Good afternoon. Hi, I'm Tim Arcuri. I'm the Semi and Semi Equipment Analyst here at UBS. Very pleased to have Seagate. Next, we have Gianluca Romano, who is the CFO at Seagate. Seagate might be the best-performing stock that I cover this year. Things have been going, obviously, very well for you. Let's just start off on the point of, obviously, supply-demand is very tight, and you've maintained your strategy of not expanding unit capacity. Can you just talk about what's driving this supply-demand tightness?
Yeah. Thank you, Tim. 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 these statements on our website. Supply-demand now is very important, of course, in every industry. It's now more than two years that demand is above supply. For us, and I think for the industry, the important thing is to increase exabytes, not to increase units. We think in the longer term, the way to meet demand, or at least get closer to the real demand, is moving our customers into our highest capacity drives. With the HAMR technology, we can go from 30 TB- 40 TB, that is, in quote today, to 50 TB and beyond. We don't think it's needed to increase the units.
I think the best way is to generate what our customers need, that is, more exabytes, through our product roadmap. I think the industry is fairly well aligned on this.
Is there a point, though, that your customers force the issue? Do they come in and they prepay for you to expand capacity? Have you been offered any prepayments from your customers? If you were offered prepayments, would you even take them at this point?
Actually, we do not really need our customers to pay for capacity. If one day we will decide to have more capacity in terms of units, we will pay for that ourselves and not have any link or any constraint from our customer on how we use our manufacturing. As I said before, we do not see the need right now to increase the units.
Is there a—what's the end game? Is there a point at which gross margin gets to X, 45, 50, maybe even? Is there a point where you feel like you're sort of earning your portion of the profit pool of the ecosystem where you then say, "Okay, we'll start to add supply"?
No. We focused on improving our profitability more than two years ago. I would say 10 quarters ago. We started to implement it this quarter, and we started to implement a certain pricing strategy that I think is not disruptive to our customers. It is very consistent, but it is not increasing pricing in a way that is surprising or in a way that is disruptive to our customers. That has worked very well for our profitability. We improved our gross margin by—we doubled, actually, our gross margin in less than 10 quarters. We do not have a specific target. I think the situation is today very similar to what it was 10 quarters ago. There is no reason to change our strategy. I think we need to continue to be consistent, to move our customers to higher capacity drives.
Moving the mix up is where we improve our profitability the most because with a fairly consistent pricing, we can take all the reduction in cost per terabyte and use that as an improvement for our gross margin and operating margin.
Got it. Now, this question might be hard for you to answer, but is there a way to assess how much you're under shipping demand? I guess I ask the question because you're booked out through 2026. You're booking into 2027. I guess by definition, because you're booked all the way out through next year, if you were shipping to demand, then your revenue would be 2X what it currently is. What is the sort of—if the supply chain was totally unconstrained and if you were producing as many drives as what everybody wanted, how much higher would your revenue be?
It would be significantly higher. I said a few times today in other meetings, the gap between supply and demand in the last six months has actually grown. I do not think hard disk is the main constraint into data center build-up. There are other parts and other components in the data center that are probably the main constraint. We are fairly high in the list, and it is nice to be there, but we are not the blocking point. I think with our product roadmap and with our customer qualifying fairly quickly our highest capacity drives, that will generate enough exabytes to supply what is really needed in the short term. Of course, it is not covering extra inventory or safety stock, but it is covering the data centers that have been built and are being built today.
I have no evidence of any data centers that are built and there's not enough disk drives installed. I think this is a very good situation for us. We have this gap between supply and demand, but we are not the blocking point in building new data centers.
What do you think it is?
I think today is power, but in the future, it can be other components. In the past, some semiconductor components were the blocking point, and right now, it's probably power.
Let's talk about demand. There have been some emerging—I mean, obviously, AI is now trickling over and is driving an increase in demand for hard drives. Can you talk about some of the applications that are AI-driven that are driving demand for Seagate?
Yes. I would say AI is one of the applications that is driving storage. It's not the only one. For sure, in the last few quarters, we have seen a fairly huge increase in storage consumption from AI. Part of that, but I think it's still the beginning, is the video part. AI video is starting to be used, and it's starting to consume a fairly important part of the storage. In the past, we have seen the new application to be adopted maybe a little bit slower than what people were thinking. I think probably even AI in the last two years was not adopted as fast as some people could think. Now, we see an acceleration. The AI infrastructure is getting bigger.
