter, we issued new bonds and…

[Operator Instructions]. And our first question will come from Toshiya Hari of Goldman Sachs. Please go ahead.

Hi, good afternoon. Thank you so much for taking the question. Matt, I had a multipart question on your data center business. It sounds like the outlook has improved since 90 days ago. Just curious how if you can size the AI business for us where you landed in the October quarter, it sounds like you'll be in excess of $200 million in Q4, but the outlook there. And then more importantly, into calendar '24, if you can speak to visibility you have across your optical business as well as the compute business that would be helpful. Thank you.

Yes. Thanks, Toshiya, for the question. We're very pleased with the performance of our data center business grew over 20%. In Q3, we're guiding it up 35%, in Q4, as you noted, that is driven by AI and the Q4 exit rate is well north now of $200 million. What's very encouraging as well is that the traditional cloud infrastructure piece of it is also growing nicely. That's come back very strong for us. And those two will be growth drivers for us into next year. We see continued weakness in softness, and this is a broader market statement than the on-premise piece and that probably persists for some time, but the mix now of cloud and AI is so much higher that it's driving the overall segment in a very positive trajectory, both in the third quarter and into the fourth quarter. And then as we head into next year, as you've seen, most of the strong growth we saw in the current fiscal year in AI and cloud infra for that matter has been in the optics area, but we are tracking well for growth their next year as well as the ramp of our custom silicon programs. And in my prepared remarks, I talked about the strong progress we made on new product development and starting to plan for ramps there. So while we're not resizing those specifically, those -- the performance clearly in the fourth quarter as well as what you would how you think about next year is much, much stronger than when we first signaled the AI opportunity for Marvell a couple of quarters back. So I think things are tracking nicely. And as you said, compared to 90 days ago, I think the overall aggregate data center business, as you can see from our results, is doing quite well.

Thank you.

The next question comes from Tim Arcuri of UBS. Please go ahead.

Thanks a lot. Matt I also had a multipart question. So enterprise networking is going to be down about 35% from the peak in fiscal Q4. And the customers are still though reporting that their inventory levels are actually going up. So is this Marvell product? Is this product from another supplier? And can you talk about just the dynamics going on in that segment? And then also for fiscal Q1, do you still think that revenue can grow? I know you said that networking is down and carriers down. But data center would be up. Do you think that total revenue can be up? Thanks.

Yes. Thanks, Tim, and appreciate the creativity on the 2-parter, but no problem. On the enterprise side, it's hard to comment. I mean if you look, balance sheets are still pretty heavy from the OEMs out there. We've been saying, I think since even going back to December of last year that enterprise was going to come down this year and it's just continued to come down. I think there's a combination of both inventory management at the OEM level as well as from some of the projections you see out there at the end market level, a softer and more weakening demand environment. So we still feel very good about our position. We had grown this business significantly from where it was just a few years ago. We had great years the last couple of years and even in the first half of this year. So we're going through what would be kind of a normal inventory correction cycle. It's taking a little bit longer than we thought if you went back to the beginning of the year, but I also think the macro and the environment has deteriorated more than we would have anticipated at that time. And so we see it down, and we see that having to work through that issue for the next couple of quarters. On Q1, while we don't guide specifically, I understand what you're looking for. I think the way to think about it is that, and I guess I gave the information. Carrier is down after a really great run in Q4, that's going to stay weak. The telco environment and CapEx spending is very constrained out there and the end customers seem to be having some trouble. We talked about enterprise being down. And then on consumer, which actually did…

Wonderful. Thanks.

The next question comes from Vivek Arya of Bank of America. Please go ahead.

Thanks for taking my question. Matt, so I think in the data center, the value of the optics business, I think, is well understood and appreciate it. The challenge is still for us to how we value your ASIC business. So is it 1 or 2 customers? Is it more customers? How is the visibility for the next 1 or 2 years? Like are we talking $100 million to $200 million next year? Are we talking $300 million, $400 million next year? Because unless we have a good way of sizing what this business is and what the visibility and what the growth potential, it's just very hard to value and give appropriate value to Marvell for this business. So could you just help us understand what is the right way you think about your ASIC business, can the lumpiness here really swing your sales next year?

