Earnings Call Transcript
Marvell Technology, Inc. (MRVL)
Earnings Call Transcript - MRVL Q4 2024
Operator, Operator
Good afternoon, and welcome to the Marvell Technology Incorporated Fourth Quarter and Fiscal Year 2024 Earnings Conference Call. All participants will be in a listen-only mode. Please note, this event is being recorded. I would now like to turn the conference over to Mr. Ashish Saran, Senior Vice President of Investor Relations. Please go ahead, sir.
Ashish Saran, SVP of Investor Relations
Thank you, and good afternoon, everyone. Welcome to Marvell's fourth quarter and fiscal year 2024 earnings call. Joining me today are Matt Murphy, Marvell's Chairman and CEO; and Willem Maintjes, 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. Earlier today, we announced our accelerated infrastructure for the AI era Investor event, which will be held in New York on April 11. Please refer to our press release for more details. We look forward to updating investors on the exciting opportunities we see in front of us from the growth in accelerated infrastructure. Let me now turn the call over to Matt for his comments on the quarter.
Matt Murphy, CEO
Thanks, Ashish, and good afternoon, everyone. For the fourth quarter of fiscal 2024, Marvell delivered revenue of $1.43 billion, growing 1% sequentially above the midpoint of guidance. In addition, on a non-GAAP basis, the Marvell team drove a substantial 330 basis point sequential increase in gross margin, completed execution on the OpEx reduction plan we outlined earlier in the year and delivered earnings per share of $0.46, growing 12% sequentially. As Willem will tell you in greater detail, we also drove another strong quarter of operating cash and increased share repurchases. We are pleased to report these results to you in a challenging macro environment. In our data center end market for the fourth quarter, we delivered record revenue of $765 million, above our guidance, growing 54% year-over-year and 38% sequentially. The strong revenue growth in the quarter was driven by the cloud portion of our data center end markets. While AI has been a key growth driver, I am pleased that our standard cloud infrastructure revenue has also grown every quarter, and we see that continuing next year. Our 800-gig PAM solutions led our growth in the fourth quarter. We also benefited from higher sequential demand for our storage products as that portion of our data center end market continues its recovery. Revenue from our Teralynx Ethernet switches also grew sequentially in the quarter. Turning to the first quarter of fiscal 2025, we expect our overall data center revenue to grow in the low single-digits sequentially on a percentage basis. We expect revenue from both AI and standard cloud data centers to continue to grow sequentially. We project our electro-optics revenue to continue to be strong, and we also expect to benefit from the initial shipments of our cloud optimized AI silicon programs. Partially offsetting this growth, we are projecting a more than seasonal sequential decline in revenue from enterprise on-premise data centers. Over the past several years, Marvell has strategically invested in technology, both organically and through acquisitions, to become a critical enabler of accelerated infrastructure. The seismic shift driven by AI in the data center market is creating new opportunities, and we are actively investing to lead the next wave of innovation. We have in place a full suite of solutions across data center interconnect, switching, and compute, and the ability to uniquely stitch these together into a unified platform, a true one-stop shop for our data center customers. While our 100-gig per lane 800-gig PAM products are currently the workhorse for interconnect inside AI data centers, customers have begun qualifying our next-generation 200-gig per lane 1.6T PAM solutions. We expect first deployments to start towards the end of this year. Complementing our optical interconnect solutions, we expect to start ramping our PAM DSPs for active electrical cables. This is a new and completely additive market for Marvell, and we expect to be shipping products to multiple Tier 1 cloud customers this year. Our DCI products, which provide connectivity between data centers, are critical to our customers' success and continue to do very well. We see exciting new opportunities ahead of us from growth in generative AI applications, driving cloud customers to build new data centers. We also expect a positive uplift from increased investment in inferencing, which will drive more bandwidth between data centers. We are shipping our 400-gig DCI products in high volume today and are seeing strong interest for our next-generation 800-gig products. I am also very pleased to report to you that in fiscal 2024, we significantly expanded our DCI customer base with design wins at multiple data center customers. We expect these design wins to begin ramping next year. As you will remember from prior calls, Marvell silicon photonics technology has been a critical enabler of our DCI modules. You will hear more at OFC and our upcoming AI event about how we plan to deploy our field-proven silicon photonics technology to enable next-generation higher density, lower power interconnects. In data center switching, our Teralynx 12.8T products continue to ship in high volume, and we are on track for production shipments of our next-generation 51.2T switch