Good morning. We're gonna get started. I'm Tim Arcuri. I'm a semiconductor analyst here at UBS, and very pleased to have Seagate for our next session, and we have Dave Mosley, who is the CEO. We'll just get started.
Great.
So Dave, maybe can you just sort of take us through the journey the past year, 18 months? You, you've, you've gone through a big downturn. You've taken a lot of capacity out of the system. You know, these products have an inherently long lead time. You know, you're now moving to a build to order. And so it begins to argue that the entire shape of the upturn could look a lot different, and that as painful as this was, that it could ultimately drive a lot of structural and lasting improvement in the industry and for Seagate.
Sure hope so. So, forward-looking statements on our website, www.seagate.com. Did I say that right, Shane? So, yeah. Okay, good. Yeah, thanks for having me, first of all. Beautiful day here in the desert. That's great. We've been through a lot. Not just Seagate, but the entire industry has. And I broaden that out to even the memory makers and things like that. It's been a tough downturn, and some of it is because in some spaces you have consumer products. The consumers are really disrupted. Enterprises are pulling back, and when enterprises pull back this time, they pulled back from the cloud, they pulled back on ad spend.
Various things happened through these periods, and a lot of major customers went from fairly consistent run rates that they'd had over a period of over a decade to virtually zero in the matter of, you know, just a few months, and stayed that way for a year. You know, you hear me complaining, I can't run factories like that. You know, we have to pay people, and we have to keep doing R&D, so that's what we've been going through. Why the change in behaviors or how do I think the behaviors will change? Well, there's just not enough money flowing through the system anymore to be building stuff speculatively.
Remember, the hard drive business itself grew as a function of client server, and I brag all the time, you know, we've done 3.5 billion hard drives in our, in our time. So we had huge factories, huge supply chains, and the cloud growth happened against the backdrop of that client server coming down. The cloud exabyte growth is still happening, so the heads and disks that fundamentally feed our products are, are still at a fairly healthy level even today, but they need to be even more, you know, in the future, and, and that's why we keep addressing areal density, which I'm sure you'll get into.
But, I think that the relative to the free cash flow of the industry and the way suppliers are treated and things like that, this has been such an impactful time with revenues coming down by a factor of two. We had to package the company very differently, be very aggressive on that. You know, I think we've, we've reached the footprint that we want. You know, I don't know exactly how the recovery's gonna happen because a lot of that'll be demand driven and depend on not only macro deals, but on the cloud customers starting to consume again. Their, luckily for us, their models are righted now, and you just heard one just get off the stage talking about the future of data, so I'm, I'm pretty encouraged by that.
But, you know, we'll continue to make sure we balance supply and demand properly and then give customers some visibility such that we're not gonna start anything speculatively. We'll give them visibility in the future of what we can do for them, and, you know, they'll pay along some different plan, I guess.
How has build to order gone so far with the customers? How have they embraced it? Do they realize, like, do they see the realities of your business? I mean, these are, there's a tremendously large amount of market cap. There are some very, very large companies that are dependent upon ultimately 2, you know, 2.5 pretty, you know, in the grand scheme of things, companies that are not that big, so.
There are different camps, for sure, of customers. There are some customers who, their supply chain people will look and say, "This is consistent with what I'm hearing from the memory makers," or, consistent with other spaces that have gone through these kinds of things before on the parts that we build, our next nearest neighbors, if you will, in the procurement ecosystem. So they'll say, "Yeah, I understand. I need to give you visibility." The total volumes have come way down, so what we're showing as far as these build-to-order plans are much smaller volumes than what the industry was absorbing two or three years ago. But, you know, against that, you give somebody predictable economics and a predictable TCO proposition, they're happy with you.
There are other people who say, "No, I'll keep trying to play the end of quarter game," and, you know, I think those are the ones who, especially with the long lead times now, those are the ones who are gonna be at some point, they're gonna run out of parts. And, you know, I do think that there's a refresh coming, not just because the cloud and, you know, some of the other mass capacity markets have been on pause for a while because of macro conditions and things like that, but there's also a refresh coming because a lot of the gear that's running is that much older.
