Okay, I believe we are live with the team from Marvell Semiconductor and CEO Matt Murphy. Matt, thank you so much for joining us. Thank you, Willem. Thank you, Ashish, as well, and thank you for everybody joining us today. Just some housekeeping. I believe if you're a viewer today, you're supposed to have a question box where you can submit an email to me if you have one for a question. You can also just submit directly to my email Christopher.Rolland, R-O-L-L-A-N-D, @sig.com. With that, let's get into it. Matt, thank you so much for joining us. Thank you, Ashish. Thank you, Willem. Some fireworks last night.
I guess the way that I kinda describe it is, you know, since you took over, Matt, it's been just a fantastic number of years. Of late, I think Marvell has hit a soft spot here. You know, maybe you can get into, you know, what's been described as kind of a perfect storm of some things coming together, some headwinds coming together against your business. Perhaps you can give us some reassurances as to the future and why it is indeed bright.
Sure. Yeah. Thanks, Chris, and I appreciate the opportunity to be here with, you and all the participants. Yeah, the first thing I'd say is, it has been a great run. I think, if I reflect back, you know, even on our results from last year, you know, the company grew 33%. You know, we had our cloud business grow, like, 50%. You know, 5G hit $600 million. That was a bogey we set a few years ago, and it's gonna grow from here. Automotive hit the $200 million milestone. The enterprise had a phenomenal year. Overall, and, you know, on track with gross margins, and we're thoughtful about expenses.
If I reflect back to where we were at the end of 2021, thinking about 2022, that's basically what we signaled, right? There was a lot of enthusiasm for the stock and the company at that time. I think overall it was a good year. Certainly to your point, we've hit a patch of softness at the end of the year, and now in the first quarter here, where, for a variety of reasons, we could dig into it, you know, demand has slowed or decelerated. Customer inventory now needs to get burned down. Then on top of that, as we look to our first quarter, there's a, use your word, perfect storm of mix issues that we're encountering relative to gross margin. Driving gross margin to a level we haven't seen in some time.
I'm happy to dig into that, Chris. I think I'll pause in a second. You can tell me where you wanna go, but probably a walk around that would be helpful, which I can do by end market. I would just say, while we're guiding 60% for Q1, that's really due to an extremely abnormal mix, and we anticipate that as this fiscal year rolls through, the mix will normalize as inventory gets digested. By the end of the year, we should be back to normal Marvell in terms of mix, and that means gross margins will float up from here.
Yeah. Yeah. Makes sense. Data center, it sounds like storage is the main culprit here. You know, maybe talk about hard disk versus SSD. Is it the same thing we're seeing there? You know, we can get into the other parts of it too. I wanna drill down there, but maybe first address storage. I think a lot of us were surprised there was another leg down. I thought we were kinda done.
Yeah.
Yeah, anything else you can describe here, normalized run rates, where you think we get back to, you know, long term, you know, where we could go? Anything around that would be great.
Sure. Yeah. I think if you think about the data center end market, you know, back in December, we were hopeful that, you know, our fourth quarter would have been the low point. It's come down since, and really I think the dynamic, and this is kinda coupled with the gross margin story, is hard drives, Fibre Channel, and actually in the data center space as well, there's a more modest correction going on on, like, our PAM4 optics, and that's really due to just the, kinda the supply chain that we feed into. Those are all generally, you know, gross margin accretive type product lines, you know, higher than the company average, and those are down in a more pronounced manner.
If you just look at data center, which has overall been a better than company average gross margin segment for us, that's down, like, mid-teens, you know, sequentially. You just sorta do the mental math, and that's a big mover on the mix. That's the biggest issue within data center.
Yeah.
Maybe, Chris, to your other part of your question, like how do we see it? You know, I mean, while, yeah, we guided it down, but on the positive side, we certainly have seen in storage, right, some of our customers start to have burned through their finished goods inventory. They're now working on working process. They're gonna start to rebuy soon. We're shipping massively below the end market, right? While it is a little painful right now, you know, the good news is, you know, you start to recover and especially in the back half, right? Our recovery is really not assuming some kind of end market recovery. It's just simply the fact that, hey, we just start shipping in line kind of with the end market.
Just to remind people, just, you know, to part of the question, kind of what's the normal run rate? I think if you remember before kind of the supply crunch and kind of the boom and bust cycle we're in right now, storage was about $1.4 billion for us kind of annually, right? 60% of that is really in data center. You do the math, that's like something north of $200 billion a quarter. That to us is kind of the typical run rate. We're obviously well below that, right? If you look at our Q1 guide, you're off by more than half, essentially, right? That's part of why the gross margin is where it is.