later this year. The Teralynx product line, which came from our Innovium acquisition, combined with Marvell's very successful enterprise and carrier switching organization. This larger scale now enables Marvell to accelerate our development to address new cloud opportunities, particularly in switching for AI deployments. We are encouraged by the traction we are seeing with both existing and new customers, which has expanded our opportunity funnel for cloud switching dramatically over the past year. Turning to our cloud optimized silicon platform, we're seeing significant progress with the first set of design wins outlined during our last Investor Day. We expect initial shipments for our two AI compute programs to start in the first quarter and are on track for a very substantial ramp in the second half of the fiscal year. Customer bring up of these programs is going extremely well, and we are tightly aligned with them on volume expectations and working in lockstep to enable the production. We now have a clear view of demand for both this fiscal year as well as fiscal 2026. We have been working closely with our suppliers and are confident that we have secured capacity for the ramp. With the visibility we now have for these programs, along with many new opportunities, we are very excited about the potential scale of long-term revenue for Marvell from this business. As the initial set of design wins reaches its full run rate, we expect annual revenue from cloud optimized silicon to have the potential to rival our fast-growing data center optics business, which, for reference, grew to over $1 billion in fiscal 2024. With the exploding investment in AI and accelerated computing, the demand for cloud optimized silicon has grown significantly. We have successfully executed multiple 5-nanometer designs in the last two years and are deeply engaged with cloud customers on many new 3-nanometer opportunities. These engagements are driving a substantial increase in the size of our design funnel. AI is increasing the cadence of new chip releases, and this plays well to Marvell's strength as a key partner for our cloud customers with a proven ASIC platform. We have a broad suite of differentiated IP, including ultra high-speed SerDes, ARM compute, security, storage, and advanced packaging, including die-to-die interconnects and chiplets. We are well positioned to act as a force multiplier for our cloud customers, enabling them to scale multiple custom programs in quick succession. Now let me turn to Marvell's carrier and enterprise end markets together. As we have been communicating, these end markets have been dealing with a period of soft industry demand. As a result, both were down sequentially in the fourth quarter, and we expect them to decline again in the first quarter. On a sequential basis, we expect revenue in the first quarter from carrier to decline by approximately 50% and enterprise networking to decline by approximately 40%. Looking ahead, we expect revenue declines in these end markets to be behind us after the first quarter and forecast a recovery in the second half of the fiscal year. Longer term, these are large and enduring end markets, which are critical to the global economy. As a result, we expect both of these end markets to eventually return to contributing over $1 billion each in revenue on an annual basis once demand normalizes, and we begin to realize the benefits of upcoming Marvell-specific product cycles. Turning to the consumer end market, revenue declined in the fourth quarter as expected and is projected to decline approximately 70% sequentially in the first quarter. This forecast reflects the completion of deliveries for an end-of-life program in the prior quarter, as well as significantly weaker demand from the game console market. Turning to our automotive and industrial end market, revenue in the fourth quarter was $82 million, declining 17% year-over-year and 22% sequentially. As expected, the sequential weakness was primarily driven by a sharp decline in the industrial portion of this end market, where order patterns can be lumpy in any given quarter. Fiscal 2024, our automotive business delivered another strong year, with revenue growing in the double-digits year-over-year on a percentage basis. We are benefiting from growth in Marvell content driven by an increase in the number of Ethernet connected endpoints in cars, coupled with the need for more bandwidth. We've continued to accumulate new Ethernet design wins across a broad swath of automotive OEMs. While the initial wave of our automotive revenue was driven primarily by EVs and hybrids, we now have also on high-volume internal combustion vehicles with major automotive OEMs. These wins tend to be multi-platform in nature, covering numerous models simultaneously. Turning now to our forecast for the first quarter of fiscal 2025, we expect revenue from our overall auto and industrial end market to be flat sequentially. In summary, in fiscal 2024, we delivered total revenue of $5.5 billion. Our data center revenue accelerated throughout the year, growing from about a third of total company revenue in the first quarter to more than half exiting the fourth quarter. As customers continue to shift investment from traditional to accelerated infrastructure, we expect data center to drive the majority of our revenue growth going forward. This transition underscores Marvell's transformation into a