Good morning, everyone. My name is Krish Sankar. I'm the TD Cowen semi equipment hardware storage analyst, and we're fortunate enough to have Seagate, one of the leader in hard drive, and it's Gianluca Romano, the CFO. Thank you very much, Gianluca. I really appreciate your time. I do have a few questions, and if anyone has any questions, feel free to raise your hand. So maybe let me start out, you know, kind of if you look over hard drives over the last couple of years, we kind of went through a cyclical downturn. Things seemed to have bottomed in the September quarter of last year, and it's slowly improving. And during the time also, I think Seagate has taken some additional steps in improving your cost structure, taking some capacity offline.
So can you just talk about how things are trending since that cyclical bottom? And, like, how to think about pricing from here going forward, capacity expansion, eventually, what that means to margins.
Perfect. Thank you. And before I start, I want to know—I want you to know that I will be making forward-looking statement today, and you can learn more about the risk associated with those statements on our website. Yeah, so we bottom, our industry bottom in the September quarter, start to improve December, March, June, now as we guided. The improvement is coming mainly from the cloud segment, that is, probably the most important segment we have. But I expect a second step in the recovery coming from the other segments, enterprise, OEM, video and image application. In general, China, that is still, lagging the recovery. So I see this, upcycle that now started after the September quarter, to possibly have now different steps. And we just went—we are just going through the first step, that is a cloud recovery.
The cost structure is now is now very different than what it was two years ago. Through the down cycle, of course, we took the opportunity to change a lot of things in our manufacturing, in our R&D, in our OpEx in general. And this is why now, couple of quarters ago, we started to say: "You know, we can achieve a much higher gross margin, and then operating margin, at actually a lower level of revenue, compared to where we were two, two years ago." And we are seeing it right now. Now, we already improved a lot our gross margin. Yeah, I don't know if it's working.
In the last few quarters, we have already discussed our estimate of what will happen in the future quarters in terms of continuing increasing revenue and continuing improvement in the profitability. Pricing is part of that improvement in profitability. We started to raise price about a year ago, despite now we were possibly as a worse part of the down cycle. The industry is very consolidated. I think the industry also has a strategy and a need to improve profitability, and therefore now was not only Seagate, but I think the entire industry decide to focus more on improving the bottom line than fighting for market share, and this adds a lot of benefit to, I would say, you know, the entire industry, not only to us.
Got it. That's interesting because, you know, if you look at many cyclical industries, you know, you were very disciplined in cutting the cost on the way down. The worry is that when things start rebounding, you know, maybe on the COGS side, capacity starts coming back online, on the OpEx side, variable cost starts increasing. So how do you think about that on the way up? How do you manage that?
I'll say we have now two different strategies. Now, one is in the manufacturing cost, and one is on the OpEx. On the manufacturing, during the down cycle, we reduced a lot our headcount in both direct and indirect labor. The direct labor, we will have to increase again. Now, it's very related to the volume that you produce. So that part will come back, but we will not need to rehire the indirect labor that now we took out from the company, so that is a permanent benefit to the company. OpEx, we focused a lot, especially on the R&D side. In the past, we had to develop two different technologies, now what we call PMR and HAMR. Starting about a year ago, we stopped doing any development on PMR.
We just focus on HAMR, and therefore, we had the opportunity to reduce the team. Now, instead of having two teams, now we just need one team. Maybe little bit bigger team, but just one team. So even for the R&D side, we don't expect to increase headcount-
Mm.
... in the future. Now, we are now going through this qualification of the first big product on HAMR, and we think all our future product will be HAMR-based.
Mm.
We are qualifying, and we actually have qualified with many customers, our last product in the PMR technology, that is a 28 TB SMR, 24 TB CMR. We don't think we will need any new PMR-based product, so all our development will be on HAMR.
... Got it. And before I get into HAMR, I kind of had, like, a few more questions on, you know, the non-HAMR, the core business. You mentioned how, you know, large part of the cyclical upturn has so far, has been driven by cloud. When do you expect the enterprise segment to start picking up, and the VR segment, video imaging for, you know, which is largely China?
Yeah.
When do you expect that to start recovering?
So the video and imaging application is improving already a little bit, so June will be better than March, at least this is our current expectation. So we already start to see a little bit of an upside. We think, September and December will be a sequential improvement. So I would say through the, you know, end of the calendar year, now, I think, especially if the overall China business is recovering, and if the, you know, the government support is really going through, the local projects and the local companies, then, of course, they will, they will focus more on the storage part, of their business. So I think, you know, through the end of the calendar year, we could see both the video and imaging application and the enterprise OEM starting to do what the cloud is already doing.
Mm.
Just with, probably, you know, couple of quarters of delay of lag.
You know, historically on the nearline side, the split used to be roughly 60% cloud, 40% enterprise. Do you think it's going to be any different this cycle?