As we get to the end of the year, whether it's exactly in Q4 plus or minus, but that is the normalized run rate. Again, fundamentally, if you look at what we supply, it's nearline HDD, it's high-performance SSDs, right, and Fibre Channel. These are all fundamentally, you know, have their own secular growth drivers. We're very confident that that is the right level. It's just a question of which quarter do you actually get there.
Yeah. Understood. You know, Matt, you'd mentioned Inphi. I mean, we had a great read-through, we believe. Lumentum talked about this, particularly one customer, your, I believe, lead customer there for PAM, PAM4. Originally it was gonna be a one to two quarter correction for them. It was described as inventory that was being digested. I believe it's inventory digested in front of the movement from 400 to 800. This was later revised this quarter by them for the full year. I guess, you know, are you seeing the same thing, and is that how you would describe the dynamic?
We, again, we're in a little different position in that we obviously feed into the broader optical module supply chain. Across a wide variety of partners for various speeds and feeds for different applications. I think that this is what we started signaling last quarter, is that, you know, while the primary sort of driver in Q4 was gonna be storage, there was some weakness that was forming elsewhere, and then we said, basically, that's continuing in Q1. While storage is down, the rest of the portfolio needs to go through that level of digestion. I think we're broad enough. What I would just say, without talking about one specific, you know, customer partner, is that it's at different phases depending on which module company and then who do they sell to.
I wouldn't put a broad brush on Marvell and say it's gonna take X quarters kind of overall. I think some are correcting now, some are correcting next quarter, some are. It's kind of rolling through the year, Chris. I think it's still dynamic. I mean, to your point, you know, what did companies think last quarter versus what are they saying now? The tone has changed, right? The information has changed. I don't think it's quite completely settled out yet as to how this is exactly going to recover. I think given our breadth and also the fact that we have so many shots on goal with new ramps coming within the optics portfolio, you know, new technologies kicking in. I mean, we just announced our 1.6 Tb solution. We'll be showing it at OFC next week.
It's a real product. People are building modules with it. You know, we anticipate that's gonna go into production next year. While there's some short-term thrash because we supply pretty broadly to everybody, you know, we think we'll get through it in a reasonable timeframe. Of course, you got growth underneath, so that helps a lot too, right? It was not a stagnant market that now needs to correct. It's been growing. I don't have an exact like it's this quarter because it's too broad, but it's already underway.
Yeah. Maybe just to add, just to want to double-click a little bit deeper, right? I think on the 200 and 400 Gb, which you would consider as kind of run rate businesses, which are typically used for switch interconnects, is I think where this inventory correction commentary applies appropriately across multiple customers. Also remember, we started to drive a very, we talked about it, I think, for the first time this quarter, is, we started ramping 800 Gb last year, and this business just took off, right? That's really being used in AI clusters. Their net investment keeps growing. I mean, you know, I don't have to say the word ChatGPT, but I think generative AI, to be more generic, I think it's certainly driving that. I think that's a different dynamic, right?
Which really there's no inventory there, right? That's more of a new growth driver. Again, within the Inphi portfolio, the other thing we have within our data center business is our data center interconnects, infra data center connectivity, where we just launched our first kind of industry standard 400ZR product last year, that did extremely well, and we see that growth continue, right? I think I don't think it's I wouldn't paint like all of Inphi, right, in the data center under the same brush. I think 200, 400 Gb certainly has some inventory correction, the other parts of the business demand looks very good.
As I look forward into next year, I even say an even bigger acceleration, especially as Matt mentioned with 1.6T, there's huge interest in AI, right? While there's certainly be the switching portion following on, I think the AI clusters can adopt these very, very quickly.
I guess finishing up on Inphi, is this all inventory or is this some of this the movement between 400 and 800? Then is there also the same dynamics taking place in ZR and ZR+?
800 Gb, it has really not found much, you know, usage on the switching side because I think as we all know, right, most of the market is 12.8. They really wanna make a jump to 51.2. You get the 4x, right? Bandwidth, it's kinda worth putting all that effort. The 25.6 would have driven 800, but I think most of 800 has really been AI, right? That's why I don't, I don't think it's connected necessarily. I think on the 100 DCI going to 400ZR, I think that's now 400ZR is an industry standard, multiple customers. I think that's on its own kind of ramp, right? That's not there's no inventory.