Yes. Thanks, Vivek. I think a couple of ways to think about it. The first is on the customer opportunity, just by design, it's highly concentrated, if you think about it. And there's just a handful of companies that can really drive the kind of silicon TAM opportunity that that's out there. And you've probably seen in the last few weeks, there's just been a tremendous number of announcements across the industry around AI, whether it's strategic partnerships that are being announced, people doing their own silicon, behind that, their own silicon. There's typically partners there, people like Marvell, who are going to participate. So a lot of activity, you can see and I would say even one broader statement, then I'll get to your question. We do see as the TAM is moving from traditional computing architectures to accelerated computing, it is really opening up the custom silicon piece of that. And so that we believe will be a larger portion of the TAM going forward. When we talked about sizing, if you remember, a couple of quarters back, our AI opportunity, we had signaled already that this year, we would do about $400 million on the optical area. Well, said another way, we do about $400 million this year in total AI revenue, most of it driven by optics and that next year would be about $800 million. That was the original sort of bogey we put out to help investors to your point, size this. And then you would assume some optics growth, and so you could sort of do the math on what the difference might be. Now AI's got a bigger head of steam than it was back then. The optics piece is higher and we've made significant progress now on new product development in terms of getting the chips actually taped out and through the fab and now starting to think about ramping. But what I would say is, and I've been saying this to investors for about 6 months, we are not in a position to call the ball just yet on how big the custom silicon opportunity can be for next year. It's still early, Vivek. I understand the investor appetite for this. I would say between us and our customers, we're not fully -- we don't fully know how big this can be yet. So I'm going to need a little bit more time to probably size that for you. But what I can say is from when we first talked about AI 2 quarters back and what the opportunity was looking out to next year, it's much, much higher. And that's both on the custom side as well as on the optic side. So I hope that helps. I understand the question, but it's a pretty dynamic market as you see what's going on. But the good news is it's all moving in a very positive and upward direction.

The next question comes from Harsh Kumar of Piper Sandler. Please go ahead.

Yes. Thanks guys. Great tremendous execution and very choppy environment. So Matt, a question for you was on the carrier side, you're guiding down significantly down mid-40s on a sequential basis. And I think you mentioned both the wired and the wireless pieces are down. I was curious if you can help us think about perhaps what is taking the bigger part of the hit. And which one do you have more confidence in, in terms of returning back to growth first?

Yes. Thanks, Harsh. And it is a choppy environment, as you point out. Look, I think on carrier, we're coming off a very, very strong Q3. It was over $300 million. And I think if you just kind of infer the whole year, it's going to be -- have been like a $1 billion business for Marvell. So I think that's been a success story for us. Clearly, some -- and we've been trying to signal if you go back over the last couple of calls, we've attempted to signal to investors the fourth quarter that there would be -- that sort of the third quarter would be the peak of 5G stuff, and then it would take some time for that to judge that through. And we bucked the trend, if you remember, on our performance here. So both of those are -- that carrier segment is a strong segment for Marvell. It will return over time, back to a normalized run rate, but it's very lumpy and volatile as carrier usually is. What I would say on the timing, I think they're both going to come back. It's just -- it's hard to know exactly when on these. I think the predictability in the segment is tough. So I don't think I have an exact answer on which one comes back sooner. I'd say wire has been trending down for some time as I think some of that pandemic led infrastructure build has waned. The carrier 5G stuff has really performed extremely well for us this year. And by the way, when it does come back, both on the carrier on the wireless and wired side, we have incremental design wins in those segments. One is incremental content base stations with the Layer 2 processor opportunity and also our new 800-gig DSPs for carrier -- for wired infrastructure, those are going to be new revenue drivers for us when we come back out of this. So timing is still unknown. But I would just say that this is a part of the base business of Marvell, kind of the core part of Marvell that over time will return to a normalized run rate.