There's benefits to power savings and space savings by modernizing your fleet and, you know, and against that demand upswing with less supply that the industry has, you know, I think that's what people are starting to look at.
Mm-hmm
... and be really serious about. Prices are also going up-
Mm-hmm
... so in some cases, they want that economic predictability, and that's, you know... I'm fairly happy with the progress on build to order so far.
You are. So how far away do you think we are from the cloud fully digesting all of the capacity? And, you know, there's been some white box compute displacement because of AI, so that's a factor in this as well. But do you see signs where your bookings are improving, and you see signs where you'd say, "Okay, well, you know, maybe into Q1, you know, cloud will still be weak, but, you know, sometime during the first half of the year, things should-"... beginning to get better?
Yes, I think we're because of the build-to-order plans, I think we're seeing a lot more predictability, which is what we want to be able to feed our supply chain, than we were seeing six months ago or a year ago. You know, the cloud builders are no longer in this mode of how much they can tighten their belt, but they're more in a mode of, "Okay, I'll give you predictable schedule for predictable economics." You know, I think people are understanding that. That's one of the reasons we feel more comfortable. Again, given the long lead times, you know, I think we need that predictability 'cause we're just not going to build a bunch of extra stuff for them.
And that'll play out, I think, over the next six months or nine months, as people will see what the true capacity of the industry is versus with the demand.
So I have to ask you the question about how things are going. You guided December flat sequentially. What are some of the puts and takes as you look into March? And also in December, is the quarter playing out as you thought it would?
Well, I think, you know, what we've been doing lately is making sure that we didn't have a lot of extra parts around, and really trying to adjust our build schedule down at the very end. This quarter, you know, we haven't really had to do that, so it's been a lot more predictable, and I'm fairly happy with that. You know, looking out over the long, long term, you know, there are signs that cloud service providers are starting to come up. The rest of the world, I think, is still a little uncertain, a little shaky. But I do think that, you know, relative to our build plans and our product sets, we're getting to a much more predictable period right now, and I'm very happy with that.
So you're not seeing those sorts of things, you know. At this point last quarter, things were still sort of under pressure, and they were still going against you. You're not actually seeing that now?
Right.
Things have gotten more predictable and more stable.
Yeah, and as you go through build to order or changing the economics or, you know, giving people a different, a different deal at the end of the quarter, they'll say, "I don't want that deal," and you'll have to back away from it and say, "No." And so, you know, we're kind of out of that period now. I think we're in, into a period now where everybody really understands the lead times and, you know, understands that they're not going to be able to come to us at the last of the quarter and say, you know, "I need an extra $20 million, $50 million of product." It just won't be sitting there, not in their configuration, for sure.
Great. So I get, so I get asked this question a lot, so I'll pass it on to you. You're, you're clearly, you know, below what any sort of a long-term trend is for mass capacity, obviously. I think you've said that mass capacity probably grows in the mid-20% over time, CAGR. But if I go back and look at the last five years, it's been roughly 15%, 16%, if I actually look at that period. So even 25% would be an acceleration over what the... and granted, it's fluctuated massively over the past five years, but it would be an acceleration over what we've seen the past five years. Now, some of that probably are some of the things that were talked about during the last session here on stage-
Right
... that make you more optimistic that the market could grow a bit faster going forward. So can you just talk about that 25 target, just in the context of, well, the math would say that that's much higher than it's grown in the past five years, and what are the drivers that could actually get you there?
A couple of different drivers. One, like I said, power and space in the data centers. You know, if you can replace a bunch of 8 TB or 4-TB drives with a 24-TB drive, you get space savings and a significant power savings. It probably self-funds, you know, some of that refresh, if you will. That has nothing to do with data growth. Data is still growing. And, you know, there's been some fairly creative things that cloud service providers have done over the last couple of years to try to contain the growth of data centers, but they're also building new data centers. So I do think that there's all kinds of evidence that the cloud is still in its fairly early innings, that people, IT departments that have moved into the cloud, their workloads are growing and growing.