leading data center company. AI was a key driver of our data center growth in fiscal 2024, contributing over 10% of total company revenue, well above our initial forecast. This was a substantial increase from approximately 3% in the prior year. Our momentum accelerated throughout the fiscal year with AI revenue well over $200 million in the fourth quarter, driven mostly from optics. In fiscal 2025, we expect this trend to continue driving another strong year for our data center end market. We expect a substantial base of electro-optics revenue from AI to remain correlated to accelerator shipments. As those continue to grow, we expect to benefit accordingly. In addition, we project significant revenue contributions from our AI cloud optimized programs, well in excess of our prior estimate of a couple of hundred million dollars in fiscal 2025. In fact, as our cloud optimized AI silicon programs reach high-volume production, we expect our overall cloud optimized revenue to exceed $200 million exiting the fourth quarter. As a result, on a run rate basis, this momentum would put our overall cloud optimized silicon revenue above the annual $800 million target we had provided at our last Investor Day. And with the full year of contributions in fiscal 2026, we expect to be way ahead of the prior target. In aggregate, we see a favorable setup for the second half of this fiscal year, driven by continued growth from our data center end market, ongoing growth from automotive, and a recovery in carrier, enterprise, and consumer. As we continue to drive revenue growth, we remain focused on strong cash flow generation and returning capital to investors. As you saw earlier today, Marvell's Board has approved the largest repurchase authorization in our history. We view the emergence of accelerated infrastructure as one of the most significant technology inflections of our time. I thank all our dedicated employees and leaders for executing on our strategy to fully capitalize on this remarkable opportunity. During a tumultuous period for the semiconductor industry, we have taken control of our destiny and are strategically investing to win. Earlier today, we announced the extension of our long-standing collaboration with TSMC to develop the industry's first technology platform to produce 2-nanometer semiconductors optimized for accelerated infrastructure. This new platform will enable Marvell to deliver substantial advancements in performance, power, and area critical for next-generation accelerated workloads. We are very optimistic about our growth prospects and our role in enabling accelerated infrastructure. We look forward to updating investors on the massive opportunity in front of us at our accelerated infrastructure for the AI ERA event on April 11 in New York City. With that, I'll turn the call over to Willem for more detail on our recent results and outlook.
Willem Meintjes, CFO
Thanks, Matt, and good afternoon, everyone. Let me start with a summary of our fiscal year 2024 results. Marvell delivered $5.5 billion in revenue with a strong second half performance from our data center end market. Driven by AI applications, our data center revenue in the second half grew by approximately 50% over the first half. GAAP gross margin was 41.6%, GAAP operating margin was negative 10.3% and GAAP loss per diluted share was $1.08. Our non-GAAP gross margin was 61.2%. Non-GAAP operating margin was 29%. And our non-GAAP earnings per diluted share was $1.51. We returned $357 million to shareholders through dividends and buybacks. Moving on to our financial results for the fourth quarter. Revenue in the fourth quarter was $1.427 billion, exceeding the midpoint of our guidance, growing 1% on a year-over-year and sequential basis. Data center was our largest end market, driving 54% of total revenue. The next largest was enterprise networking with 19%, followed by carrier infrastructure at 12%, consumer at 10% and automotive industrial at 5%. GAAP gross margin was 46.6%. Non-GAAP gross margin was 63.9%, growing 330 basis points sequentially, driven by a significantly better product mix as we had expected. Moving on to operating expenses. GAAP operating expenses were $697 million, including stock-based compensation, amortization of acquired intangible assets, restructuring costs, and acquisition-related costs. Non-GAAP operating expenses were $429 million, in line with our guidance. These results reflect the successful completion of our fiscal 2024 cost reduction plan we had outlined at the beginning of the year. GAAP operating margin was negative 2.3%, while non-GAAP operating margin was 33.8%. For the fourth quarter, GAAP loss per diluted share was $0.45, non-GAAP income per diluted share was $0.46, growing 12% sequentially. Now turning to our cash flow and balance sheet. Cash flow from operations in the fourth quarter was $547 million. I'm pleased to report to you our second straight quarter, delivering robust operating cash flow of over $500 million. Our inventory at the end of the fourth quarter was $864 million, decreasing by $77 million from the prior quarter. Our DSO was 77 days, decreasing by a day from the prior quarter. We returned $52 million to shareholders through cash dividends. In addition, we repurchased $100 million of our stock during the fourth quarter, doubling from the prior quarter. We expect to further increase repurchases in the first quarter of fiscal 2025. As you saw earlier