Thanks, Matt.

The next question comes from Tore Svanberg of Stifel. Please go ahead.

Yes. Thank you Matt, you mentioned the NVIDIA and Marvell partnership, and I do appreciate the video on your website, but could you add a little bit more color on exactly what this means. I assume this is a multiyear partnership. But yes, any additional color you could share with us would be great? Thanks.

Yes. Thanks, Tore. I think yes, the video was posted today, but it really, I think, just sort of captures a very long-term working relationship between the two companies across a number of opportunities, by the way. I think we've had a complementary and very strong partnership with NVIDIA to really help enable them and their products. In the optical area, we've been working with them for some time, and this goes back even to working with Mellanox in some of their applications. So this was really a way just to, I think, highlight the years of work we've done together and also signify that there's a lot of opportunity for us to kind of double down together on this AI opportunity. So we're proud to be a partner of them and support them in their growth. That's really what's behind it. It's not a new announcement or anything per se. It's just a recognition of a long-standing cooperation between the 2 companies.

Great. Since that was kind of a soft question, I had a question on the enterprise networking business. It sounds like the sequential decline is going to be a little bit better than what we've seen lately. Is that a sign that you're starting to see stability there? Or are the order rates and the visibility is still very limited?

I'd say it's still limited. I think it's still coming down. And I think we're -- we want to see when the customer balance sheets get back in shape to them talk a little bit more robustly about their end market demand, strengthening before we make a call there. So we'll have to see where enterprise IT spending really ends up, Tore, next year, given the macro. And so right now, we're just sort of taking it week by week, and we're monitoring what products they need and looking at their forecast for next year, and we're trying to plan accordingly. But I'd say visibility is pretty limited at this point for the full year.

Great. Thank you very much.

The next question comes from Ross Seymore of Deutsche Bank. Please go ahead.

Matt, I just wanted to ask about your business and perhaps a different way to split it. I think everybody is excited about the AI and the cloud is a part of it and for good reason. But if I took that out of the revenues, everything else seems to be down in your guidance, 30%, 35% year-over-year, somewhere in that range. I really just wanted to get what percentage of that, I think, is down just because of cyclicality, and it sounds like the first quarter might come down again, but should bounce back at some point versus the businesses you are just deprioritizing. So how much of a cyclical snapback whenever that happens, should we expect out of Marvell versus a focus away from some of those non-cloud areas?

Yes. Got you, Ross. Yes, I think maybe I'll start just super fast at the high level and we'll just dive in. I'd say that this sort of supply demand inventory cycle we've gone through in the industry has taken a lot longer than historical to play out. And its sort of the dynamic I would characterize this time is you had various end markets resetting and correcting at different times in a pretty protracted manner. And if you go all the way back to last year in 2022, it started early with PCs and what the smartphones that went to gaming and then it started with data center and cloud, and it sort of had this kind of couple year series of resets, if you will. In our business, we saw these sort of declines in that manner as you talked about earlier this year in data center as an example. And a few quarters later, it's like roaring back as we work through some of those issues. So that's the cyclicality part you're talking about. And you're asking the right question, which is, hey, over time, do these come back or not because maybe we pulled R&D or we sort of have shifted our bets. So I'd say for our business, we are committed to a diversified portfolio. We invest R&D across those various segments, including enterprise carrier. And then within the cloud, there's a number of opportunities in AI. The only one that we really have deemphasized, and this goes back 5 or 6 years ago is consumer. And we've always projected in our models, in our Analyst Days that, that would be a declining business for us over time, both in revenue as a percent of total. But then our investments in the other areas would…

The next question comes from Matt Ramsay of TD Cowen. Please go ahead.