And so the data is going to continue to grow. So that's the biggest driver that we see. And then the other thing on exabyte growth, which is a reality, is we're about to launch, you know, 3 TB per platter drives, and we can get to 4 TB per platter in two years. That's going to put a lot more exabytes out at scale and change those economics that I just talked about quite a bit for some people making longer term investments, seven years or 10 years, that they want to be running data centers. So, you know, that's why we think you'll see significant exabyte growth.
Got it. Let's, let's just, let's just talk for a moment about NAND. You've taken a different approach, and you didn't get into the NAND business as a hedge against your, against your drive business. Obviously, we've seen a ton of, you know, cannibalization, you know, happen in client. What are you seeing in cloud? I mean, you know, HAMR sort of takes you to an entirely new scaling paradigm that, that if there was ever going to be crossover, it would argue that it would push that out because, you know, you can scale costs as, as fast, if not faster, than what NAND can. But can you talk about that?
I can, sure. You know, I always start by saying NAND is a great technology, so I'm not going to say anything bad about the, the technology itself. Lately, from a business perspective, it's been really troubled waters for them. They've invested in too much capacity and, and taken a lot of losses because of it. And actually, you know, there are a lot of parallels with what's gone on in the HDD business, although I would, I would argue that the HDD business is a lot less impacted than even the memory business has been. The way I look going forward is ultimately, they'll get that supply and demand balance right, I think, and probably not go through the same kind of investment cycle. They will also roll off their depreciation at some point, so their profitability will improve.
From a hard drive perspective, you know, to your point, if we can keep pushing areal density just like we have for 50 years, you know, just then we will get to the point where we'll not only maintain the delta in cost, in effective cost. I think we said in our last earnings call, 5x, but it's probably a little bit more than that. We'll maintain that delta, and architecturally, there will be no big transitions. NAND will have a brilliant future. I think, you know, when you talk about AI or something, what, architecturally, the stuff that I see changing the most is memory interfaces of DRAM and NAND. But the stack itself inside data centers or even on-prem, with HDDs being bulk storage and then NAND being fast storage, I think that architectural, that architecture isn't going to shift very much.
Yeah, I mean, to some degree, it's kind of, you know, apples and oranges in terms of the use case. But even as much as NAND pricing came down, you didn't see any of the cloud guys taking workload that would've been, you know, going to an HDD that has migrated over to SSD, even, you know, given how much NAND pricing has come down?
No, and I, I think one other thing, in the architecture that I see that probably people don't appreciate the subtlety of, is there's tiers in everything. So there's not just an HDD tier, but there's different ways that the HDDs are used. There's tiers in NAND and probably tiers in DRAM. Look, these cloud service providers are fantastic storage architects, and they know exactly what they're doing, so they, there's not a price point where everything tips or anything like that.
I do see that, long term, there's, with some of the new applications that we're talking about, there's, a lot more possibility for a little colder storage, and then there's probably a lot more possibility, like I said, about memory interfaces, with really, really fast storage, that may be DRAM-based or maybe, you know, some new memory type. And that those, those are trends that these guys have definitely got their fingers on. But, you know, fundamentally, architecturally, between NAND and HDD, I don't see very much change.
So as a NAND asset potentially becomes spun back out on the market, sounds like you're, you're not of the mind that there's a lot of synergy between the drive business and the NAND business.
Yeah, I don't think so. I mean, we didn't believe that, you know-
Yeah
... seven years ago, and we don't believe it today.
Yeah.
You know, great technology is useful in different application spaces. There's a lot of applications that are growing. I really believe in the growth of the edge again. And I think data and data gravity at the edge will play a big role in the future. But I think NAND will benefit from that tremendously. I think, you know, HDD will have a place as well.
So let's actually talk about HAMR. We've made it, you know, halfway through the session, we haven't talked about HAMR yet. It, you know, you've given some unit metrics, and if you play out those metrics, and you assume that things maybe grow again into the back half of the year, you can get to a number. Now, this is my number, not yours, but you can get to a number where, you know, somewhere in the mid-teens of your exabytes next year could be HAMR, which is, you know, quite a bit. And you've proven that you can ramp, you know, new technologies, you've done that time and time again.
