Good morning, everyone. See a few familiar faces here. So I'm Asiya Merchant. I work for Citi Research. I cover the tech hardware, tech supply chain stocks. Really happy to have Seagate CFO here, Gianluca Romano. This is an open fireside chat, so we do have, you know, a set of questions. However, if investors do have questions, I'll allow some time for that. Please do raise your hand so we can bring the mic to you, so that will enable it to be webcasted efficiently. So before I kick off with some questions, I'm going to turn it over to Gianluca Romano, I mean, to go over some prepared commentary.
Thank you. Thank you, Asiya, for inviting us here today. As usual, before we start, let me remind everyone that I will be making a forward-looking statement today, and you can learn more about the risks associated with those statements on our website. I also would like to announce that we have recently achieved a new qualification for our HAMR product. So we now have four major cloud customers qualified. Three are in the U.S., and one is outside the U.S. So we are going very fast through all those qualifications, and we still expect to complete all our top seven customers in the cloud space qualified by next June.
All right. Gianluca Romano, you guys have had a really strong quarter. You know, I've been asking all my companies here, as we sit in calendar 3Q, and we kind of look at, to though it's the start of the year, there was, you know, all kinds of events that made my life quite miserable. There was like D.C., there was tariffs, there was no tariffs, Section 232, whatever. As you kind of sit back and reflect on, you know, how your end customers, cloud customers, you know, video imaging application customers, et cetera, when you kind of sit back and think about how demand has evolved, you know, can you share some thoughts on, you know, how you're looking at calendar 3Q, where you're sitting right now versus where it was at the start of the year?
Yeah, you're right. No, life is complicated. No, all those items that you mentioned, plus, no, many more.
Yes.
When you run a business, there is always some complication here and there. I would say, despite all those complications, the calendar year is coming out, as we were discussing at the beginning of the calendar year. No, we said through this calendar year, we will increase our revenue sequentially every quarter. We will increase our profitability sequentially every quarter. And no, we have done in Q1, Q2, we guided Q3 sequentially higher. So I think we are executing what we said. And no, we are happy, of course, in particular for our qualification in the HAMR space. Now, this gives us the opportunity to ramp up exabyte volume through mix. As you know, no, we don't want to increase units in terms of disk and heads. But having those quals allow us to move customers from a 24-terabyte PMR to a 32-terabyte HAMR.
And as you probably know, we have started qualification of our 40-terabyte HAMR in July with our first major cloud customer. So we are progressing well. This is our way to at least be close to the increase in demand. No, it's not unit volume, it's exabyte volume. And to increase exabyte volume, we need to drive the mix up in capacity.
Okay. And then, you know, interesting, like, you know, to be close to demand, which means, which would imply that you're still seeing strong runway for demand as you guys think about 6-12 months. And this whole notion of LTAs with your customers, it's interesting for a lot of investors. So just help us think about some of the puts and takes to kind of that growth indicators that you're thinking about looking ahead?
Yeah, we see demand growth to be, you know, around mid-20s for nearline. Of course, no, every quarter is different. Every year is different. I think in the last four quarters, nearline grew like 90%. So no, every period is a bit different. But demand is growing very rapidly. It's well above supply today. And this is why we focus a lot on the quals as we were discussing before. We need to move customers up in terabyte per unit in order to stay at least close to where the demand is going. But for us, I would say the focus in terms of customer partnership is based on build-to-orders. The build-to-orders is very simple. It means we need to have an order before we start a product. So we need to have three or four quarters of visibility. And visibility is not, no, theoretical visibility is an order.
Right.
So we get an order, we start the product, and we know exactly who will buy that product, when, and at what price. This is what we call a real partnership. They tell us what they need, and we produce, and we sell, and we have an agreement. Of course, no, in this period of time, customers are willing to go even much longer. So we have LTAs that are longer than three or four quarters. But where we really focus is the visibility of our customers in the next three, four quarters based on the data center that they have. They need to refresh the data center that they are completing, and they need new hard disk to be installed, which is the most important. Of course, having even a longer visibility is a positive.