Thank you very much Good afternoon, guys. Matt, I wanted to go back to some of the custom compute and ASIC programs and maybe ask a couple of questions about that. One is, I mean it's been a heck of a year from a gen AI perspective, when you just go back 12 months, we were just starting to hear ChatGPT then the whole thing has blown up. So the customer base all had to think about that, react to that. And so I wanted to think about the programs that you have in flight and the interactions that you have with the customers and the limited set of customers in that space that are all going to be large. Have you seen them lean more into merchant, more into using ASIC houses or more into doing some of the silicon directly with the foundries themselves? There has been a change in that mix. That's the first part of the question. The second part is when you look at AI overall, the leader in that space, which you guys announced a partnership with today, leads not just because of the silicon, but hugely because of software. And I was curious as to whether with the custom programs you're doing for AI compute, are you seeing the customers invest at the scale and software that can make those programs successful at volume over time?

Yes. Thanks, Matt. Yes, I think it's -- you're right to kind of reflect that how much things have changed in basically a year. I think it's a pretty simple answer. On the first one, I'd say, every one of those different sort of segments, if you will, that you mentioned, all of them are up in terms of the activity. You can see it from the leaders' numbers and their guide, you can see it from some of the other companies in the ecosystem talking about ASIC and custom. I mean, look at where we've come from even on that, from having very kind of modest expectations to this being a significant revenue driver for us next year. And then there's various business models, I think that all the large companies are trying to pursue to figure out how to, I think, do all of the above, which is kind of have a nice mix of custom for things that are very, very specific to them as well as take advantage of kind of the market unlock that's happening because of the work of, say, somebody like NVIDIA, right, that's actually helping to create a market. So I think all of that is going to be -- it's not a zero-sum game at all. And I'd say on the sort of customization side and working with those hyperscalers, the design activities through the roof at this point. And so a lot of excitement in that area, both for the custom and then all the stuff we can do around it from an optics and networking perspective. And then I'm not probably the best person to comment on the second one. But clearly, these are very, very well-capitalized smart companies. They understand their customers and they're going to do what they need to do for their workloads and for their custom silicon to make sure, I'm sure that the software and the service offerings are more than competitive, but you've got a great company that's also out there that's driving the market. So I think both are going to exist. And my last point would be to reiterate, this is clearly not like in traditional computing, where it's more of a zero-sum game. It's just highly hard to grow the pie. In fact, it may be hard to keep the pie even flat. In this area, there's going to be a lot of winners and there's going to be a lot of opportunity created for the companies that have the right positioning, customer relationship and IT portfolios to support these emerging needs. But it's very dynamic at this juncture. And I think for us, that's what creates a lot of opportunity.

Thanks. Appreciate the perspective.

The next question comes from Ambrish Srivastava of BMO. Please go ahead.

Hi, thank you very much. Matt, I had a question on gross margin. You surprised us negatively early in the year, and then you gave us good counsel, don't panic. And you're back at 64%. So just wanted to think through for the next year and what's the right way to think about the gross margin profile, given you have customer is coming in, but typically is lower margin than the corporate and maybe it's not for you guys, but what's the right year to think about the margin profile for the company near to medium term Matt? Thank you.

I'll let Will take a victory lap on this one. So, Willem go ahead.

Yes. Ambrish, so thanks for the question. So first of all, really pleased to be guiding back to 64% at the midpoint, right? I think the team internally has done a phenomenal job controlling what we can control in a pretty volatile environment. When we look at it to next year, we've been discussing some of the uncertainty on the recovery on some of our core businesses, right, enterprise networking and storage. And then also the timing and the scale of the ramp on custom. I'll just remind you, prior to this last year, we scaled our carrier and custom business is very significantly while maintaining our gross margin, right? And so looking ahead at next year, if custom on a relative basis grow significantly more, yes, clearly, that would put pressure on our gross margin. However, our view is that they'll be very accretive on operating margin and on EPS. So yes, so that's how we're looking at it. Hopefully, that's helpful, Ambrish.