The AI applications are starting to generate more and more data, and video could be a major drive for more storage needs in the next several years.
How do you prevent—we hear some examples of customers migrating what maybe might have been HDD over to SSD? I know there's this push-pull between what's going on in SSD and what's going on in HDD. Can you just talk about that? I know that you don't view it as really competing for the same workload. Maybe you can talk about that.
No, I don't think—as I said before, I don't think there is any data center that is short in hard disk today and that pushes our customers to spend six times or eight times more to buy the storage that they need for that specific data center. I think it's more a theoretical solution in case hard disk becomes the blocking point for new data centers. In that case, maybe some customers could decide to spend a lot more money to buy different components to do storage, a component that is not used for storage today in the data center, but is used to run the application. It could be used for storage, but the cost is so much higher that we have not seen it happening, and I don't see it happening.
As I said before, I think all the data centers that are being built will have enough hard disk to cover the storage part.
You think that to the extent there are some examples where SSD lead times are shorter, and so they're just buying just in case? They're maybe buying SSDs just in case?
No, I think NAND and DRAM and other components are growing in the data center because they are used to run the application. As I said before, many applications, including AI, are growing. They need more DRAM and more NAND. Because they generate data, they need more storage. The storage is on hard disk.
Got it. Just the point about—the idea that your lead times are 52+ weeks now, and these are very, very wealthy companies, the wealthiest companies in the world. If they need—they're going to make sure they get what they need. How do you prevent them from double-ordering? Maybe double-ordering is not even the right word because they would just place orders out farther. It is not like you could ship anything else today anyway because you do not have the capacity. How do you prevent that from happening? Would you say, "Fine, let them double-order. All that does is just fill our backlog out farther. If they do not take it at that point, then we'll kind of deal with it then"?
Yeah, that's the right point. We push demand to the future. Today, not serving the full demand, we are just pushing part of the demand to the future. This works well until eventually hard disk becomes the bottleneck of a build-up of data centers. As I said before, we have not seen this happening. We are just pushing this as an industry, I think. We are pushing this demand out in time. As every technology industry, I think we will have cycles even in the future. Pushing this demand to the future will at least decrease the impact of the down cycle when and if it will happen. It works well for us and for the industry.
I think it will work well also for our customers because until this industry is providing enough exabytes for what they need today and tomorrow, they will not have problems on storage.
Great. I know you do not give us orders. You do not give us book to bill. I guess maybe to get an idea of how the trajectory of bookings are, are bookings volatile quarter to quarter now, or have they continued to get better? If you did give us a bookings number, it would be up every quarter. Or do they come in lumps?
No, of course, there are different negotiations with different customers in different quarters. We already said for calendar 2026, we have already basically allocated all our nearline capacity to our top customers. When we go longer, for us, it is less important because what we want to have is orders covering products that we start in our manufacturing. Because the lead time is about three quarters, especially for a HAMR drive, we want to have orders in place so that we have a certain mix with a certain price, a certain time to deliver. We have covered more than three quarters. We have covered four or five quarters already. When we go longer, the reason why we go longer is only because customers are asking.
They ask to have a certain confidence on a certain level of exabytes that we will allocate to them even after the next four quarters, for calendar 2027 and maybe calendar 2028. Those are different kinds of agreements. They are not firm orders like we have for calendar 2026. It is basically an agreement on how many exabytes we will allocate to them in the longer term. When we get closer in time, based on what they are qualified, which products they are qualified, we will define the product, the volume for that product, and the price.
Great. Let's talk about the technology roadmap. Obviously, we'll talk about HAMR. It sounds like five CSPs are qualified and three more are being qualified in the first half of next year. It seems like the velocity of these qualifications is actually picking up. Maybe is there some seeing the customer that went first, seeing them go first now, has that sort of greased the skids to now say, "Well, they've qualified it, and now we all want to qualify it too"? Can you also speak to sort of when we should expect the exabyte crossover to be for HAMR?
In general, customers care about exabytes. To get more exabytes, they need to qualify drives that have higher capacity per unit. This is why they want to qualify HAMR drives because they are the ones with the highest capacity per unit. We needed a little bit of time to qualify the first customer in the cloud space. After that, when we found the right configuration for the first customer, every other customer went very fast. It actually went faster than what we were expecting, which is why after our investor day, we gave a certain model, and we have a little bit outperformed that model until now, mainly because we were able to move more customers quickly to HAMR and move the mix up. We have now five. We have two or three more.