In fact, I think we were short on supply across the industry till not that far ago, right? I think, I think it's a different dynamic.
Yeah, the highest level, they're not transition issues. It's just more growth was here, now growth is still, you know, cloud's still growing, but it's not growing at the same rate, and that delta between the two is the, the digestion. It's not like a product transition per se that we're dealing with. The 12.8T platforms with 400 Gb optics with them have a lot of legs.
Right.
'Cause in a lot of cases, people are just gonna wait to leap all the way to 51.2T.
Yeah. Innovium, I would have expected a strong attach with PAM4 and Innovium, but they can move on different cycles and so forth. Maybe describe Innovium. Is it different than what we're seeing in PAM4, or are there any inventory issues there or buy ahead or lower demand or anything, or you're still fairly bullish there? I think you guys gave a revenue number at some point. I wanna say $150 or $200 million. Any updates there?
Yeah, no, it's gone well. It took us probably a little longer than expected to get it to ramp. We bought it in the middle of the supply crisis and, you know. Part of the value we brought was actually being able to get the capacity that was needed, but that took a little time to bring online. We've ramped it, and it's in production now, and we had set a target for, you know, data center switching for Marvell, you know, to be in that range. That's the range that we achieved, and that's where we are. That's gonna grow from here.
That one's not really got any inventory issues. Part of that is, we never could quite supply enough early on, quite frankly. We were throttled. I think that one actually seems okay. I think had we been able to supply it, we may have, we may have actually overshot. In some way, it was probably okay that the supply came on when it did. Generally, that's been tracking. Then we announced, you know, yesterday, along with our 1.6T optics, also the whole platform, which includes our next generation product, which has a combination of Innovium and Marvell IP in it. That's our 51.2T solution.
You know, it's kind of a nice combination of actually Inphi plus Marvell plus Innovium and all the IPs coming together in one platform solution.
Great.
Yeah, Chris, we expect data center switching to grow nicely this year. Just to kind of, not to put too fine a point on it, but we absolutely expect our data center switching business to grow again this year.
Excellent. Good. First question from the audience. Again, if anybody has any, probably the easiest way, would just be email me directly at Christopher.Rolland, R-O-L-L-A-N-D, @sig.com. First question is, how is your visibility into hyperscale customer inventory? Obviously, it must be fairly low as it was only at CES that the COO said data center have been de-risked.
Yeah. It's still, you know, obviously became, you know, I think those companies, especially I would say, we've learned a lot in the last eight weeks or so in terms of what their plans are. I think it's been a very dynamic environment. If you look from kind of December when we were guiding what transpired through December and January. I mean, back when we guided in December, there was a little bit of a shock that somebody was, like us, was even gonna call out the fact that cloud had some weakness. You know, we were actually early in that, and our view certainly was that that was largely de-risked. Those companies are still going through their own rationalization, if you will.
I think we're feeling better about, you know, certainly based on our Q1 guide, you know, our hope is that this allows the right level of inventory rebalancing to start and then start working up from here, right? We can actually grow overall as a company in the second quarter. Yeah, it's the visibility in some ways is challenging because remember, some of that business in data center, like in the HDD area, we don't sell directly to them, right? We sell to a hard drive company, and then they sell it to a system integrator, and then it goes into... The same on the optics side. We try our best to triangulate all the inventory in the supply chain, but when you get into storage and optics, there's at least one, if not two, steps in between.
That's just become more challenging to figure out. It's just more opaque. It always has been. Anytime you sell into a guy who sells into a guy who sells into your customer, you've got the traditional bullwhip effect that you deal with. We're, you know, our best estimate right now in Q1 is that that, you know, takes all that into account.
Got it. Data center, $400 million this year, $800 million. These are custom ASICs, primarily for hyperscalers. You know, you now believe it's probably gonna be $200 million this year because we have some push-outs. Maybe just put people at ease that this is indeed a push-out, that these designs are coming. As much as you can, I'd love to talk about them. I know you can't get into huge detail, but I'd love to understand, you know, is this one or two customers? Is this 10? Is this primarily around just networking, or is it broad? Just compute or is it broad? Anything you can talk about, I think, really help investors.
