Seagate Technology Holdings plc (STX) Earnings Call Transcript & Summary

All right. Great. So we'll get started here. Thank you very much. So again, my name is Erik Woodring. I lead U.S. IT hardware coverage based out of New York. Quick disclosure from my end. Before we begin, please see the Morgan Stanley Research disclosure website at www.morganstanley.com/researchdisclosures for important disclosures or reach out to your sales representative. So I am delighted to be joined today by Seagate's CFO, Gianluca Romano, been a mainstay with us for a number of years. So Gianluca, thank you for coming. Excited to host you today.

Thank you very much, Erik. Thank you for inviting us. And before we start, let me remind everyone that I will be making forward-looking statements today, and you can learn more about the risk associated with those statements on our website.

Perfect. So I want to start very high level because it has felt like there's been a demand inflection over the last few months, not to say this year hasn't been very strong. But the last few months have felt very strong from a demand perspective. And I'd love to get your view on what you believe is kind of driving that underlying strength because -- we cite video very often, but video is like multimodal still feels early days, a great opportunity, but still early. Token growth is very significant. So just like where is the demand coming from as you see it? And maybe if there's any difference relative to the Analyst Day that you had back in May, what you're seeing?

Yes. I think when we have the Investor Day in May, we discussed about a certain revenue growth, especially in terms of exabyte revenue growth of about 25% in the nearline space. I would say from that point to today, we have seen demand even to be stronger. It was already very strong before, but I'll say possibly, the gap between supply and demand has been growing a little bit. There are many drivers, as you said. And it's difficult for us to know exactly what is the main reason of one growth in demand compared to a different application. Video is, for sure, one and I would say normal video and AI video. Video is consuming a lot of storage. And of course, when it start to increase in terms of volume, volume for the storage consumption becomes really, really huge. But it's not the only one. I think in the near future, you will see other applications like autonomous driving or robotics. Those are -- they're all applications that are in development today and will start consuming exabyte storage fairly soon. So we are very optimistic, not only for the short term, but also for the long term. I think the world is going into a direction where all the important applications that will be used by companies and by people are based on data and data storage.

Okay. That's a perfect segue because we're kind of talking about a world that's moving into more AI workloads, like we haven't -- we just haven't necessarily seen and correct the record if I'm wrong, but like these huge orders coming from AI native vendors, whether that's Neo clouds or AI labs. Where did they play in the world of HDDs? Or is it just too early to say because they're either so focused on training or leveraging compute that they're just maybe not at the storage layer yet. I just love your view on that.

No, I think they are very important players as they focus on running the AI application. Usually, they work for or in conjunction with big cloud service provider like AWS or Google Cloud or Alibaba, Microsoft Azure. So they import the data from data center that are run by those other customers of them. They run the application and send the data back. So the data and where the application is run is not in the same location. And this is a different model compared to a normal, I don't know, AWS or Google Cloud data center, where in the same location, we have a lot of storage and you have the application where they run. The Neo cloud, they only focus on running the application. So they import the data from those other cloud service provider, they run application and send it back. If one day, they become, let's say, a competitor of those cloud service providers, they will have to create their infrastructure. And that adds some complication for them. It's a more complicated infrastructure to run. But it's a step that eventually they will have to go through is a tend to go up into the stake and compete with their -- today their customers.

Okay. And you kind of answered the question I was about to ask about how you think about kind of the midterm growth of exabytes, stronger than the kind of '25 nearline you gave at your Analyst Day. Is there a way that we can kind of gain conviction in that by either thinking about some sort of like exabyte per megawatt math or historically, we've been able to say a certain percentage of cloud CapEx? But GPU spending has maybe made that relationship a bit murkier. But is there a way that you think high level that we can all approximate kind of HDD demand top-down using some of these kind of simplistic metrics?

Probably the best way is still looking at the CapEx of our customers. Now we are a percentage of that CapEx. We are probably in the mid-single digit of the CapEx. Of course, the CapEx is growing. And so the part that is allocated to storage is going to grow. Depending from the quarter, depending from the year, there is more focus on one component or the other. When our customers, they were very focused on buying more GPUs. Now people were saying, "Oh, this is detracting from your business." And I said, "It's just a matter of time." All those GPUs will generate data and that data has to go into our hard disk to be stored. And actually, the time arrived and demand for us is very strong.

Yes. Yes. Okay. Let's touch on supply because Seagate and the industry more broadly has maintained a ton of supply discipline since the last down cycle. And instead of saying, will you expand unit supply? Obviously, this is kind of a hot debate. Maybe my question instead is, is there some form of catalyst that would get you to change your mind on expanding unit supply? Or what would it take to expand unit supply from your perspective, if anything?