It is. I had a quick clarification for you, Matt, sorry. You mentioned something about one to one pull attach rate for your business. I was just wondering what was it before. And kind of what has changed the -- what has changed in the dynamic to push it to greater than one to one? Thank you.

Yes. Thanks, Ambrish. It's been a little bit of a journey on this topic. I think if you go back two quarters ago, we were still we were still building models out for how to think about the attach rate and help investors size the opportunity. And we ball parked it at one point one to one initially. And subsequent to that, as we sort of look at the kind of -- all of the optics through up into the switch and in the network. And we've -- I think we've updated -- I think we were really reiterating this. We've updated this number before that it had grown to greater than 1. I think that was more of a clarification we were doing. But initially, it was a little bit of how do we actually get our hands around how big this could be and people wanted to know, balance that against how many GPUs they thought were going to shift. But I wouldn't say that's really incrementally new news. I think that was more of a reiteration, if you will, because it's not just the optics attaching from the accelerators to the switch, but also in between switch layers, and that was sort of the nuance that I think we needed to clarify. So that's the kind of rate on that one.

The next question comes from Harlan Sur of JPMorgan. Please go ahead.

Good afternoon. Thanks for taking my question. After 6 quarters of sequential shipment declines in nearline or capacity optimized HDDs by your customers, cloud and hyperscale excess inventory of drives appears to be normalizing. And then on top of that, you have cloud storage utilizations to keep on increasing. I think your storage customers are cautiously optimistic on sequential shipment improvements on their new 20 terabyte platforms. You guys are still shipping about half the rate of your pre-slowdown run rate on storage. I know you've been growing slightly sequentially the last few quarters. But are you guys starting to see signs of a sustainable pickup in data center stores moving into next year?

Yes. Thanks. I think this is one of those markets, Harlan, where any good news is good news. I mean it's been very, very rough out there for that part of the electronics industry, and you certainly saw the impact on our revenues. Look, I'm encouraged to see the same reports of inventory coming down and some of the end customer commentary. We, I think, aren't going to call this until we really see the backlog build again. I mean it's come off the bottom and it's grown a little bit. And it's -- but even if it's sort of improving off the bottom, at least the end market side, it's still kind of flattish units, right? I don't think it's necessarily off to the races. So we're staying pretty cautious on this. We got hit pretty hard on the downside here, and we're going to cautiously kind of guide our way back into that market normalizing. But I will say the end market signs were hearing and seeing are positive, and it certainly gives us some hope for next year but we're not ready to call any kind of recovery in that area with any certainty or timing. But it has come off the bottom and it's continued to grow, which is a good thing.

Appreciate the color. Thanks, Matt.

The next question comes from Christopher Rolland of Susquehanna. Please go ahead.

Hey guys, thanks for the question. I'm kind of hearing mixed signals on the custom silicon opportunity. And I just wanted you guys to clarify on that. I guess, first of all, are you guys above or below $200 million expectation you guys had for the year? Second, are you expecting this to kind of double off of that $200 million or even $600 million originally that had been an $800 million number you had discussed. And then third, where are we on that $800 million long-term goal that you had? Is that like a 2025, 2026 opportunity? Or is it something beyond that?

Got you. Yes, you're going back to the original like 2021 Investor Day, hey, here's kind of the custom silicon opportunity, which a couple of quarters back, we had reset to $200 million this year and then over time, getting to $800 million. I just want to clarify, that's what you're asking about?

Yes, that's right. I think that was in March. You guys talked about that $200 million for this year.