Two of the five are already qualifying the second generation HAMR, which is a 40 TB drive. They go as fast as they can. We did not have any issue on qualification, except that first delay on the first drive with the first customer.
Relative to crossover, I believe you said that crossover will happen in the back half of next year, second half of calendar next year?
Yeah, we said by the end of this fiscal year, so by June, 40% of our exabytes will be sold with HAMR product. And four quarters later, so at the end of our fiscal 2027, 70% of the volume will be sold with HAMR products.
Why would it take, if everyone's qualifying HAMR, I would think that the crossover would happen faster. Can you just talk about that?
I would say first we qualify, then we take the order, and then we have to ramp and sell. It is maybe a little bit different than what was happening in the past. In the past, when we had a new product, we were ramping and then trying to find home for those products. Now we are in a very different situation. We want to be sure that we have a customer qualified before we dedicate manufacturing to a product that otherwise remained unsold. Of course, demand is so strong today that you do not want to have manufacturing allocated to something that you need to wait another two or three months before you can sell it. We go a little bit slower, but we are very reliable, and we are very consistent. We need to qualify more customers and then ramp.
I think going to 40% of our capacity in basically three, four quarters and then going to 70% is a good ramp. Of course, if we can go faster, we go faster. I think it's a good model.
Let's talk about gross margin for a moment. Gross margin has expanded four to five hundred basis points in the last calendar year. How much of that's from better pricing versus better utilization versus other factors?
I'll say mix is a very important factor to improve our gross margin. Of course, if you look, our gross margin 10 quarters ago was probably half what it is today. Part of that improvement at the beginning was the underutilization charges that started to be absorbed. After that, between the pricing that has been consistent for 10 more quarters and the mix, moving customers to the higher capacity drives where we have the lowest cost is actually the main factor to improve our gross margin. That's been for more than two years now.
Do you think I remember a conversation we had maybe on a bus trip or something. We had a conversation about how high you can push gross margin. I think your comment was, "Well, there are a lot of companies in the hardware supply chain that in the continuum of all these companies, you have very, very low margins." I think your point was there is no reason why someone who sells that's mission-critical on the product as we sell, no reason why gross margin cannot be much higher. I think you were not saying 50, but I sort of read that there is no reason why if you look in the continuum of who captures the value, there is no reason why your gross margins could not be 50. Can you kind of talk about that?
Do you think is there a point where the customer begins to push back and say, "Listen, I can't let you have gross margins this high?
First of all, I don't think our customers are looking particularly at our margin. I think they look at the return that they get from our product. I think they get a very high return. We are very consistent in the pricing strategy, but there's no reason to change it right now. Actually, I think if something eventually happens with gap between supply and demand, it's a little bit bigger, not smaller. I don't see any reason why we should change. Very importantly, we are just at the beginning of our second generation HAMR. This is a product that will reduce our cost per terabyte much more than the first generation HAMR. That will be another boost to our profitability. We don't have a specific target.
I don't think our customers are looking at a specific limit for us and then starting to push back more than what they do today. I think we will be consistent. I think for the next two or three years where we have visibility, it looks no different than what has been in the last two or three years.
If we kind of play this out and if and when things do reach a peak and we look back and we say, "Oh, that was the first sign of you as a company. That was the first sign we saw that happen." In retrospect, that was the thing that should have made us concerned. Would you expect it to be the demand in the longer term would begin to fade? Because you're booked out in the near term. You'd think that the stuff booking into 2027 would soften first if we did see demand soften. That's where it would soften first. Is that sort of how you think about you have people looking at this and what's the first sign that you're looking at every week to make sure things are healthy?
Yeah, I think you're right. I think when we start looking at new orders coming at a lower volume than the current orders, that's probably a sign that already there's a slowdown in data center build-up or they have built some inventory so that they needed to be less for a while. Another sign is when we can build a little bit more exabyte in a quarter and we put that volume in the market, today is going to be sold very quickly, even if the price is actually higher than the normal orders. If we don't see that volume being purchased quickly, that's another possible sign. We monitor those. Of course, we talk a lot with our customers also. We are not even close to that situation today, but there are two very important metrics that we need to monitor.