Can you talk just about HAMR, what the advantages are, you know, what the challenges you have in, you know, ramping yield and sort of how we could expect the customer adoption profile for HAMR to actually look like?
Hmm. It's a, it's a complicated discussion, I think, because a lot of people just say HAMR equals laser. You have a laser inside the drive on the recording head now. And that is true, but really, HAMR is more about the recording media, the disk, the surface of the disk. You're now opening up a new material set. And it's taken a long time to get that material set on all fronts to yield and reliability, that this is the same as the old material set. We know those old material sets very, very well. We're very confident, though, with the data that we see in the lab, and the lead times are so long that we are filling up the pipe right now because of that confidence.
There's also, with now solving some of those problems, there's the opportunity to kind of unshackle other things inside the phase space to grow areal density. That's what one of the reasons why we've talked about, we're not gonna stop at 30 TB or 3 TB per platter or something like that. We're gonna keep going to 4 TB per platter and even beyond that, because we see the capability for some of the other stuff to now be unleashed. But the industry has been through these kinds of transitions before. Once you learn how to do it, you wonder why it took you so long.
But, you know, I think the technology is getting really hard, and it's, you know, nanometer scale or less engineering that you're having to do on multiple fronts, whether it's a recording head or the disk itself. And, you know, I, you know, I'm really proud of the team, how predictable we've made it. It's taken a little longer than maybe I would have wanted, but, you know, screaming works sometimes, I guess.
It seems like the ramp is happening largely on the back of a single customer, at least initially, next year.
Well, I think we can qualify against many, many customers. Yeah, I don't think anybody should take away that it's one customer.
Okay. From the customer perspective, what's the decision tree to go to HAMR versus just, you know, stay on SMR and, you know, put in more disks to basically scale above 30?
Well, most customers don't use SMR, to be frank. So, there are a couple of customers who use SMR, and they do it very differently, one to another customer, so, you know. And we have developed SMR a long time, shipped our first cloud drive for SMR, that had SMR in it in 2014. So, you know, we'll tailor any drive the way a cloud person wants it, but most people still don't use SMR. But the HAMR drive itself, it's independent of SMR. If I shipped you a 30 TB, 32 TB drive. I'm expecting to be able to put that in distribution, and people use it as a CMR drive, just like they would one of our current products, and not even know what's inside the box.
So, you know, there's some places where there's subtleties about growing capacity inside that person's workload, and there's other places where people just say, "Great, a bigger drive. Let's go." And that's what we've engineered it for. There's no real changes in HAMR that require any customer, you know, nuance. They don't have to change code or anything else, as opposed to SMR, where you do have to, you know, really tailor the system if you want maximum performance out of it.
And can we talk about just the impact of HAMR on gross margin? There's this push-pull because, obviously, you know, customers want HAMR. I mean, the customers seem somewhat indifferent to what to, you know, whether it's CMR or HAMR. But from your perspective, you know, want to make sure that you ramp HAMR at a time when your yields are, you know, high enough, and it's not going to dilute your margin. So can you talk about that balancing act? And maybe I should ask Gianluca, he's here. But will it be accretive to gross margins in the entirety of 2024?
Is there, like, a yield point where you're sort of, I'm not saying you're metering the ramp, but you're trying to create that balance where you don't ramp, you know, HAMR too fast because your yields aren't high enough yet?
It's true. You know, I'm very happy with the yield progress, and I'm pretty hard to please on yields. I mean, like, the 20 TB drives that we had were fantastic yielding products, not just, you know, at the final assembly, but also upstream in wafer fab and media fab and things like that. So, you know, there's a very high bar that the HAMR drives have to get it through. But I'm very confident in what the team's put together, in getting there. And I actually think some of the, some of the concerns about the cost delta are really unwarranted. You know, from my perspective, you know, HAMR's going to continue to drive with virtually the same boxes as the PMR drives that we have, so there's really not much of an impact. We have to go work the yields.
And we do that really well, so I have confidence that we'll be able to do it. I think one other subtlety that, you know, I've heard people talking about is that with the new technologies that are coming, with all these new opportunities that we have, we can continue to push the TCO proposition better and better for customers. So at 4 TB a platter, it may be even more than that. And the more we can push those technologies, then the more we get into a balance of with SMR or something. We get into a balance of the same relatively the same cost, but a higher price point, relative price point, 'cause we're, you know, we're getting paid more for it, and that's what really buoys up the margins.