Gotcha. Yeah, on the first one, you know, it is a push. You know, Chris, when we look back, historically when we've guided investors on our new growth areas, we've sized them, but we haven't put a timeframe on it, and that's allowed investors to understand, like, our automotive story and our 5G story. I think if you even look at last year, you know, we had talked about $100 million in automotive at some point. It did $200 million. We had talked about $600 million. That's kind of worked for us.
I think the difference on the cloud-optimized is we committed to two timeframes, which in retrospect, sure, wish I hadn't done that because I still think it's gonna go 400, 800, but it's not gonna be in that exact 4-quarter period that we had anticipated a couple years back when we sized it. It has moved. It has moved to the right. If to the extent any of the programs were to be canceled or rescheduled, then we would talk about that as being part of the issue. Right now that's not the case, and even in a kind of reframing in all these companies' own budgets and their own deployment plans, you know, we've figured out where we fit in, and these programs continue to have high priority.
We have parts in the fab, stuff that's taping out. There's a lot of pressure to get these chips executed, and that's kind of a way that we also look at, you know, what's the R&D intensity that we're applying to it and put upon us. So those are intact. Then, again, it's a little, you know, because of the sensitive nature of these, you're right, it's a little bit like how do you talk about them? There's, first of all, there's only really four major hyperscalers, plus there's the guys in China and then the tier twos. This is really, these are really centered around more of the large companies in the space. That's where our design wins are.
It's broad within that segment. Then the applications range from, you know, compute-based applications leveraging our Arm expertise, some for offload or for acceleration. Then there's products that are levered to AI ML. Then some of these are also in the networking area as well. My hope, Chris, is as we get closer and these things start to come to fruition and to the extent we could even get some public, you know, announcements around these, that would really help, I think. We're early enough on, and the sensitivity level's high enough that I think what I've given you is probably a hopefully a helpful mosaic for now. Then, you know, as we get closer, we can try to hone in.
How much visibility do you think your customers are gonna give us into these wins? Do you think we'll see half of it, or do you think it'll be all opaque, or you can't tell?
Don't know, but I'd assume it'd be opaque.
Okay. 'Cause these guys, they like to brag. They like to brag about their custom silicons.
It's their chip, you know? I mean, Look, we're very content to be a key partner. The guy behind the guy. Help them achieve their silicon ambitions, no problem with that. I know there's a desire for everyone to say, "Well, how's Marvell exactly connected?" I get that. Love it if it could just be all. The reality is we're very blessed to have these relationships, these trusted relationships with these companies, and I'm more than happy at the end of the day for it really to shine for them. As long as they're delivering what we thought they were gonna deliver in terms of the economics for Marvell and the long-term growth, we're thrilled with it, right?
Eventually if it's in the P&L and you guys are seeing it, you're gonna be perfectly happy too, no matter what the little, what the thing is that we built.
Perfect. another one for the audience. Broadcom's infrastructure and storage grew 31% year-over-year in January and 20% into April. why are you undergrowing in this segment, year-over-year?
Yeah, Chris, I think, you know, the interesting thing about this cycle we're in, first of all, it's a very long cycle, and if you I'm just taking it back to 30,000 feet relative rather than getting into this company versus this company, 'cause every company's got its own moving pieces and composition, and you gotta look at also people's results over a longer period of time as well. If you step back, what's surprised me is that normally when you have these episodic, inventory, you know, supply-demand imbalances followed by inventory correction, typically they're fairly bunched together. You know, companies are within 1 or 2 quarters of each other. Everybody kinda got over their skis, and it's a little bit agnostic company or market, pretty much everybody feels it unless they've got some completely, you know, unique growth driver underneath that they power through.
This is a case where large semiconductor companies as early as like last June started talking about, you know, as interest rates increased and economy started to slow down. You've got this long period of time where it started with consumer and it's rolling through. I think my view right now is every company's on their own journey, and every company's got their own way that they're managing it. All I can comment it right now is on Marvell, and I would just say that what we're seeing and saying is highly correlated with what the end customer and end market is saying and doing.
To the extent people are not doing that, I think you need to just ask them because clearly they've got their own set of initiatives and things that they're doing to successfully manage their companies. It's hard for me to comment, and I haven't studied all of it, and I don't, I don't know that specific comparison, but.
Fair enough. This is probably where we can talk a little bit about gross margins as well. The next question is, pure play ASIC shops typically have 35% gross margins. Is that what we're thinking about for your business? I would imagine not. You have said, that that was that that was lower. How does the margin profile change over the life of a project as we go from NRE to first production to high volume?