Let's actually talk about cash for a moment. You're generating a lot of cash. Let's talk about capital return. How should we think about capital return as you pay down debt? Because you're essentially almost at the point where you want to get debt to. Can we see a scenario where you ramp up share repo? What are the metrics there?
Yeah, we will. I would say we have reduced the debt from more than $6 billion- $4.5 billion at this point. Maybe we can reduce it a little bit more, but probably not much more. We have addressed part of the convertible in the current quarter. Maybe we will address a little bit more in the next few quarters. Generally, we have always been very focused on shareholder return. We have increased our dividend again in the last board meeting in October. We have started share buyback in the September quarter. I think in the next few quarters, you will see a higher level of share buyback.
Great. Let's talk about demand growth. I believe demand growth somewhere in the mid-20s CAGR over the next three years is sort of what you've said. In the near term, though, it seems like it's higher than that. Is that the right way you're thinking about demand growth, CAGR?
Yeah. When we gave the model, it was not for a few quarters. It was actually for three, four years. So that 25% exabyte CAGR, you need to look on a longer period of time. In the short term, we have done better in terms of volume, in terms of revenue, in terms of profitability. I think the model is a good model. I think it is a model that will show a lot of improvement in the company, top line and bottom line. Of course, every quarter, we try to do as well as we can and to optimize all our manufacturing, all our opportunity to improve profitability. We have done better than what we were expecting for a few quarters, and we will continue to do it.
Great. Let's go back to the HDD versus SSD question. Obviously, NAND pricing is going up a lot. The gap, as you can push to higher capacity drives, you would think that the gap probably does not close a whole lot, if at all, given that NAND prices are going up so much. Can you just talk about that? Where does the TCO gap stand today? What is your outlook as you bring on HAMR? Things seem to be going the opposite direction. You are scaling density and NAND prices are going up.
That's correct. I think the gap between the NAND cost and the hard disk drive cost is actually going to increase in several quarters and years. HAMR is just at the beginning. Usually, the cost decline that you get from a new technology is more relevant at the beginning. When you go from 30 TB- 40 TB, you increase your capacity per unit by more than 30%. Then 40 to 50, you have another 25%. Then start to maybe decline a little bit. The cost improvement that you get at the beginning of a new technology is huge. Like NAND, when they move from planar to vertical was a huge decline. With the time, tend to decline. NAND is still declining in term of cost, possibly not in term of pricing, but that is very variable.
Hard disk, at least for us, we have HAMR. We will have a major cost decline right now and for the next several years, especially when we go through the first two, three, four generations of HAMR.
How do you plan to free up more capacity for HAMR? Is there a scenario where you'll start to buy heads from TDK? What is the biggest constraint right now on capacity? Is it heads?
That is a little bit of a problem, right? Because when you move from the old technology to the new one, the cycle time is longer. Cycle time for the heads, cycle time for final test. To build the same number of units is not so easy. We need to improve our efficiency internally in order to try to keep at least the same number of units so that the 25% improvement that we can get from the technology transition will actually result in a full 25% and does not get declined because the units start to go down. We have a fairly, you said we are not very aggressive, but I think we are fairly aggressive around HAMR. We are the only one producing HAMR heads, so it is not that we can use someone else to help us in getting more heads.
We can have eventually, if there is capacity available, but I doubt today we do not buy any external heads. If in the future there are some heads available that we can use to keep the units where they are, so not to decline the number of units because of the longer cycle time, we can, of course, use it. I think we do not talk in specific customers or suppliers, but I think TDK in general is supporting another producer of hard disk. I think they consume the majority of the heads. If they do not for any reason, I will be happy to take some of those heads in the future just to keep the units where they are and reducing the impact of the longer cycle time of HAMR.
Again, eventually, it's very short term because when we move to HAMR, we don't have any opportunity to go outside.
Sure. Maybe just last question. There's been this trend on these conference calls to ask companies, "What's your portion of if there's a gigawatt announced, how much do you get from that gigawatt?" It all began with Jensen saying that he gets, ultimately now, between $35 billion and $40 billion per gigawatt. Not on a per gigawatt basis, but on a customer CapEx basis. When a data center customer spends $1 billion, is there a rough way to think about how much of that goes to HDD?
Statistically and looking historically, I would say we are mid-single digit of the CapEx in our disk industry. Every quarter is different. If you look at the longer term, mid to high single digit is probably where the hard disk is.
Great. We have run out of time, but thank you, Gianluca .
Thank you very much.
Thank you.