But the major, major focus is, no, three, four quarters is a good time for us.
Okay. And then remind investors, you know, why are these data center customers, if you're giving them this three to four quarters, is that how long it takes for you guys to produce the exabyte that they need, the disk that they need? Is it a technology HAMR qual issue or just help investors think about why three to four quarters is the right?
Yeah, it's a time to produce, especially when you're going very high capacity HAMR drive. It can take about three quarters to produce. So having three, four quarters is good enough to have a good plan.
Okay. All right. And when you think about, you know, the confidence that you have in that, I mean, I think people also want to understand that, you know, there isn't any element of like over-ordering or, you know, what gives Seagate the confidence? Is it CapEx plans that they share with you? Is it goes, you know, level deeper, which gives you the confidence that yes, indeed, what they're buying is actually being consumed and not being stored in a warehouse somewhere?
Yeah.
Yeah.
This is also why we want to focus on three, four quarters. When you go longer, the visibility for our customers is difficult. They have a number of customers that is even difficult to think about. They have a lot of people, a lot of companies, they have governments. So they cannot call their major customers and ask, "What is your plan two years from now or three years from now?
Okay.
You're talking about millions of clients and customers. So they cannot do it. So they need to do in a different way. But if you look three, four quarters, they know their internal plan when they want to refresh a certain location, data center location. They know exactly what are the data centers where they are building and when they will be complete and when they need to install the new hard disk. So this gives us confidence of where our hard disks are going and based on what internal plan of our customers. Then when you go longer, again, eventually, even if they buy the drive, you don't know if they buy for inventory or they buy because they are installing. So it's a bit far in time, it's more complicated. So we like to stay close to that period of time.
Okay. Fair enough. Section 232 tariffs, I was expecting them to get resolved here and hopefully, you know, hear from my companies about some updates because a lot of them have been subject to exemption under that. You classify under semiconductors as well for that? So just, you know, if these were to come around, how are you guys just thinking about potential scenarios that could unfold as these exemptions, you know, come into play?
Yeah. Today, there are two rules that are exempting hard disk from tariff. One is a 232 and one is a USMCA.
Sure.
So it's difficult for us to have a plan until we know what is going to be, you know, eventually a new decision or if the current decisions are going to stay. So we need to wait and see. We don't expect any problem in the short term. If both rules are changing in a way that we are not exempt anymore, our manufacturing location for the product that we import in the U.S. is Thailand. So the tariff in Thailand today is 19%. Even that one could eventually change. I think competition is in the same location. So I think we are in the same situation and we will see what happens later. But right now, we have two rules that are exempting our disk.
If the exemption were to require manufacturing or increased manufacturing in the U.S., can you remind investors, what do you manufacture in the U.S. and what you could potentially use as a lever?
In the U.S., we have huge manufacturing. We are producing all the heads. Particularly the HAMR heads are in the U.S. We have two locations for heads. One is in Minnesota and one is outside the U.S. in Northern Ireland. But we have a huge site in Minnesota. We also have a lot of R&D. A lot of R&D CapEx is actually spent in the U.S. Again, we don't know what are eventually the new rules. If our CapEx is good enough eventually, or there is an expectation for higher CapEx, if that is the case, we will consider what is the best solution for the company. But of course, we evaluate and we consider all the opportunities.
Okay. All right. Let's jump to demand. You know, you guys, one of the questions that I often get from investors is the cyclical recovery, you know, after the downturn, cloud investments, and of course, they need to build new data centers. But then you also have the kicker from AI, which is more of a secular story. So help us understand, you know, how you're thinking about, how Seagate's thinking about the incremental demand that you're seeing from AI. Where is this demand? You know, how sustainable is that demand? How could it grow further? Just walk investors through that scenario. Yeah.
I'll say there are two positive impacts to storage from AI. One is retention. Companies that want to use AI, they want to have a lot of data. And therefore, they keep data for longer and longer and longer. And now that first impact, we saw already many quarters ago, like six, seven, eight quarters ago. Now we see the second positive impact, which is AI is using the data to generate something that is valuable for those companies or people. And they keep it. So now AI is generating data that is increasing storage. And I think this second part is probably the most important and is the one that will continue to grow by a very high rate. I would say it's probably the first time when you have an application that is generating data by itself.