Yeah, I think at the highest level, maybe a benchmark we've already given is, you know, when we, when we first really got into this business, we purchased Avera, you know, from GlobalFoundries in 2019, which was, you know, the original IBM custom silicon team. We've been very happy with that purchase. The team's done a great job. We said at the time, you know, that business was around 50% gross margins, you know. The NRE component of the projects was such that, you know, we believed and it, and it has delivered, you know, very competitive operating margins over time because we get an NRE offset to the R&D, which delivers very solid profitability.
That business, by the way, has grown probably as fast or faster than overall Marvell, okay, since we bought it. It's been in the numbers, Chris. It's been in the P&L, it's been flowing through. Again, even if you assume it grew a little faster than the overall, you'd say, "Well, how did you go from where you were now to?" Well, that's because we have other higher margin accretive products in the diversified portfolio we have that's allowed us to manage this. I think it's been a great blend to date. We've always said we're gonna have a diversified mix of product lines with different structures, but it's our job to manage the mix such that it delivers to the model. You know, so it's not 35 as an example.
It's probably, you know, it's five handle type of stuff versus six handle would be the company average. Then some of the higher gross margin things we do that are very differentiated, that are merchant, that are very unique can be, you know, seventy plus. It is a range in Marvell. It's a portfolio, and I think today we've managed it very efficiently. Q1, obviously the volatility was such that we couldn't, we, you know, highest gross margin product lines are down, the lowest ones are growing like crazy, and that's the result, and it is what it is. It's not due to some other extraneous event like we didn't cut the price or there's not a weird, you know, input cost problem or, you know, a lot of that stuff is getting back under control.
It's, but the mix within each of these segments has moved around quite a bit.
You know, we haven't talked a lot about 5G. 5G has been aging, there may be some specific drivers for you guys that are interesting, like for example, India and Samsung's presence there, also Nokia's presence there. Can you help us balance all of this and understand what growth might look like for 5G moving forward?
We had very strong growth last year and, you know, we sort of ticked off the $600 million bogey, which was great. We've guided up our Q1, you know, which is, which is positive to see. India certainly is very helpful in this regard. It's a whole big geography that's finally kind of got legs. Other markets have gone through their deployments and still are. This was always one that was in front of us. I would also say though, Chris, that this carrier business always, no matter which company you're talking about, semiconductor company, it's always a little lumpy. We are seeing a pretty big, you know, growth, but it's growth, you know. It's kinda stepping up our run rate in 5G a little bit more.
5G overall, even if it's a strong first half, if you think about it's still gonna be up year-over-year again over last year. I think that's been a positive story, and it's because we have a diversified set of customers and we're pretty much in all the geographies except China, and those are the ones that are kind of still in front of us. Then we have some content gain rolling through and some other things going on. It's, you know, despite the lumpiness of carrier, it's been a pretty good opportunity for us.
Yeah. There is a question here around that as well. you know, as you're guiding for growth and others like ADI have seen a downtick post CES, is the delta there around content growth?
Yeah. I think the way to think about that, again, without comparing ourselves to any specific company, I'd say that a lot of the large semi companies that supply into carrier and into the wireless segment typically had pretty decent positions in 4G. You know, they had a good content in LTE systems, they sort of leveraged it to 5G, they probably got some content gain. You know, it's been a good story for them. You know, we're kind of 5G only almost, if you will. We're in the segment of the CapEx spend that's still actually got momentum and legs. If you've got some legacy business that's probably doing some offsetting, it would be my guess.
You know, over time, when 5G becomes fully deployed in a few years, then, you know, this business will normalize, right? That's still got growth in front of us right now. That's what I think, that we don't have a lot of the legacy carrier because we really didn't have a big content in 4G. That'd be my guess.
Next question is Samsung was the sole supplier at Jio. Now they're sharing the business with Nokia, Ericsson. Does that impact blended dollars per unit? Does just, you know, more broadly, does that affect, you know, your outlook?
No, I think we're okay just based on that particular system configuration and based on the content we have with all the players, which varies, but there's still content kind of, you know, broadly in India. When we look at that mix of probably who's gonna get what, and you can see it already in our, in our guide for next quarter, you know, we're forecasting the business to grow. I wouldn't, I wouldn't point that out. I think to be specific for that investor question, I wouldn't, I wouldn't worry so much about, hey, did Samsung get more? Did they get less? I think we've got enough content now broadly that some points of market share move wouldn't be what would drive volatility in 5G this year.