So to some degree, I mean, people just you know look at your margins versus your competitor's margins, and they say, "Well, your competitor has higher margins." Well, part of that is because you're investing in HAMR, so-
Oh, right now, there's a lot of stuff that's underutilization charges versus not. We are building a lot of HAMR stuff that's, you know, going to see the light of day a year from now because the lead times are really that long. So yeah, I don't really get... I don't look at today's optics, especially with so much excess-
Yeah
... supply that's still in the industry. Yeah, I think we need the industry, the suppliers and everything else. We need the industry to come back to a healthier demand-supply balance before we really start judging margin.
How far can you stretch HAMR? You know, once you convert to the HAMR roadmap, how far can you stretch it? And being first mover, do you think that it opens up a lot of opportunity? I know you, I know you're not managing your business for share, but if you have a more scalable roadmap, isn't the logical outcome that you're going to gain exabyte share?
We're going to try to apply our technology in every price band, if you will, cost point, you know, so on and so forth. So will there ever be a 10 TB drive with HAMR? You bet. So it's not just about the highest capacity point. We have to go do market development on some of that. Like, how would somebody use a 10 TB drive? I mean, today, some of the lowest capacity drives are. There's still a legacy PC market out there, but there's also things like NVRs, DVRs in smart city boxes, you know, surveillance boxes, hospitals, things like that. So, you know, we'll go do market development on that, and we'll go apply that technology all the way through the portfolio. So, you know, I'm confident five years from now, that we're going to be applying HAMR across the entire technology portfolio.
Do you think, is it not a logical outcome that you ought to be able to gain exabyte share over time?
It's not something we would go for, but right now I want to fill up factories. So, you know, given the long lead times, balancing act, but, you know, ultimately, I want to fill up the factories for sure. We want to get it, make sure we get our footprint right.
Just in terms of how much capacity has been taken out of the system and how much is cyclical versus structural, I think you've let about 10,000 people go. It's hard to bring those people back. It would take time. Of course, you could if you wanted to, but there is a structural element to the fact that, you know, even if you brought those people back, the capacity of the industry will be lower over time. So can you talk about how much of the capacity reduction is structural versus cyclical?
Yeah, hard to quantify, but I think it's there is some structural, to your point. There's 15%-20% that's a structural number because of underinvestment through this period and because of, I'll say, utilization. Some of the new technologies are not as efficient inside the factories. You know, they take longer. There's more cycle time, if you will. And then there's the piece of mothballing facilities or letting people go, which is really the painful thing for us. It's not just us, it's also our suppliers. So keeping the orchestra all in tune is a big thing, because if one supplier says, "I'm not building anymore," then all the other suppliers having excess capacity could actually hurt the profitability of themselves and then, therefore, the whole industry.
Getting that right is what we have to really look out for as we grow back into the right footprint. I'm confident we'll be able to do it, and I, you know, I think we'll be able to have suppliers come with us along that journey. But it's been a tough road, and I don't think we're exactly there just yet.
Yeah, I wanted to ask you about that because a drive is, I mean, it's, it's a very, very complicated product, and it has a very complicated, you know, chain of supply, very, very long lead times. So it's not just your capacity, it's the supply chains as well, to your point. So how active are you in looking out for, you know, trouble spots? I mean, it's been a challenging time for everybody, and there may be, you know, some of your key suppliers that have had a very, very challenging economic experience over the past couple of years. And, you know, any one weak link is gonna really hurt you. So how active are you in managing that?
Could we rightfully worry about there being a weak link that is gonna cause a, you know, hiccup in the ramp?
Yes. I think we're through the period largely, but, you know, not with every supplier, so we still have some work to do. But I think we're talking to each other. I mean, it, we had a similar event in 2011 with the Thai floods, especially with mechanical suppliers, who their margins were relatively low then, and they were totally dependent on the hard drive business, you know, with billions and billions of parts going in, you know, coming out, especially mechanical parts. So recently, some people have been able to diversify, which is great. They're healthier businesses than just totally reliant on hard drives. But there are still some companies that are super reliant on the hard drive industry, and that's why I always say we have to look out for the entire industry.