today, Marvell's Board has approved the largest repurchase authorization in our history, increasing our current plan by $3 billion, which brings our total available authorization to $3.3 billion. Our total debt was $4.17 billion. Our gross debt-to-EBITDA ratio was 2.19x, and net debt-to-EBITDA ratio was 1.69x. As of the end of the fourth fiscal quarter, our cash and cash equivalents were $951 million, increasing by $225 million from the prior quarter. Turning to our guidance for the first quarter of fiscal 2025, we are forecasting revenue to be in the range of $1.15 billion, plus or minus 5%. We expect our GAAP gross margin to be in the range of 44.5% to 47.2%. We expect our non-GAAP gross margin to be in the range of 62% to 63%. We are forecasting a sequential decrease in non-GAAP gross margin due to lower revenue impacting fixed cost absorption. Looking forward, we expect that the overall level of revenue and product mix will remain key determinants of our gross margin in any given quarter. For the first quarter, we project our GAAP operating expenses to be approximately $676 million. We anticipate our non-GAAP operating expenses to be approximately $455 million. This forecast includes a step-up from the prior quarter due to typical seasonality in payroll taxes and employee salary merit increases. For the first quarter, we expect other income and expense, including interest on our debt, to be approximately $48 million. We expect our non-GAAP tax rate to be 7% for the first quarter. We expect our basic weighted average shares outstanding to be $866 million, and our diluted weighted average shares outstanding to be $875 million. We anticipate GAAP earnings per diluted share in the range of a loss of $0.18 to a loss of $0.28. We expect non-GAAP income per diluted share in the range of $0.18 to $0.28.
Operator, Operator
Our first question will come from Ross Seymore with Deutsche Bank.
Ross Seymore, Analyst
Clearly, it's kind of a tale of two cities, the data center and AI side is going to be really strong, the rest of it not so much. So why don't we just get the bad news out of the way first? The magnitude of the drops ex your data center are kind of shocking. Can you just walk us through how much of that is something that is not going to come back, some of the stuff in 5G, et cetera, versus what do you view as just taking the cyclical medicine and then a snapback should it ensue soon thereafter?
Matt Murphy, CEO
Yes, definitely a tale of two cities. On the carrier enterprise and consumer side, the way to think about it is, sure, the drops are pretty steep in Q4 and Q1. We're definitely going through a cyclical downturn in the industry, you've been doing this a long time, so have I; we've seen that happen, and that's what we're going through. The way to think about both of those businesses is at their peak, together, they were about $2.5 billion in revenue, and that was during the pandemic and some of the supply chain issues that went on. You can now see we're shipping well below that. And so I said in my prepared remarks, both of these are very solid businesses for Marvell. Both will recover to greater than $1 billion in revenue over time; the question is when. And remember as well, within these businesses, these are very long product life cycles, typically like seven years in production. These are designs. In some cases, we won 3 or 4, 5 years ago; some of them we just won a year or two ago. So we feel good about these businesses, but they are going through some demand softness and some inventory correction, but we expect to see that behind us. And to be very specific, none of this is due to any business that's not coming back or share loss. In fact, when things recover back to a normalized run rate, we do have Marvell product growth drivers in both of those segments, carrier and enterprise, to drive growth going forward. So we're just managing through it in the short term.
Operator, Operator
The next question will come from Matt Ramsey with TD Cowen.
Matt Ramsay, Analyst
I guess kind of a two-part question on the custom silicon business. The first part of it, Matt, you gave us some good color in your script as to how you see $200 million exiting the year and exceeding the $800 million for next. You kind of referenced the billion dollars at scale. And I just wonder if you could maybe square the rest of that; circle with that just for business you've won today for fiscal '26 or is there more to come? And I guess the second part of the question, my observation is you guys probably have, A, more shots on goal for custom silicon, and B, each one of the shots on goal could be potentially a very large or larger than expected piece of business given what's happening in the AI trend. But I guess my concern is as you evaluate those opportunities when those programs will be shipping 2, 3, 4 years in advance and trying to figure out how those programs would sort of align with the dominant player in the space's roadmap 2 or 3 years out and the pace of that innovation. Like how do you guys evaluate the potential for the size of the business when lots of shots on goal, but the probability of each of those shots ongoing being successful might be hard to gauge given how quickly the industry is moving. Hopefully, that makes sense. I'm just trying to get an idea of how you're scaling the business and taking on orders given all the volatility and rapid technology advancements in this space.