Mm
... or do you think it's probably still the same ratio?
Well, if you look at now, two years ago, before the down cycle, it was probably revenue was still probably close to 50/50.
Mm.
Then the down cycle, of course, depending from the quarter, everything was very different. Right now, with the cloud recovering, we see, you know, of course, a bigger cloud compared to enterprise OEM. Right now, we are probably in that 60/40. Cloud will continue to grow faster than enterprise OEM. Enterprise, of course, will grow also, but between the two segment, cloud is expected to grow faster. So probably in another year, you will see, you know, maybe a 65, and with the time will go to 70.
Mm.
Not because enterprise OEM is not going to grow, but because cloud is expected to grow much faster.
And on the-
Now, AI could be a little bit of a variable, right? Because AI is not clear if it will impact only the cloud or a lot of the on-prem data center.
Right.
So we could see a period of time where also the on-prem data center are growing faster than they usually grow. So that ratio could change a little bit, but we still expect cloud to grow faster.
Yeah, that was going to be my next question because it's kind of interesting. When you look at many of the cloud spending, many of it is going for AI applications versus general purpose HPC, at least, so far or as of now. But do you have the... You probably don't have the level of visibility or do you have that, where your customers tell you what workloads these are being used for?
Well, in general, they don't need to tell us, because they don't need a special hard disk. No, they buy a normal storage hard disk, and then they utilize in their data center, and possibly they utilize for different application. Some can be AI, some can be non-AI. Of course, now, when you see this improvement from the cloud business, because their business is improving because of AI, we expect that a part of this improvement is actually coming from AI. We had a lot of conversation with customers, so of course, they are telling us, now, their storage is becoming more and more a focus for them, because of AI. They are very focused also on other components, but they still need to build the infrastructure, and that is good to us.
Now, they need to build the infrastructure so they can develop the AI application, and they can store the data from AI into our hard disk. So it's a matter of time. I think the time is coming. It's probably already right now happening, but will be even bigger in the next few quarters.
Got it. And then maybe switching on gears a little bit to HAMR. You know, in the last call, you kind of mentioned about how you expect to be qualified with one large hyperscaler, and start shipping, like, probably a few hundred thousand units by June time, kind of, timeframe. So curious how that call is going on. Is that timeline still the same?
So the call is still ongoing. All the data we have until now demonstrate that the component issue that we had in March has been solved, so that will not be, you know, in our way anymore. During this time, we also has the opportunity to improve many of our manufacturing processes, and we also changed firmware of the drive. So this will have a positive impact to the product in term of yield for us and product stability for the customer. So to me, those are all good improvement and good news. Of course, they have to be qualified also, so possibly the qualification will extend into the September quarter and, and not be finalizing in this quarter.
So because my understanding was that, you know, if you're going to be shipping a few hundred thousand units, it means you are already qualified. In the past, you said by June, so that might happen more in calendar Q3, in the September quarter.
... So we are, well, whatever we don't ship on HAMR this quarter, we will ship on PMR drive, so we don't expect any change to our revenue forecast. Actually, pricing is also a little bit better than what we expect, so possibly, you know, EPS could be better, and we will be more precise later in the quarter. But I don't see any negative impact from this change. We have volume in PMR drive that we can use to replace any shortage on HAMR.
Got it. Just on that point, because in April, I think you'd already identified the component issue. You went to the other supplier to, like, supply this big customer, and at the time, the expectation was you'll be qualified by June. Now, it's kind of pushed out to, by three more months. What is the reason for? Is it just because getting the component took a while, or, like, the qual timeframe is getting longer?
I would say the component was, of course, the reason for the delay in March. Now, through that time, we found opportunity to improve our own processes, manufacturing processes, and working with a partner, we also decide to change our firmware. So are all positive changes, are all things that will reduce our cost, will increase the stability of the product, so better quality. Longer term, this means better-
Mm
...gross margin, better result for us. The only negative is that instead of, you know, qualifying in June, will take a little bit longer. But this is a long game.
Mm-hmm.
2 or 3 months is not changing anything. We have so much advantage on this technology, that the important is that we have the best product with the best quality, and also the best cost for us, that means better yield.
Gotcha. Is it kind of like a Catch-22 situation? Because, you know, that customer's demand can be met by PMR because HAMR is unqualified, and the longer this goes, does the HAMR qual gets pushed out and the opportunity, the time opportunity for HAMR comes down because PMR is meeting the demand?
I would say, in terms of, you know, units, no PMR can meet the demand until a certain level, then I think the industry is a bit supply constrained. But possibly not this quarter.
Mm.
But in the next, you know, few quarters, we could be in a situation where, in the cloud space, a very high capacity, you know, the industry could be constrained. But what PMR cannot do is to provide the same benefit of a much higher capacity drive.