Yes. Right now, we don't see any reason to do it. I think with our product road map and with the industry product roadmap that has been defined by the various competitors, I think the industry will generate enough exabyte capacity without increasing the unit to serve the short-term need of the data center. And when I say short-term need means data center that has been built recently or are going to be built in the short term. We are probably not covering the additional demand that is the inventory or safety stock demand or maybe what our customers thinks they will need in 2 or 3 years, but it's time to go there. I think customers look at more unconstrained demand. So their demand can be really huge, even going back to the video application and video AI application. But data centers are complicated. There are a lot of components and there is always some bottleneck at a certain point. So we -- our job is to bring that unconstrained demand to something that is maybe more constrained. And therefore, we think if the industry grow about 25% in terms of exabyte will be enough to serve the short term, not having any data center that is being built and not having hard disk. So we will not be the bottleneck for data center. But at the same time, we will not oversupply, and we will not generate that level of inventory that we had in the past.

Right. Okay. So touching on another question related to supplies. Is there a way that you can help us all understand how -- like the gap between supply and demand now? And how much like -- and we'll get into technology innovation because it leads to this question, but can your supply or should your supply kind of improve in calendar '26 and '27 as you come out with higher capacities of HAMR as you kind of cross that crossover thresh point -- threshold in kind of mid-calendar '26? So the question maybe is how short of demand are we and how much can supply improve as we look out maybe the next 2 years?

Yes. Our product road map for the next 4 or 5 quarters is mainly based on second generation of HAMR. This is a 4 terabyte per disk, so 40 terabyte drive, is a huge improvement comparing to the first generation drive, HAMR drive that is a 30 terabyte. So of course, we don't move all the mix from one to the other, but we can generate a fairly strong increase in exabyte when we move customers and when customers qualify bigger drives. And this is what we are going to do quarter after quarter during calendar '26. We have the first 2 customers that are already in qualification for the second-generation HAMR. So we think we can start probably selling some units in the first half of calendar '26 and ramp much more in the second part. And after that, we will go into the 5 terabyte per disk, or 50 terabyte. At our Investor Day, we show a further road map that actually is continue to grow up to 10 terabyte per disk, so 100 terabyte drive this is an important step in HAMR technology, but it's probably not the end of HAMR technology.

Right. So what does this mean for your visibility today? You've kind of updated visibility at each kind of successive point where you've been able to interact with Wall Street, whether that's earnings or conferences. And it feels like every time we interact, some data point on visibility has elongated. So at earnings, you were largely committed for calendar '26 nearline. You had some LTAs extending into calendar '27. Where do we kind of stand now on kind of supply visibility?

I would say, we are similar to what you just discussed. We want to have firm orders when we start the product in our manufacturing. And to build an hard disk you need between 6 and 9 months. So it has to cover 3, 4, 5 quarters so that we don't get any surprise after we have started material in our manufacturing. So this is what we really focus on is the next 4 or 5 quarters to have orders in place, well defined with all the terms. After that, it's more a customer request. So customers today, of course, are very interested in getting longer-term agreements. So we have LTAs in place for our calendar '27, in some cases even longer. This is not as well defined as a purchase order, it's more a discussion and an agreement on overall exabyte. And exabyte that we can generate and we can allocate to them. So they know what they will get from us. And therefore, they can plan their data center around all the components, but at least knowing what we will provide to them.

And just one quick clarification point on that, which is kind of -- so purchase orders are firm orders, LTAs are, let's call them, demand planning tools. Is that the right way that we should all be thinking about the difference between the two?

Well, every LTA is different. So overall, some LTAs that are a little bit more defined. But in general, yes, I would say for us, it's not -- so it's more a customer demand. They want to have those LTAs. We want to have purchase order for when we start production in our manufacturing. This is the most important. We want to support the real demand. So if a customer doesn't want to have hard disk, it's okay. We can move that production to someone else or we can align our manufacturing to the real demand. Building inventory is not a solution. That means at a certain point, the demand goes really low because there is a lot of inventory. So forcing someone to receive a product, it's not a great strategy.

So -- and this is a great point. I'd love if you can help us understand because I think like the pushback that I would get or that you get is, well, there's double over ordering when you talk to customers. How do you fight back against that risk? You've kind of alluded to it.

Yes. We don't serve the full demand, as we said before, and that demand is actually moving out in time. So we see that demand not going away. If they're shifting out in time, that will help with eventual cycle at a certain point. We don't take the big demand, but possibly the down cycle will not be as deep as it has been in the past. And when we can generate a little bit more exabyte out of our manufacturing, we see. We put that in the market at a fairly healthy price and see how quick it goes away and usually goes away very quickly. So that means demand is actually here.