Matt Murphy, CEO
Okay. Let me just tackle them both. On the first one, yes, super excited about the ramp we're seeing now on the custom silicon programs we won a couple of years back. So we're shipping in Q1 on both programs. It's fantastic. We gave some color in the prepared remarks on the exit rate being $200 million plus. I don't want to cap that at this point, but certainly, that's sort of the way to think about it as the floor, and then we go from there. So that would imply, obviously, in fiscal '26, a much larger number than the $800 million we had called out. And again, I think it's still dynamic and still early but we're very excited from the bullish forecast we're seeing from our customers. So '26 looks very strong from the programs that we've won, and we should be way ahead of the plan that we had articulated a ways back. On the second question, yes, I don't think I've ever seen more design win activity in my career right now, in AI, custom silicon, networking, optics, you name it, this whole transformation that's going on in data center architectures is creating this massive opportunity for Marvell specifically on the custom portion of it. I mean, we're talking about a TAM creation that's going to be in the tens of billions of dollars range. And we'll talk more about that at our AI Day. But very significant sort of opportunity that's emerging in front of us. That being said, I view that as very complementary to the merchant leader in this space. I think both have a place, given the workloads that are required, given the scale of computing that's needed. And I don't think it's a zero-sum game. I've been saying this for a while. I think both segments, the merchant side as well as the custom side are going to grow to be very enormous TAMs. And so yes, we're right in the mix on all the major opportunities and programs. We have assembled an incredible team, an incredible portfolio of technology. We announced today our continued partnership with TSMC with a very full and robust roadmap of IP and technology to specifically address AI and accelerated computing applications for the next generation. So that's got a lot of excitement with the customers as well. So yes, overall, we're very bullish, and we think all the boats are going to lift on this one and certainly for Marvell creates a huge new TAM opportunity that we really didn't have just a few years ago.
Operator, Operator
The next question will come from Christopher Rolland with Susquehanna.
Christopher Rolland, Analyst
Matt, thank you for sharing the custom ASIC opportunity numbers again. This likely indicates you have a solid pipeline. I would like to explore this further. Can you discuss the opportunities regarding training versus inference, and how this compares to general compute? Additionally, do you have a unique advantage that is contributing to your success? Your competitor has leveraged their leading-edge SerDes IP to achieve wins. Is there something special that you have that is driving these successes?
Matt Murphy, CEO
Hopefully, some of the custom numbers were helpful. There was significant interest from investors last year and this year regarding the size of our market opportunity. Now that we have more visibility, it’s beneficial to express our views on that. Regarding your question about training versus inference and compute, I would categorize it broadly as creating a substantial total addressable market. I’m not prepared to break that down into specifics on this call, and I think it’s uncertain where that will be in three to five years in terms of exact proportions. However, I can say that our opportunities span all of these applications. In terms of our technology and competitiveness, it's quite impressive. When we secured our designs at 5-nanometers, it was our first time competing in this sector, particularly in the data center area. The origins of this business trace back to our acquisition of Avera in 2019. We shifted our roadmap during the closing process from 14-nanometer technology to 5-nanometer. Despite it being our first attempt, we were competitive with SerDes, packaging technology, interconnect, manufacturing scale, and strategic relationships throughout the supply chain. This capability enabled us to capture a significant share of the ASIC market, which is now materializing in fiscal years 2025 and 2026. We then progressed to our 3-nanometer platform, which is now a vital part of the opportunities we are pursuing today. We are in a significantly stronger position now compared to when we first worked with 5-nanometer technology due to the experience and readiness we have gained. We are fully committed to not just advancing nanometer technology, but also in areas like die-to-die interconnect, packaging, SerDes technology, optics, and chiplets, and figuring out how to integrate everything. Additionally, with Marvell's scale and our strategic partnerships, we are seen by our customers as a reliable, long-term partner who can help them achieve their AI silicon goals. That’s where we stand currently. We performed exceptionally well with 5-nanometers, we're set to excel with 3-nanometers, and we’re optimistic about the future.
Operator, Operator
The next question will come from Quinn Bolton with Needham & Company.