Right.
So as soon as the drive is ready for this customer and other customers, of course, they, you know, they would all want to have that product. We are very confident, actually, in the product, and actually, we decide to start the vast majority of the cloud customer qualification in the September quarter. So we will not wait any longer, so next quarter, we start qualifying almost all the CSP in U.S. and outside the U.S.
Mm.
So we think the product is actually very ready.
Is there a chance, you know, the current customer who has been kind of your, hopefully, the first one to qualify, is still, you think, the first to qualify?
Yes.
You think another hyperscaler could qualify-
No, no, they will be first for sure because they are through the end of the qual. The other guys will start the qual in September. Based on this experience, I would say the qual could take about three quarters, for the new CSPs that are starting the qual in September quarter. So the big, I think the big volume ramp will be, let's say, mid of 2025.
Gotcha. And that assumes you already have a few cloud customers qualified, and at that point, are we talking about, like, a couple of million units a quarter run rate, kind of a thing, or...?
Well, it will be a, you know, a material volume. It's a little bit far in time, so we don't communicate exactly what we-
Got ya
... expect as a volume, but will be material volume.
And also, I think Gianluca mentioned in the past about how HAMR is margin accretive, gross margin accretive. And let's say, you know, you are qualified in multiple hyperscalers, couple of years down the line, I think one of the questions that many investors have been asking is that, you know, if you do the math, you could probably get to a 40%+ gross margin. It feels like so far you don't want to underwrite that number, but you're still sticking with the low 30% gross margin target.
Well, we are staying with that target only because the volume of HAMR is not here yet, so we need to wait a little bit longer. But, of course, we always said HAMR will be accretive to our gross margin. Is a very good product. As I, you know, explained a few times, the bill of material will remain the same from 3 TB per disk, so a 30 TB drive, to 4 TB per disk, to 5 TB per disk. This is a big benefit of HAMR. You don't need to increase the bill of material. You don't need to have more media, more heads.
Right.
It's exactly the same bill of material. Capacity is going up, but increasing capacity will have, of course, a price, and will bring product after product, a better gross margin. But as I, you know, as I always said, even at the beginning, with a 30-terabyte drive, you know, is significantly higher than a PMR drive.
Mm.
'Cause you compare the 30 terabyte CMR HAMR drive to a 24 PMR drive, and if you want the SMR version, today we have a 28 SMR, HAMR will be 34. It's still a huge difference, so the interest of the customers will be very high. We think we can price that well, especially when you start to have, now, three, four, five more customers, and we expect positive improvement in the gross margin.
Got it. We just pause for a second to see if there's any questions from audience. But let me follow on, you know, obviously, your competition, Western Digital, is probably behind in HAMR. You know, they probably believe their ePMR can extend up to, like, probably like 38 TB. I understand that, you know, as they scale up, I mean, it might be margin-negative margins because of adding more disks. But the flip side is that, you know, I'm sure the hyperscalers don't want to give extraordinary amount of market share to you, right? They probably want to maintain the duopoly balance. So how do you see it playing out?
Where do you think HAMR adoption gets artificially curbed because there is no second source, or do you think that, you know, there is a 1-2-year opportunity set for you in terms of share and margins before the competition catches up?
I, I think you're right. No, this is not a competition between us and our main competitor. No, the industry is already very consolidated. There is not enough supply for us to supply the full volume, and as you said, customer will not want to go single source. And now, we said several times in the last, you know, three or four years, market share is not our focus. No, the industry needed something different a few years ago that was more profitability-
Mm.
and less fighting on the last 2%-3% of market share. And now, we have done that consistently in the last several years, even during the down cycle, and even now during the recovery period, and we will do the same later with HAMR. Now, we will have a certain volume, that now I think will be very interesting for our customers, but our focus is not increasing the market share. It's taking the right profit from this product that needed many years to be developed.
Mm.
And as you said, at a certain point, possibly, no competition will have a similar drive. So even put in place a lot of capacity, and then having, you know, a another company or maybe two other companies having a similar technology, then we create again that additional supply, that now will create oversupply and negative input to pricing and profitability. So that's not our objective. Our objective is to be very reasonable with the capacity we put in place, and of course, sell all the volume that we will produce on HAMR, but not not focusing too much on market share.
Got it. And the reason I ask is also, you know, there's, I think, recently, news reports about Toshiba-
Mm
... shipping some HAMR sample with TDK heads. So I'm kind of curious, you know, they were always, like, the marginal player, at 10%-12% share. So if it's true, what kind of a lead do you have in terms of timing, or years, or months, or whatever it is on HAMR, which is even like a Toshiba?