So it's a big change and is possibly huge volume, especially when the video part will continue to grow. Video is, of course, consuming much more terabyte or exabyte. So that part is even more interesting to us and it will come in the near future. So AI is extremely important for storage. We also believe, now looking at the past, we also believe that you can have cycles. And keeping demand above supply is also a way to reduce the impact of that cycle because you don't take the full demand that is available. So that demand is shifting to the future and eventually reducing the impact of a future down cycle if that happens. So I'll say, you know, different situation, but we are ready to manage also cycles that we have eventually seen in the past.
Okay. And when you talk about the video application, I think Dave's talked about, you know, autonomous vehicles, for example, or even robots on the factory floors. I actually saw that at Computex where they get trained before they hit the floor. Cars get trained before they hit the road in the factory itself with, you know, synthetically generated data that you talk about. Are there other applications? Are these the two main ones? Just kind of, you know, help investors.
I know the nice part is that innovation never stops.
Yes.
And innovation is always based on data. So storage will benefit from all the applications that we still don't know today, but maybe in two or three years, we will be talking about. If you think about AI, two or three years ago, we were not really talking too much about AI yet. And now it's a main generator of storage. Possibly in two or three years, we will see new applications or some of the ones that are today very small will start to generate a lot of data and a lot of need for data storage. So that is the good part of our business is based on data, and data is what is growing into the world in many different applications, even one that we still don't know about.
Okay. When we talk about, when I was at a, in August, I was at the Flash Memory or Future of Memory Summit, as they call it right now. And there, you know, it was interesting. They have obviously the massive data lakes that you said, unstructured data, that videos and other data that lives on HDDs. But now you're also starting to see a little bit of these high-capacity SSDs that are kind of in between feeding to the GPUs versus pulling some data from these data lakes. Just talk to investors about, you know, where you see that use. I think Matt has talked about this tiered storage. You know, how do you see that use case evolving and the impact that you think it could have on this potential demand, continued demand growth for HDDs?
Yeah, no, we are both important components inside the data center. Hard disk and NAND are used differently. The storage is done on hard disk. When you run the application, you use SSD and DRAM and GPUs and CPUs and many other components. So data actually moves out of hard disk into an SSD before you run the application. And then when you have the result, you save it back into hard disk. So it's not one or the other. No, we are two different components. We are growing a lot, as you know. SSD is growing a lot. I think that is a very good future for both of those components, but they are used differently. Sometimes people are confused because maybe they look at the laptop and they say, "Oh, there is no hard disk in the laptop anymore." So SSD is replacing laptop.
Now, if you think a little bit more careful, the storage of your laptop is in the cloud. So it's still in hard disk. It's just in a different physical place, but it didn't really change. So you use SSD mainly to run the applications on your laptop and do a very minimum part of your storage because you can use SSD for a minimum storage. But the vast majority of the storage, especially if you are in engineering, especially if you use AI, is in the cloud or is in the on-prem data center. In both cases, it's still on our hard disk.
Yeah. Good point. When you talk about demand and the fact that you are still under shipping demand, both you and your major competitors here, but then you also have NAND and SSDs, which are plenty available. And so just help us understand, you know, when you think about where to invest, how to invest, and how to kind of walk this delicate balance between not bringing on too much capacity, but making sure we don't leave opportunity on the table either. How are you thinking about that balance?
I think to be perfectly right is kind of impossible, so eventually we leave some opportunity on the table. There is some demand that is not satisfied, but as I said before, it's not really bad. That demand is just shifting to the future, to the future quarters, to the future years, but keep a little bit of tension between supply and demand that is good for pricing, and again, it's also good in case there is another cycle later on to reduce the impact of that cycle, so I would say you don't want to leave too much revenue on the table, so you want to try to be close in terms of growth, in terms of what you can do internally and what you see the demand growing, but it's okay if you are slightly below. It's not a real problem.