Yes, exactly. So yes, so we're tracking, I'd say, close to the $200 million, okay, for this year. And then what we had said at the Investor Day and kind of the long term was this $800 million, and that was between, sometime between FY '25 and '26, that was what the -- if you looked at the slide from a couple of years back. And what we had said, I think 2 quarters back or it was a quarter back that, that number would be bigger over time now because of the AI piece of it, even though some of the stuff had shifted around that wasn't an AI. And I think that's still largely on track in that time frame. We never gave an exact kind of -- it's going to happen in XYZ quarter. But in that FY '24, '25, FY '25, '26 time frame it should be able to get towards above that number we gave before which is the $800 million. So I don't think there's any mixed signals. I don't think there's any update, which I think we're -- I think there's some enthusiasm around, but nothing's changed from a quarter ago. In fact, I think the thing that's positive is that the chips are looking really good to go to production for next year, and that was always a risk. Chris, if you remember in our commentary was, hey, these are extremely large die, very complex and a respin either by us or by our customer would -- could cause some delay. And so far, knock on wood, both these programs look like they're in good shape. The real issue is given the sort of growth that we've seen in gen AI. We just don't know the total magnitude, but it will be bigger than what we sized before. And I think we're largely tracking to what we thought was going to happen.

Thank you for that Matt. One last one on carrier. Is Q4 going to be the bottom? Or do you think there could be some more yet to drop? And then I think you have some additional content coming at one of your customers at the end of the year. Is that going to be a meaningful lift for the segment? Thanks.

Yes. So there's -- as I think I said in my remarks, there's going to be continued softness into Q1 in carrier, okay? It's going to take, who knows how many quarters. And it really depends, I think, there'll be some inventory and then you've got to also look at kind of where the CapEx ends up during next year and where carriers are actually going to spend globally on their deployments. But over time, there is a lift. There's a content lift we get from the additional sockets. And we still don't know kind of where -- let's go a couple of years out over the next couple of years where the market share is going to go as well relative to the sort of western suppliers versus the Chinese suppliers on a global basis. So there's a few factors in there that need to be considered. But yes, over time, sort of the number of base stations per year has been relatively well understood. Our content is going to be larger. And it's a solid business for us, I think that we built from a very kind of nascent position just a few years ago. But it will come back and it will have a little tailwind because of the content.

Thanks so much, Matt.

The next question comes from Srini Pajjuri of Raymond James. Please go ahead.

Thank you. Matt, I have a couple of clarifications at this point. First, on the custom silicon, understanding that it's a little bit early to size the opportunity. Given the complexity of these chips and the supply chain issues that we've been hearing about on the HBM and the CoLo [ph] side, I would think that the lead times from your customers are fairly long. So my question is, when do you expect to start building inventory, I guess, in terms of wafer inventory and packaging and et cetera. What time frame should we kind of expect that if you, I guess, are looking to ramp next year?

Well, the planning certainly is happening now, Srini, to your point. And I think between ourselves, our supply chain partners, and our end customers who have a vested interest to make sure this all goes well. I think we've been working quite well as a kind of a team to make sure we've got the necessary capacity that's required from our wafers packaging and kind of third-party component perspective, including HBM. And yes, I mean, all of those are either capacity that's reserved or we're in the process of starting wafers. I mean again, there's multiple programs. So it's not sort of a -- each product is at a different point. But yes, I think the answer is, well -- and I think I was saying this even a quarter or 2 ago, we'll have a lot better visibility kind of come, call it, March when -- what this is going to look like because at that point, given lead times and all the other factors we'll just have much better visibility. So that's sort of the, I guess, the pro, if you will, of some of the complexity we're dealing with is you do have to plan in advance and you do have to start wafers. So all of that is kind of in motion. I would say the capacity planning has been done some time ago. So we feel comfortable with where we are on that, even though it's tight. But yes, we're in that process, but I don't think we're going to get that granular on when exactly we're starting wafers, and we're just going to try to size the opportunity for investors when we have a better view in terms of what the revenue is, but the capacity side is looking to be in good shape for us.