Quinn Bolton, Analyst
I'm going to follow-up on the cloud ASIC question as well. But I guess I'm coming at it from a margin perspective. Matt, you look at some of these programs on the cloud compute side to the extent that they integrate high-bandwidth memory. It certainly does nice things to the ASP, but there's also a pretty significant costs that you may have to pass through. And so I'm wondering, can you talk about the gross margins on some of these ASIC platforms, how does that compare to the corporate average? And as you ramp, especially if cloud ASICs get to be similar in size to the optics, which I know are good margins, how do you see the margin mix changing with this now very, very strong ramp of cloud ASIC programs? And then I've got a quick follow-up if I can sneak it in.
Matt Murphy, CEO
Sure. Yes. I mean on the way we've talked about that business, actually since Avera in 2019, is that it would always, due to the nature of custom business, be at a lower gross margin overall than the company average and the target. At the same time, because of the NRE and non-recurring engineering that we receive as part of the customer funding of these programs, the operating margins of this business tend to be very competitive and over time, in line with our company targets. When you start talking about these very, very large programs, you've got opportunity given the volume and scale on the OM line to really drive a lot of leverage in the model. But the GMs, just due to the nature of that business model, will always be a little bit lower than the total. I would say, though, at the same time, because the sort of follow-on sometimes is, we'll see if that happens and what happens to the overall company gross margin profile. And remember, we still have very strong gross margin merchant businesses in storage and in networking, automotive. Some of those were even under shipping demand right now. So as the year plays out, some of those businesses will come back, and that should be a little bit of a margin offset to some of the custom stuff. So we're still managing to our long-term target model. But yes, the cloud stuff, especially if these volumes certainly lower gross margin. You got a quick one for the second? We can do it; otherwise, we've got to get going.
Quinn Bolton, Analyst
Yes, quick one on the second. You mentioned sort of ramping AECs at multiple hyperscalers this year. Just wondering if you might be able to give us some sense. Is that meaningful? Any sense on the magnitude of that ramp?
Matt Murphy, CEO
Yes. I think what I'd say is it's absolutely a real market. We felt that way for some time. Even back when we acquired Inphi, that was sort of a shot on goal we thought was possible. I think the big news is that we're qualifying at multiple hyperscalers now. We're starting to go into production this year. We'll get a very nice, solid kind of more full year-ish ramp in fiscal '26, but it's definitely an incremental additive category to Marvell. And it really leverages our high-speed or high-performance DSP and PAM technology, which is really now inflecting in this wave of AC before it was based on the older modulation techniques which we didn't participate. So yes, really excited about it, but let it play out for a few more quarters.
Operator, Operator
Your next question will come from Harsh Kumar with Piper Sandler.
Harsh Kumar, Analyst
Matt. So AI is hot, there's no other way to put it. You are in kind of a pole position with your optical piece, Inphi piece. Can you give us a sense of the magnitude of the size of that business, either quarterly or last year? Anything you can provide us that helps us size it? And also kind of what kind of growth rate you expect to see from this business? And then, Matt, when you say the bottom is for some of the pieces like consumer and networking, the bottom will be this 1Q timeframe, what piece do you think will come off the bottom in 2Q? Or do you expect all the pieces to come off the bottom in 2Q?
Matt Murphy, CEO
Sure, I'll address the first question, and then Willem will handle the second one. To start with the first, we provided some substantial insights. The exit rate for AI in Q4 was significantly above $200 million, primarily driven by optics. When you consider this, you also need to factor in the revenue Marvell generates from standard cloud optics, which adds a healthy amount. Additionally, there are contributions from our carrier segment, specifically related to Inphi and its coherent DSPs along with some ASICs we have. Overall, the Inphi business has exceeded our initial expectations since the acquisition, proving to be an extremely valuable strategic asset. This includes the technology we've acquired, the talented team, and the synergy with our existing product lines. Overall, the outlook for this opportunity appears to be much larger than what we anticipated a few years ago. Willem, would you like to address the second part?
Willem Meintjes, CFO
Yes. Thanks, Matt. Yes, so we're really working with customers to focus on Q1 being the bottom, really confident that that's the bottom. And then we see growth resuming in the second half across enterprise networking, carrier, and consumer. So we're really just trying to make sure that we put this behind us really quickly and see growth in the second half.
Operator, Operator
The next question will come from Vivek Arya with Bank of America.
Vivek Arya, Analyst
Matt, one more on electro-optics. I think near term, it seems in Q1, there's probably a little bit of a breather after several strong quarters. But the more important question is that should we assume that your electro-optics related to AI should grow somewhat proportional to the growth in accelerators? So for example, if you look at accelerator growth this year, I've seen various numbers between 50% to 100%, kind of year-on-year growth. So should the AI part of your electro-optics grow kind of commensurate with that? Or is there something different? And then when does 1.6T start to play a role in your electro-optics business?