Right. Okay. So let's talk about kind of the reasons for being able to expand exabyte capacity. So HAMR, you were the first to market with HAMR. You've already qualified 5 customers -- at least 5 cloud customers, at least on the Mosaic 3 platform. You just mentioned some of the timing around Mosaic 4 and the 40 terabyte platform. But can you just help us understand again, how we should think about the underlying mix of your exabytes going into HAMR or going into 40 terabyte versus 30 terabytes? Like what's the time line as we should think about mix here?

Well, I would say this fiscal year that will end in June is mainly in terms of units sold, is mainly for HAMR first generation, so 30 terabytes. So we will have some of the second generation through the end of the fiscal year, but the majority of the units that we are producing today and are going to be sold in the fiscal year are mainly Mosaic 3, so 30 terabyte. And we said by the end of this fiscal year, 40% of our exabytes sold will be based on HAMR technology. So in this case, mainly the 3 terabyte per disk. The next fiscal year will be a fairly big transition from 3 terabyte per disk to 4 terabyte per disk. And we said by the end of the following fiscal year, 70% of the exabyte will be sold with an HAMR technology. So there is an increase in the overall HAMR volume and there is a transition out of first generation HAMR into second generation HAMR. Second generation HAMR, if you remember our Investor Day, is probably when we also get a major improvement in terms of profitability. We have a slide that we use for Investor Day, where we show the cost per terabyte. And if you compare our last PMR drive and the first generation HAMR, the cost per terabyte of the first generation HAMR is a little bit lower but not much lower. When you go to the second generation HAMR, the cost per unit for us to produce first -- a second-generation HAMR is fairly similar to the first generation HAMR, but we increased the capacity -- but we increased the capacity per unit by more than 30%. So this is a huge cost reduction for us in terms of cost per terabyte because finally, we sell terabyte. And of course, the price per terabyte is based on supply-demand. So we think with the second-generation HAMR, our profitability will continue to increase.

Right. So I'll get back to HAMR in a second, but I wanted to touch on that pricing because there's a lot of moving pieces, in some cases, because of supply/demand is imbalanced. You've been able to raise prices. You've been able to lock in pricing with some of these purchase orders, but at the same time, as you're mixing into higher capacity drives, that price per terabyte is a little lower, so you could give back some of that margin to your customers. When we add all of that together, how should we all be thinking about price per terabyte in totality for Seagate as we think, again, like 2, 4, 6 quarters down the line? What's the right way to be thinking about that today?

We have used this new methodology since more than 2 years, like probably the last 10 quarters. So if you look over that period of time, we have quarters where the price per terabyte is a little bit higher sequentially. Sometimes where it's slightly lower. But I would say, it's generally fairly flat. So let's see what will happen in the future. But the strategy is the same. So I expect a fairly similar result from allocating the same strategy. Of course, every quarter is different because it depends how many agreements you are renegotiating, who is moving to the next product, who is not is getting a price increase. And the rent because finally is how many units are you selling of that product in the specific quarter. So sometimes the impact in the P&L is a little bit later than the agreement. But I would say we have been fairly flat until now. So I say flat to maybe slightly up is what we expect.

Okay. Okay. And so to get back to HAMR, the overarching question I have for you is how much of a competitive advantage does HAMR provide Seagate? And I ask because if maybe demand wasn't as strong. You can clearly deliver the best TCO with HAMR. Demand is so strong that you have customers giving you LTAs into early calendar '28. So -- just I'd love to know how you think HAMR gives you the competitive advantage in this type of kind of extremely strong demand environment?

Well, you are right. Customers are looking for exabyte, and this is what they want to receive from our industry. I would say if you have an higher capacity drive, HAMR is a way for us to generate higher capacity drive, not only with the first generation, but going to the second generation, which is a 40 terabyte and then 50 terabyte and 60 terabyte. So continue to increase capacity is giving you a better financial return. Because with the pricing strategy that we have discussed, the majority of that cost savings remain into your profitability. So this is probably the advantage of HAMR. And of course, we need to see what competition will do in term of HAMR development. But now if they have a similar product, I guess, we will have a fairly similar financial return. If they have a product that has a lower capacity per unit, of course, we will enjoy the best cost per terabyte.

Okay. We've kind of talked about the near-term risks that some investors have around the longevity of the cycle, around unit capacity. Longer term, how is Seagate thinking about the kind of SSD disruption risk? Some of the players on the NAND side believe that they'll have a kind of a cost competitive product by 2029 using hybrid bonding and whatnot. Just I'd love your reaction to how you think about the competitive landscape beyond kind of the shorter term?