That makes sense. And then my other follow-up, Matt, on your Q1 guidance. And obviously, you said half of your business is going to be down. But I would have to imagine that the data center will continue to be healthy and strong, especially kind of what your customers are talking about in terms of AI investments continuing and also NVIDIA kind of sounded fairly positive about next year as well. So my question is, given the strength that you saw in the last 2 quarters, I mean, are you concerned about any inventory build in your components? Or is that what's giving you pause about guiding for growth in the first quarter for the overall business? Because even if 50% of the business is down because looking at how much this business has grown in the last 2 quarters and if that trend continues, I would have to imagine that the total revenue should grow in Q1. So I'm just wondering what's giving you that pause looking for growth in Q1.

Yes, yes. No, no problem. I'll maybe give you a little bit of a bad time here. I think we're trying to get to the Q4 call first. And then we could do the Q1 call, okay? So maybe first things first. So that being said, I mean, look, I think I would agree with the market commentary you gave, which is year-over-year, we do expect strong growth. I'd say on our side, if you look at kind of the growth we've seen on AI, we're having a very, very strong fourth quarter. And some of that we're catching up from the upsides we had. I mean our supply chain team has done a phenomenal job, right? So I'd just say that we're not guiding Q1, it's dynamic at this point. And I need to really see where the orders flow in and what people really need. And right now, we're just focused on executing Q4. But yes, overall demand looks good. Overall growth next year should be good. But I think trying to guide 2 quarters now with this precision, Srini, is just not going to be helpful for anybody given how fast things are moving. So I'd prefer to give you my Q1 guide when we guide Q1.

Got it. Thanks, Matt.

Our last question will come from Chris Caso of Wolfe.

Yes, thank you. My question is on the cloud part of data center. It sounds like that was an area which may have surprised you a little bit coming back here. Can you talk about the driver of that? And to what extent -- to the extent you're seeing in kind of traditional cloud business is it cyclical or product cycle-driven?

Yes, thanks. Our thesis earlier this year, which actually did play out, I think we had a lot of predictions and in some cases, the market didn't cooperate. But one area it did was that we did not believe that our traditional cloud business was going to be really impacted by the shift to AI. And in fact, it might even provide a tailwind. And our position was that, hey, because of our position in switching and in optics, that broader networking build was going to be required in data centers, in particular, multi-tenant data centers where you had to actually put in increased networking bandwidth to handle the AI capability that was being put in and that's played out. So I'd say there's been some on the product cycle in terms of new products kicking in at that time. But also, I'd just say that it's increased demand, increased demand for products, we added 400 gig and increased demand for our 12.8T switches, things that had been a little bit depressed earlier in the year with inventory correction. So that's been a real positive. It's kind of a nice combination of strong growth on the products we have as well as new products wrapping up -- ramping up, sorry. So yes, and that's played out, and that should be a nice driver and tailwind for us into next year as well. Chris.

Got it. Thank you.

All right, fantastic. Hey, listen, I think we're pretty much at time. I'm going to conclude the call, maybe just a few words. First, I appreciate everybody's interest in the company. A couple of final comments would be despite kind of the challenging macro out there, I'm very, very pleased with the Marvell team and our performance. And I think what it shows is that even when you have a lot of volatility from a cyclical perspective in some of these end markets, right now, it's enterprise and carrier as an example, the diversified business model that we've really put together at Marvell has been able to have some strong offsets to those things. And like right now, we're seeing our data center business come roaring back with strong growth in the third quarter and then a very, very strong growth in the fourth quarter. And that's the commitment we made several years ago. It was to diversify the company by end market, focus on data infrastructure, which, in our view, is going to be probably the best TAM growth opportunity in semis. And we still believe that and while we have some inventory digestion and end market weakness we're seeing in some segments that's going to revert over time in enterprise and carrier. And I would view that as, and describe that as kind of the core foundation of Marvell. And then you have these nice growth drivers with large TAMs on top of it in cloud, in AI, in things like automotive that are going to drive our growth over time. So I think our model is working, and we're managing in a tough environment, and we're managing what we can control at this point. And with that, I will conclude the meeting. Thanks, everybody.

The conference has now concluded. Thank you for attending today's presentation, and you may now disconnect.