Matt Murphy, CEO
I think it really depends on how you define a breather. The electro-optics business has ramped up significantly from Q2 to Q3, and then from Q3 to Q4, and it remains strong in Q1. While it may seem like a breather in terms of sequential growth, we haven't actually seen any slowdown. As for your second question, regarding the 1.6T, that will ramp up later this year and everything has been progressing well with the qualifications and product introduction. We announced it at OFC last year, and there has been a lot of demand and interest. Now, about your second inquiry related to AI?
Vivek Arya, Analyst
Yes. So should your electro-optics grow kind of proportional to the growth in accelerator?
Matt Murphy, CEO
Yes, the accelerator is definitely important. We expect our business to grow year-over-year in line with the growth of accelerator units. We see a strong correlation between the two. There may be some timing challenges related to the supply chain, but that’s a straightforward way to understand it. Different models exist regarding the specific numbers, but we believe our optics business is highly linked to accelerator shipments.
Operator, Operator
The next question will come from Gary Mobley with Wells Fargo.
Gary Mobley, Analyst
As a way to think about the win or hit rate for your custom ASIC business when engaging with prospective customers in the future, how should we think about the increasing competitive environment from the Arm ecosystem, especially with the more versions than Neoverse, its total design ecosystem, its compute subsystems and as well as the ASIC design community? First of all, how do you feel about Arm essentially becoming more of a competitor? And then as well, how do you think about the way in which you'll compete against that and succeed?
Matt Murphy, CEO
Look, I think there's a lot of excitement in this market today because of the opportunity in front of us. I mean the way that we're working is there's a number of companies that are our partners, solution providers, et cetera that may or may not have their own offering. Whether it gets taken up or not is something we will have to see. Where we come together really is we pull it all together. We're able to actually deliver with Marvell intellectual property, design methodology and techniques, our IP, and our know-how pulled together with our customers RTL and IP and partners, the total solution, which means we actually lay the chip out. We package it up, we manufacture it in high volume. We QA it, we test it, we yield it, we drive reliability, we drive cost improvement, and we drive it through the cycle. And that is a very different business model than providing a piece that goes in. And so Arm has been our partner for a long time. We use them in our products kind of all over the place. And as they evolve and others evolve, we'll figure that out too. But in the end, it comes down to, we believe, for these large volume mission-critical big compute silicon chips, it's going to require somebody like Marvell with the manufacturing scale, the design techniques, and know-how to really execute it. We haven't seen any change there. And in fact, I think at the sort of 3-nanometer node we're at and beyond, it's going to be even more critical to have a partner who's got your back, who can help you get to production quickly and spin derivatives quickly, which is increasingly looking like what's going to be needed: more SKUs, faster beat rate in terms of when products are released, and then obviously ramping them very hard and quickly given the competitive environment.
Operator, Operator
The next question will come from Harlan Sur with JPMorgan.
Harlan Sur, Analyst
Matt, you mentioned the initial shipments of your AI ASIC program. Can you just clarify because I know last you updated us, these programs were in qualification. So have you guys passed qual on both these programs? And is it sort of the initial start of the full production ramp? Or maybe you're still in qual, but you've got enough line of sight to passing the qual just given you're at the tail end of this process? And maybe more importantly, have you guys secured the follow-on AI programs for these two initial projects?
Matt Murphy, CEO
So yes, we are in the initial start of the production ramp on both products. And then as far as the follow-on, like I said, the opportunity funnel we see across all of the various opportunities right now is significant, and we're involved in, we believe, in each of them. So yes, and there'll be more to come sort of at our AI Day. But I would just say our 3-nanometer funnel and our 3-nanometer hit rate and design win rate is very encouraging and it really gives us this tremendous confidence in where this business is headed. It also has a side benefit of driving this advanced technology for the custom ASIC side, as it pulls along the technology development that benefits all the other businesses in Marvell, like our high-performance switching, our DSP for optics, et cetera. So there's actually kind of a virtuous cycle happening where being at that bleeding edge is now we're able to show our other solutions that interoperate with this custom silicon, really a best-in-class roadmap there. But more on the overall strategy and opportunity size at the AI though.
Operator, Operator
Your next question will come from Tore Svanberg with Stifel.