Well, I would say HAMR technology is a way to reduce the cost per terabyte, right? And I think, especially in the first 4, 5 years of the new technologies when you get the majority of the cost decline. Going from 30 terabyte to 40 terabyte, you're increasing your capacity per unit by more than 30%. And then you go from 40 to 50, another 25%; 50 to 60, another 20%. So there is a lot of opportunity for us to reduce the cost. Now cost and price are two different things. I would say, NAND is not really reducing their price today. It looks like price is going up. So that gap is actually increasing, not decreasing. I would say, for storage, but it's no reason why a customer should pay 8x more to buy a terabyte of storage. So they will always give priority to hard disk. And that is if you look at the last 15 years, the data center infrastructure has not changed. And we had very different dynamics between NAND and hard disk, but the architecture is already optimized and the storage is on hard disk and running the application -- part of running the application includes the NAND. So NAND volume is actually growing a lot, but not for storage. It's growing because it's needed when you run the application. So the data has moved from hard disk into NAND. You run the application together with DRAM and GPU and CPUs and other components and then data is going back to hard disk, is very simple and is not changing for many, many years. That is -- there's no reason to change it now when actually the HAMR technology can drive that cost per terabyte even lower than what was the prior technology.

Right. Okay. And maybe on HAMR, can you maybe speak to the qualification process? Obviously, you guys came -- were first to market with HAMR, there were some hiccups along the way that you seem to have triage and that is now in the past. How do we think about the qualification time lines now? Can that accelerate? Has anything changed? Are there any bottlenecks as you see it going from 3 terabyte per platter to 4 terabyte? Yes, let's just -- in totality there, maybe.

No, in general, there are no differences, right now is a normal product qualification. Of course, the first product with the first customer had to go through an intensive testing, and we had to find the right configuration for the drive to work into the cloud that require a more intensive usage of the device. But once we find the configuration, we qualify customer number one. And after customer number one, we qualified 4 more customers in a very normal time, usually to do well, you take between 2 and 3 quarters because that's the time of the test. It's not for any other reason. So -- after that, everyone is qualifying with a normal time. So we don't see any reason why a second generation HAMR should take longer than that time.

Right. Okay. Last two questions to keep you on time. First is just we've talked about kind of flattish pricing. Obviously, with the benefit of moving up the aerial density curve, you should be able to reduce cost per terabyte. We are -- I think last quarter, your gross margin expanded by 220 basis points sequentially. Maybe my question is, does that pace continue? I mean if we see this kind of wedge continuing? Is there not a clear path to 50% plus gross margin?

Well, it's not that we have a specific number in mind. We are executing our price strategy. We have done for many, many quarters, and we will continue to do it because the supply-demand situation is actually the same or actually maybe even a little bit more disaligned today than what was in the prior quarter. So there is no reason to change our pricing strategy and the mix is moving in the right direction. So of course, moving from the last PMR product into a 30 terabyte HAMR has some benefit. As I said before, when you go to the 40 terabyte HAMR, you have even better results. So we will continue to go in that direction. And depending from the time of the quote and time of the ramp, our gross margin will continue to improve. We gave a model just a few months ago at our Investor Day, we have done better than that so far, and we will continue to optimize as much as we can. I'm also saying that model when you run that model for 3 or 4 years is giving already a fantastic result. So if we can do better, it's all good. But is not a bad model anyway.

I have to touch you on it, of course. And then maybe last question. How do we think about -- we're in a strong up cycle right now, how do we think about capital allocation? Any changes to capital allocation? Obviously, you have your dividend and that's protected. That's kind of your first priority. But how do you think about delevering versus buybacks? You obviously have a convert out there. Just help us all understand maybe what the path forward is because obviously, your cash generation should be very strong as we keep looking forward.

Yes. So we said we will -- we'll return to our shareholders at least 75% of our free cash flow, which is, of course, over a period of time that we discussed at the Investor Day, that was a 3-year period of time. So every quarter is a bit different. In the short term, we gave priority to reducing our debt. A few years ago, we were above $6 billion. Right now, we are at $4.5 billion. Early this quarter, we actually reduced the convertible by $500 million. I think we will reduce our debt a little bit more, but we have also increased the dividend. So January payment will be the first payment with increased dividend. And we have restarted share buyback in September. And I think we will do substantially more in the future quarters.

Perfect. I'll keep you on time.

Perfect.

Thank you very much. Appreciate it.

Thank you, Erik.