It's not going to impact overall end demand the way you think about the Exabyte CAGR. All right. Before I talk a little bit more about HAMR, you know, you do also ship to enterprise OEMs. I know there is a smaller percentage now of your end user base, but just help us understand what do you think about demand there, how that's evolving. Because it clearly seems like hyperscalers are doing much better.
It's a very good and a very important segment for us and for storage in general. But as you said, if you look at the last probably two years, the majority of the growth happened into the public cloud. So it looks like there is a little bit of a different trend in terms of storage growth between public cloud and enterprise OEM. Now, we think cloud is probably going to grow to be the major segment for us in the future, but enterprise OEM will also grow. On-prem data centers are very important. You never know. Depending from the new application, where companies want to store their data is on-prem or in the public cloud or both. We will see. For us, it's not really different. We produce hard disk. We can ship a hard disk on a public cloud or an on-prem data center.
It's the same product. In general, the highest is the capacity, the better is the profit for us. So public cloud has this benefit for us, but it's not so material. It's good also the enterprise OEM. We are happy to ship there.
Right. Yeah. I guess people sort of want to understand if there is, if you're seeing anything, you know, we're talking a lot about inferencing, agentic AI, which typically lends itself to a little bit of AI on the edge. People want to be a little bit sensitive about the data. So just wanting to see if you're seeing anything?
Yeah, no, we see. Enterprise OEM is a big business. But public cloud is going to be a huge one.
Okay. Fair enough and then you also do sell, you know, other mass capacity applications as well, right? Whether it's video imaging for cities and things like that, so anything to think about there? Was there any pull forwards there?
Pull forward, I don't think so. No, there is a part of our business that is still seasonal. Not the public cloud, not the enterprise OEM, but almost everything else has some kind of seasonality, like the client, consumer, the video and image application. Generally, March is the lowest quarter of the calendar year, and then the business tends to grow through the calendar year. But it's also true that year after year after year, the nearline is becoming bigger and bigger. And right now, it's probably 75% of our volume. So the seasonality in the future will be less impactful compared to what we have seen in the past.
Okay. Coming back to HAMR, that's great. You have the fourth call there. How do you think about the market share at your key cloud customers? I think you've said you're not necessarily chasing share here.
Oh, yeah. No, we grow based on mix. So the more customers we qualified on HAMR high-capacity drive, and as I said before, now 40 terabyte is starting the qualification. Actually, already started with one big cloud customer in July. The more exabyte we will produce, and then depend what competition is producing in a certain quarter and how their mix is evolving. Market share can change a little bit. But it's not when we both sell all our exabyte, it's not so important how much is the market share. What's important is what we produce is sold, and it's sold at a good price. And we always try to improve our internal efficiency to get the lowest possible cost per terabyte and generate more profit. But exabyte market share, especially in this period of time, is just a matter of supply. It's not a matter of demand.
It's not a matter of which customer. It's a matter of what we can produce because what we produce is actually sold.
Okay. And then just remind investors again about margins. I think you had some pretty strong statements that, you know, as we ramp HAMR, our incremental margins are pretty robust here. So just remind investors, you know, how to think about these margins, especially as you're ramping HAMR.
Yeah, no, margin depends a lot also from the cost per terabyte. And of course, the higher is the capacity, the lower is the cost per terabyte. We presented at our analyst day a slide in terms of cost per terabyte where you can see fourth-generation HAMR, which is a 30-terabyte drive, already present a slightly lower cost per terabyte than our last PMR drive that was a 24-terabyte. And then when you go to 40, there is a much bigger decrease in cost per terabyte. The main reason now between the two technologies, the main difference is with the PMR technology, to grow capacity, you need to increase the bill of material. So you need to increase the number of disks and number of heads. So if you have a two-terabyte hard disk drive, you open the box, you have one disk and two heads.