I think what's gonna drive volatility is it's just lumpy. Always has been, always will be, we know that, right? Which is why, you know, we went to the cloud, we tried to build that business, right? The whole company goal has been over time to diversify among end markets and smooth out to the extent we can. Now, Q1, we're not smoothing out. Q1, we're going through inventory correction. In general, that's been the business model, right, is get attached to secular growth drivers, drive enough diversity in the portfolio to weather the storm. Yeah, 5G will be a little lumpy, we have other markets too.
Next question is, can you talk through gross margin by segment and various variance within each segment, including how does semi-custom gross margins compare to ASIC gross margin? Is semi-custom at corporate average, or 65% actually was the number they gave.
I think there's kind of two questions in there, right? One is kind of the what's going on with each of the segments. We talked about data center, right, and the dynamic there with our gross margin product lines being down more than the rest. I think if you go to enterprise, you know, the merchant part of the portfolio is going through its own inventory correction, channel inventory, customer inventory. That's on the merchant side. Those typically are higher gross margin. We have custom ASIC ramping within enterprise. Carrier, we talked a little bit about the 5G, which carries a lower GM. You know, on top of that, the wired the other half of the... Or not even half, but the other portion of the carrier business being wired, that actually does carry a higher gross margin.
It had a great year last year, but that also has its own lumpiness to it. That's one that's down. If you kind of even think of our three big segments, you know, within each of them, they have their own mix issue that's being driven down. Then on the sort of semi-custom question, I think it really depends on the job. You know, some of these are more, "Hey, basically, you're not contributing a whole lot of IP. We need really strong back-end and packaging services. We need this or..." Okay, fine. That's got one margin profile. If we're building the whole chip to spec, then that's another one. In general, high volume, very digitally intensive products are generally, you know, blended below the corporate average, right?
You would just expect that given the nature of the beast there. Again, you know, we offset that with, you know, higher margins in the physical layer, higher margins in storage, higher margins in automotive, higher margins in optics. That's the goal over time, Chris, is to continue to drive a blended portfolio that's more stable. We're obviously not doing that in Q1.
Yeah.
That's an aberration out of, I think, a long track record of being able to manage a diversified portfolio of gross margins and product lines.
Yeah. I think you guys said, you'd be returning to the low end of the range by the end of the year. In terms of gross margin, you know, what could help accelerate that? Could we be above or like when do we go back to that, you know, getting back into that range? Is it early next year? You know, what has to happen for that to move in?
I think step one is we just gotta get all the inventory that's in there normalized back to customer demand, right? That's sort of just like step one. I mean, it's obvious, but that needs to happen. And then on top of that, kind of get back to where we're shipping to consumption. We believe we're shipping below consumption now. If you think about exiting the year at kind of normal Marvell, you know, and maybe the GMs are creeping up, maybe they're in the range, maybe they're a little below the range, but they're heading in the right direction. Clearly, we're really looking out to our fiscal 2025, calendar 2024 kind of full year with all of that behind us and looking at the blended portfolio to drive back into the model.
There's a lot of wood to chop between here and there, so that's what we're gonna go through right now. That's the way we think about it, Chris. I think, by the way, any goodness next year as we get, you know, growth drivers can get higher revenue, that'll also will help GMs. Just like we're taking a little bit of hit on overhead absorption as revenues have fallen, you will get that benefit. Plus we're, like, laser focused on gross margin, cost reduction initiatives, you name it, right? We're going at this very hard so that the setup, at least heading into fiscal 25, is to, you know, is to deliver the performance that people expect of us.
First half of calendar 2023, our fiscal 2024, we gotta go deal with what we gotta go deal with. That's just the market. That's just what people want. We do think, you know, as we get to the end of the year, the goal is to get to normal Marvell, you know, by the fourth quarter.
Yeah. Okay, good. That's, that's good. In, in the few minutes we have left, there's a couple more questions, and then I may have one. Can you elaborate on the lags to the 400G cycle, and does that mean we should expect a smooth baton pass to 800 and 1.6?
Yeah. For the, for the inside, you know, inside data center business, Here's the way we think about it, and I'll have Ashish add it at the end here. The 400 cycle has been underway. It's been a strong ramper for us. It's going through its own little correction we talked about right now. As long as those 12.8T systems remain in the fleet and in production, that's just gonna keep going. We see there's probably some, you know, interim step that some of the customers that will adopt 25.6, but that is a little bit of a tweener, is what it sounds like right now, that really people wanna leap over and go to 51.2.