If, you know, capacity-wise, if one person goes down, it's everyone's problem, and I think we're making sure that we share the burden and the gains once they ultimately come with those critical suppliers that have invested in the platforms that we want to take forward.
You know, one thing that I think is pretty amazing is that you've done a really great job generating free cash flow throughout this entire downturn. I mean, there's, you know, I can look at, you know, you look at any company, you look at any memory company, and you look at what's happened to them. I mean, you know, you're in the same vertical, and you've been able to generate cash flow throughout this entire downturn. How much of that, though, you know, you've cut CapEx. CapEx is about half of what it was in fiscal 2021. Whenever you see CapEx come down that much, the inevitable conclusion is, well, you're sacrificing something out the other side. So how have you managed that process? You know, you're trying to ramp HAMR, yet your CapEx is much lower.
How, how do we not get concerned that maybe there's some longer-term impact that could happen to your roadmap?
Yeah, it's a good question. I would say that one of the things I'm most proud of is that during this downturn, that was so painful on our company and our people, and we were able to execute on the technology roadmaps and continue the investments in the right type of CapEx to make sure we get back up out of this thing. I think, you know, maybe it's just a reflexive behavior that I have from many, many years, but it's like you see a downturn, and you say, "Get to the new technology or the new products." You know, this is exactly what happened in 2009. We said, "We don't like our old products. Let's get to the new products," and we went from, you know, the depths of despair to record profitability in five quarters.
I mean, that to me, it's always go to the next. And so we have not, while we've been very, very careful on CapEx, we have not in any way compromised the ability to ramp HAMR really hard. And, you know, we'll have to watch the demand go and then, you know, be as aggressive as we can because there's long lead times on that CapEx as well.
Can you just talk about pricing? You've gone to your customers, again, I asked this upfront, but you've raised prices. It's just as astounding how these large companies are still reliant on you. And, and so is there a point where your customer is gonna maybe have to start to co-invest alongside of you or, you know, prepay for supply? Can you—I guess I'm asking just a big picture about what the lasting structural changes will be-
Mm-hmm
... in this industry over the past five years or next, you know, five years. Given how painful things have been, could you envision a scenario where actually they have to literally prepay for capacity?
I think the answer to that is no, but I do believe that the build to order is a form of that.
A form of that.
So basically, if you say, "Hey, I need nothing this quarter or next quarter," right in front of me, then I'm gonna zero you out further out in time. I mean, that's just the way it has to be. And given the lead times, then, you know, you're probably not gonna have product for a while. And you may make that decision based on the economics of what we're offering today, but then you're seeing the inability to garner the economics that we're showing you in the future. And if those future economics are based on a much better TCO proposition, you're probably just hurting yourself. So in some way, we're asking everyone to be much more predictable on their demand rather than bursty, and because we've seen that in the past.
And, you know, I think, especially at the supply levels that the industry has today, it's a reasonable request that's being met. I, you know, I think people realize that they, they bought 2x or 3x more two years ago, and they may get back into that mode someday. So they, if you wanna, you know, go at x now as, as the predictability factor or whatever, that's, that's all that our industry really has the money for. I mean, we, we just don't have the, the free cash floating around to be able to speculatively build stuff, so.
Yeah. Yeah, I guess just last question, I mean, it just comes down to pricing and really just the tenor of the discussion has... Do the customers realize this? I mean, it's, you know, sometimes there's a subtle change in the tenor of the discussion, and you can see, oh, actually, they do value me as a more important partner.
I think so. I think also as we look forward, one of the things we can do is we can get people to go to the next node, and the next node, and the next node, right? So you get them to leap to the new drive that has a better TCO proposition for them, and relatively for them, they're still seeing benefit. That's the way to think about it. And as far as selling the old stuff, you know, we don't want to sell the old stuff anymore, so prices will go up there.
Great. Well, we're out of time. Thank you, Dave.