For a 20-terabyte drive, you open the box and you have 10 disks and 20 heads. HAMR with 10 disks and 20 heads instead of 20 terabyte can produce 30-terabyte in generation one, 40-terabyte in generation two. So it doesn't need to increase the number of disks and heads anymore. All the increase in storage is coming through areal density, which means more storage per disk. This is why HAMR is so important. The space in the box is limited. So there is space for 10 disks, of course. There is space for 11 disks, but I think competition is already doing because if you don't have HAMR, you need to grow through bill of materials. But it's probably space for another disk, go to 12. After that, there is no space anymore.
So if you don't have HAMR, your ability to grow on terabyte per unit is limited. So it's capped by the number of disks. If you have HAMR , you don't have that problem anymore. You can stay on 10 disks and go from 3 terabyte per disk, so 30 terabyte drive, to 4 terabyte per disk, 40 terabyte drive, 5 terabyte per disk, 50 terabyte drive. And we showed our analysts a very long-term roadmap showing up to 10 terabyte per disk, so 100 terabyte drive with only 10 disks and 20 heads.
Okay. And then help investors understand because you still talk about, you know, some cannibalization, you know, if you were to think about cannibalization, the arguments like, you know, the TCO is still very much more attractive using HDDs. And there's a price differential. And then if you include total cost of ownership, it's still much more attractive. How does HAMR change that equation for the TCO?
In terms of, if you just look at the cost per terabyte, as I said before, it's declining very rapidly because you don't need to significantly change your bill of material and you add a lot of terabyte per unit. Of course, the cost and the price is not the same. So of course, the price depends from supply and demand and how much demand is growing and how much supply is growing. So you can have differences between how you reduce your cost and how you price your products. But I would say, first of all, NAND is still much more expensive. Yeah. Second, also for NAND, there is a cost and there is a price. And you have seen in the past how even for NAND, the price can be very variable.
In different quarters or different years, that gap between hard disk and NAND can change. And then we have HAMR. So because HAMR is a beginning, the cost reduction from one product to the next is much bigger than what you can see in NAND, where the big change happened many years ago going from planar to vertical. And that was not 32 layers. And then they went to 64 layers. You still double the gigabyte per wafer, so the cost is much lower. But on third generation is 96. So you don't have the same impact. Right now, they are close to 200. So you can see how the change from one tech node to next tech node is actually giving less and less and less cost benefit. HAMR just started. So HAMR is at the beginning of the curve.
So actually, I see an increase of gap in terms of cost between NAND and hard disk for the next at least four or five years. But again, there is a difference between cost and cost curve and price.
Okay. All right. I'm going to ask the audience if they have any questions. If you do, please raise your hand. Anyone here? Okay. I'm going to continue here. CapEx, a little bit. I know. How should investors think about, you know, the mid 20%, I guess, Exabyte CAGR that you have? And as you're planning your transitions going from Mozaic 3, then Mozaic 4, then Mozaic 5, you know, how do we think about the CapEx as a percentage of revenues? I think you've kind of laid that out, but, you know, is it getting a little bit more capital intense? Is it getting less capital intense? And just giving you a free cash flow, if you can talk a little bit about CapEx.
Yes. Our model assumes a 4%-6% of revenue for our CapEx. So this industry is very low CapEx intensive. If you compare to, you know, we were discussing about the NAND before, we are talking about 30%. So the opportunity for this industry to generate very strong free cash flow is actually very high. And this is why Seagate and the industry in general is always generating a very good shareholder return for our investors. So we don't think this will change. We already went through all the transition from the PMR technology to the HAMR technology. So the R&D CapEx has been spent. A big part of the manufacturing CapEx has been spent.
So of course, every year you need to do some more CapEx, some different CapEx, depending where your bottlenecks are, what are the tools that you need to replace, what are the new tools that you want to buy. But still in that 4%-6%, I think is a very good model and does not add capacity on top of what we think the demand will be.
You're also becoming, so you have the gross margin improvements that you talked about, you know, both on the cost side, keeping pricing stable, but you also have becoming more efficient with OpEx.
Yes.
And that's kind of, you know, your target model that you guys laid out at your investor day. So just help investors think why that is, you know, why there are legs to that OpEx improvement.