I think on the traditional sort of 400, 800 to 1.6T, I think it's gonna be more of a 400 to 1.6T jump on the sort of the switch attach. The 800, which been in volume production on and, you know, product's doing extremely well, that's almost been driven completely by AI, you know. That typically always leads it, right? That's kind of on its own track. I think AI will be the main driver of 800 Gb, and then that will convert to 1.6T. I think 1.6T is probably where it all comes together, would be my view. That architecture we've developed, we believe can scale pluggables to 3.2, which would be even beyond that.
That's kinda how we think about it. It's, it's dynamic. We'll see who ramps what. We're kinda well positioned across the whole product portfolio to supply 400, 800, 1.6 for the variety of applications that are out there.
Yeah. Maybe the only thing I'll add, Chris, is, there is still a set of customers which hasn't actually gone to PAM-based products, right? They're still using NRZ lower speed products. We do see them start to adopt, right? They'll probably start more with probably a 400 Gb product. That is still also in front of us, right? There's been, call it a few hyperscalers which have really converted to PAM in volume, and that's been kind of the run rate business. There is a set of, you know, customers which still haven't actually, just because of their internal architectures, right, they were able to live with NRZ for a little bit longer. You know, as time goes on, as each link goes up in speed, right, there's pressure.
I think you'll start to see some of those convert to 400 as well. That's also completely in front of us.
Great. Maybe just one more for the audience, and we'll wrap up. Hard disk, you've talked about $200 million a quarter as a normalized run rate. How confident are you in that $200 million moving forward?
Yeah. Just to clarify, I think that's really a total storage number, right? For data center, right? Our overall business is a lot bigger. Again, you go back to on a normalized basis, the last time we used to break storage, that was like a $1.4 billion business, right? Call it 60% was in data center. I think confidence is high because if you think about, you know, this is really supplying nearline HDDs, right? Which basically are the volume driver for cloud storage, right? That's gonna continue to grow. Once the inventory correction is done, right, that business basically goes back to where it used to be. Then we've actually got growing share, quite frankly, on the SSD controller side, right? For enterprise data centers, right?
We already have new design wins in PCIe Gen 5, Gen 6 in front of us, right? Finally, a small amount of the business is Fibre Channel, which, you know, that's again, generally been a fairly stable business, right? Yeah, there's gonna be some, you know, a seasonality in Q1, but that business grows back again. Overall, I think the comfort level is high. We're not gonna call a particular quarter, but we're shipping so far below kind of the end market, right? You started to see those customers start to bring down their inventory, especially on the finished goods side. That's gonna over time translate to buying more components.
Overall, I think, again, not calling a particular quarter, but yeah, I strongly believe we'll get overall storage, you know, not just data center, you know, back to kind of a typical run rate, and then you go back to slow growth from that base.
Great. Well, again, Matt, I think it's a soft patch for you guys, but if there's any leader out there that I have the confidence, that you'll get this thing back on track, it's you. I guess with that, if you just wanna take us home, that'd be great.
Yeah. Thanks, Chris. No, it's, appreciate the time today to be able to talk to everybody. This is one of those periods where it is softer, it is much more dynamic than what we've experienced the last few years. We're well equipped to manage through it, and we will manage through it. We're laser focused on our initiatives, right, that we can control. That's things like making sure capital allocation is spot on and we're continuing to invest in the right projects, right? Driving gross margin initiative improvements, really looking at project costs. Those are things we can do independent of kind of the market and the cycle and the inventory correction, right?
Making sure we have the right leaders in place in the right, in the right organizations to really drive the growth through, you know, coming out of the cycle. That's our goal. I mean, we were there in 2019. We had a reset at the beginning of the year. There was this industry downturn. We had to go kinda you know, reorient, that was a year where we doubled down on automotive, we doubled down on 5G, we got the ASIC business in place, you know, we're shot out of a cannon coming out of that cycle. We gotta go double down. We appreciate the support from our investors so far, we will continue to execute for you guys and come out of this even stronger.
We'll have to manage you in the short term, and we're gonna go do that.
Great. Thank you very much, Matt. Thank you, Ashish. Thank you, Willem. We greatly appreciate your time.
Thanks, guys.
Yes. Thanks, Chris.
Yes.
Thanks. Thanks, Chris.
Thank you.
Thanks, Chris.
Bye-bye.