Tore Svanberg, Analyst
Matt, I had a question on the 800-gig upgrade cycle. So you mentioned that ramping, but you also mentioned 1.6 Terabit ramping later this year. My understanding is that we're still in the early stages of the 800-gig. And I'm just wondering if you expect a more broad-based 800-gig upgrade cycle before 1.6T really starts to take off? Or will we potentially see kind of both ramping at the same time?
Matt Murphy, CEO
Yes, I think there's three different pieces we should talk about. So the first is 1.6T. I would just call that early shipments, first volumes type of thing in the second half, end of the year, and then more next year. And that's really, I'd call the early adoption of that technology. But it's very promising because that's obviously what's going to follow. The 800 gig for AI has still got tons of legs, right? And it's going to be very, very strong this year, very, very strong next year. And then there'll be some transition over time. And then I think what you're also indicating is, at some point, you're going to see an 800-gig transition on traditional cloud infrastructure from either 200-gig or 400-gig to 800-gig and that is also true in something we're planning for. So I think they're going to end up happening kind of all in parallel a little bit, right? You're going to have leading-edge people adopting 1.6T. You're going to have the bulk of the volume on AI still driving 800-gig for the next few years; and then you're going to have a transition on the traditional PAM side as well that's going to upgrade. So I think they're all going to be in production in parallel for different reasons, which is great because it just ends up being multiple customers, multiple SKUs, and kind of a rich portfolio of solutions that we’re offering so that people can optimize their systems for the best TCO and the best performance that they need.
Operator, Operator
We will take our last question today from Srini Pajjuri with Raymond James.
Srini Pajjuri, Analyst
I want to follow up on the earlier question regarding the custom silicon opportunity moving forward. There are four major cloud customers who are hyperscale, and from what we have observed, most of them appear to be investing in one or two custom silicon chips. This suggests that the number of opportunities may be limited, although these are significant opportunities. I'm interested in your perspective on how this might evolve. Do you think these customers will continue with just one or two programs, which represent multibillion-dollar opportunities, or do you anticipate seeing a broader range of opportunities as AI expands? I would like to hear your thoughts on this.
Matt Murphy, CEO
A few years ago, when AI was starting to emerge, there was typically one chip and one program to focus on. That process would take several years before gearing up for the next iteration, following a serial approach, which was acceptable at the time. However, what we’re observing now is that multiple programs are being developed across all major customers. I prefer not to provide specifics regarding my customers, but we are hearing a strong need for each of them to develop highly differentiated and unique solutions, managing more SKUs at a much faster pace than we've seen previously. This shift is creating a significant new opportunity in custom silicon that didn’t exist before, both in terms of the total available market and the volume of chips required, along with the urgency for speed in delivery. If we consider the current 3-nanometer cycle, it represents a much larger opportunity compared to the 5-nanometer cycle, even when adjusting for the increase in 5-nanometer total addressable markets due to market growth. The trends suggest we will see continued expansion in this area over the next few years. However, it's important to note that high-volume data center applications will likely continue to be concentrated among a few key partners, which is just how the industry operates. It’s an exciting time with a lot of activity in this space.
Operator, Operator
Yes, sir. Go ahead, sir, take the floor.
Matt Murphy, CEO
Thank you for joining the call today. We are currently experiencing a significant transformation in the semiconductor industry, particularly with the overhaul of data centers and their architectures. At Marvell, we have dedicated the past four to five years to evolving into a data center-focused company. As indicated by our recent results, the data center has become our largest end market, where we are concentrating most of our research and development efforts, positioning ourselves for significant success. We strategically decided several years ago to engage strongly with accelerated computing and infrastructure, with AI as a key component. This opportunity has grown much larger than we anticipated a few years back, and we are committed to investing in it. We expect considerable growth in our data center business over the coming year and are enthusiastic about this potential. Some programs are just beginning, while others are accelerating in the latter half of the year, providing a solid foundation for our fiscal '26. Reflecting on the insights gained through customer interactions over the past few months, particularly from a recent discussion at JPM, there has been a lot of excitement regarding our production plans. Our confidence in our growth and the strength of these initiatives has only intensified since then. We are thrilled about our current position and appreciate everyone's support and interest in Marvell. We look forward to seeing everyone at AI Day. The team is working diligently, and they will exceed expectations. Thank you, and see you soon.
Operator, Operator
Thank you for attending today's presentation. You may now disconnect.