Mainly, you know, is a big part of the reason for the decline in our OpEx compared to where we were a couple of years ago is the transition from the PMR technology to HAMR. So for many years, we had two groups, one working on PMR technology because we were still developing a lot of products in PMR technology and one working on HAMR. Now we don't develop any more PMR. So we have only one group doing HAMR. It's a little bit bigger group than what it was in the past because now we are developing product fairly quickly. But again, compared to the overall number of resources we had in the past, we were able to reduce. I think we have the right level of resources. So I don't see a major change in our headcount.
Of course, every year you have now the salary increase and you have no differences based on variable compensation, one year compared to another, depending on how we achieve our internal goals, et cetera. But for this quarter, we got it $290 million. This quarter has 14 weeks. So we said probably for the next two or three quarters, maybe we'll be around $280 million.
That's including the extra week then rolls off.
Yes.
Okay. All right. One of the questions I've been asking all our customers, all our corporates that are present here today, have you been using AI internally? And what kind of productivity efficiencies are you seeing internally as well?
Yeah, we use in manufacturing, especially for quality. But we use in many different parts of the organization. Even Investor Relations is using a version of AI for their research and analysis. We use in other parts of finance. We use in IT. So yeah, we use extensively for what is available today. I think in the future it will be much more.
Okay. All right. When it relates to capital allocation, you guys have, you know, obviously you have a dividend. You can maintain that dividend. And now you are looking at capital allocation. Just help us understand two things, two parts here. One, let's just say if you were to increase capacity, would this be and generate, put some of your free cash flow towards that that you're generating, would that be, you know, is that mothballed capacity? So relatively speaking, it would be lower intensity to bring it on. And then part two, assuming you're not increasing capacity, how should investors think about, you know, the debt levels, continued paying down debt, increasing buybacks, all that stuff?
So we are not thinking to increase capacity. And we don't have, let's say, capacity at this point, we don't have capacity available that we can put in place. So eventually we'll be new capacity. But as I said before, we are not thinking about it. That we achieved, you know, the first goal that was to be at about $5 billion. Maybe in the future we will reduce that even a little bit more. But no, at least we achieved that goal and actually we achieved it earlier than what we were planning. So we are starting share buyback in the current quarter. We are already active in the market with some share buyback. And dividend, as you know, we have protected our dividend during the down cycle. And usually during the up cycle, we increase a little bit our dividend.
We generally discuss with our board around the October meeting. So it's fairly close. So we will decide with Dave on what we want to ask the board to eventually approve.
Okay. As it relates to cash flow generation, anything on working capital? I know at one point you were running, you know, cash conversion cycles that you said were very elevated. So how should I think about anything on working capital as you're now generating profitability and your suppliers? Your suppliers probably are asking for.
No, I think we are good now. We had a couple of years, our fiscal year 2023, fiscal year 2024, where we were for sure stretching our working capital. Yeah. But during last year, that ended in June, we did the opposite. So we actually generated less free cash flow than what we usually generate with that level of profit just to, let's say, normalize our working capital. But I think that is done. So in this new fiscal year, we expect a much better free cash flow generation.
Okay. And then just as we wrap up here, Gianluca, just, you know, maybe remind investors, you know, is there anything about the story that's underappreciated that investors should keep an eye out for?
I think, you know, we had a very good analyst day. So I hope we were able to explain our business to our investors. But our oldest new investors that maybe are less familiar with the story, I would say, you always need to look at the past, but always consider that the business is different. It's very consolidated. Demand is very strong. The capacity that was created in the past for client is now fully absorbed into the cloud space and demand is above. So we're in a very different situation. And we want to keep this situation. And I think, as I said before, AI is one application. I'll say the application of the present, application of the future are some that we don't know yet about, but will generate even more data and the need for data storage.
I would say it's always good to look at the past, but I think the future will be better, even if, as I said before, we expect this business to grow, but to have some cycles during this growth.
Okay. All right. Thank you very much.
Thank you.
That wraps it up here, and good luck with the rest of your meetings here.
Thank you very much.