Yeah. Well, I would say, it's very difficult for a company that is not vertically integrated in terms of producing heads and media to really develop HAMR. Now, there is a lot of work that you have to do to put the heads and the media together, especially when you talk about mass production volume. Now, you can put together 1 drive, but try to put together millions of drives in manufacturing is very complicated. So, they said they will have some samples next year. I would say we had samples probably 2 or 3 years ago.
Mm.
So the delay is huge. We are not sure, but they finally will be able to, you know, ramp high volume production of HAMR. I would say, you know, we still consider maybe WD as being the main competitor on HAMR, not Toshiba.
Got it. And, you know, when you look at, like, data centers, like obviously AI data center, power is a big, you know, question, challenge, consumption, et cetera. I understand hard drives are, like, more like 7-9 watts consumption. It's not a big met number, but if you go to HAMR, since you're adding a laser, does that additional component consume more power or it's not-
No
... or it's de minimis?
Actually, the consumption per unit is very similar, and the consumption per terabyte is actually lower-
Mm
... because you have many more terabyte per unit. So it's actually good news for, for our cloud customers, 'cause for the same location, now they will have more terabyte and a lower consumption per terabyte.
Got it. Got it. And then, I had one more question on HAMR before I jump into the ASIC part of it. You spoken about, you know, like, 3 terabytes of disk going to 4, and eventually even 5 terabytes a disk. Does that increase the component, i.e., COGS, associated with it, or is it more R&D intensive, that it's more an engineering thing, versus incremental component additions?
Well, when you go from 3 to 4 to 5 TB per disk, mainly is areal density. So it's all the work that our R&D team is doing every day. You know, we already have the 3 TB per disk. We're already working on the 4 TB per disk. We actually said in the last earnings release that, you know, we expect to have the product ready by the end of calendar 2025.
Mm.
So, it's not far in time... and in our lab, we already have the 5 TB per disk. So it's not really a bill of material, it's not really a manufacturing process, it's really, you know, an R&D development, how to store more and more data on the same disk. So really, an areal density work. But in terms of cost per unit, we don't see any major change between 3 TB per disk, 4 TB per disk, 5 TB per disk.
Yeah.
Where's the big opportunity, I think for us, and for the industry, you know, you have a unit cost that is the same, but you have 10 TB more per unit-
Mm.
Those 10 TB, of course, will have a price, and that is all gross margin.
Got it. And then, you know, I think last quarter you mentioned,
Yeah
... about the $600 million ASIC facility sale to Broadcom in Singapore. But my understanding is, you still buy the ASIC controllers from Broadcom, Marvell, and others. I think one of the concerns some investors have is like, you know, how much pricing negotiation power do you have if Hock raises the price on you two, three, four years down the road? So is there any color you can give on, kind of, your guardrails against that?
Well, first of all, I would say we have a great partnership with that vendor, as we have a good partnership with other, other vendors. We already had a supply agreement in place, now, with this specific partner. Now, we have a longer supply agreement in place with the partner. Of course, in the supply agreement, there is a certain volume, there is a certain price, so I... No, we will don't, we don't expect any surprise on that side. I think, now our partner saw the opportunity to take a very good team, that is designing ASIC for the hard disk business, and can design ASICs also for other businesses. So they can use the team for, now, other projects, and of course, the IPs that, that, team has developed in, now, the last several years.
Of course, I think there was also a big interest in having this long-term supply agreement and developing our new SoC for the future.
Got it. Got it. Then maybe a final question from me, Gianluca, is, you know, your friends at Western Digital are spinning off the NAND business. So now you're gonna have two pure play, you know, hard drive businesses. You know, is that like a not even for you? How do you think about that? Like, does it change the dynamics in any way, or?
I don't think it will change a lot. Of course, depend from now the strategy of the people that will manage this business. But assuming that is consistent with what is going on today, I don't think there is any change, because today they have two business units, so they have basically two different ways to manage the two businesses, and now, possibly, the hard disk will remain for the vast majority with the same people they have today, with the same processes, with the same locations. So I don't really expect a lot of change, but will be probably good to have a competitor that is really only focused on hard disk-
Got it
... and with the same objective and the same needs that we have as an industry, a pure hard disk industry.
Got it. All right. You have a quick question? Okay, go ahead.
It seems that the biggest variable for having those margins is manufacturing yields. I wonder if the delay of, that qualification should also delay the confidence in achieving the right yields for corresponding-
I would say no. I would say the opposite. Now, we had this opportunity to continue to work on our processes and find opportunities to improve the internal yield. And of course, as I said before, we also had to modify the firmware. But actually this will be a positive to our gross margin in the future, and we are very confident right now in where the product is, and so we have decided to start the qualification with almost all the CSP in U.S. and outside the U.S. in the next quarter. So despite this little delay, that of course was not welcome, but at least we took out the best from this time.
All right. Thank you very much, Gianluca.
Thank you very much.
That's super helpful. Thank you.
Thank you.
Thanks a lot.