Western Digital Corporation ($WDC)

Hello, everybody. Thank you for joining Tom O'Malley here, just kidding. Tim Long sitting in for Tom O'Malley. I'm the IT hardware, com equipment analyst at Barclays. Tom couldn't make it today trip from his team is here. Very happy to be hosting Western Digital. Kris CFO, thanks for coming with us. I have covered the stock for a fair amount of time in the past. Tom covers it now. So would be good to catch up. Obviously, a lot of good things going on with the stock in the industry. So before we get into it, Kris, I think you got to start off with the safe harbor.

Yes. And thanks for hosting us here, Tim, at this great conference. We have a lot of good meetings already. But before we start my compliance department asked me to make sure that today, we will be making some forward-looking statements based on our current assumptions and expectations as it relates to our product portfolio, business plans and performance, market trends and future financial results. Those forward-looking statements are subject to risks and uncertainties. So please go to our SEC reporting, Form 10-K and other filings for more information about the risks and uncertainties that could cause actual results to differ materially from expectations. We also will be talking about some non-GAAP measures, and you can find a reconciliation between GAAP and non-GAAP on our investor website, investor.com or investor.com.

Okay. Great. And I will give an opportunity with about 10 minutes left for folks in the audience to ask a question. So first, just maybe just mark-to-market here a little bit. This year, obviously, there's been a lot of success the last few years with your business. We're starting to see AI industry transferring from a little bit more of a training based market to inferencing. Could you talk about that transition, how that relates to WDC and HDDs? And do you see that as a positive? Or how do you see that playing into the business? .

Yes. Obviously, we are very well positioned as a strategically focused hard disk drive company in this data economy, right? It hasn't been that long in February of 2025 just after the separation of our flash business the management team was indicating that we expect the exabyte growth to be mid-teens year-over-year CAGR for the next 3 to 5 years with a potential upside if and when AI kicks in. While in February of 2026 at our Innovation Day, we've actually already started seeing AI kicking in, and we were comfortable indicating that we now expect or in February, expect the exabyte growth to be 25% CAGR over the next 3 to 5 years. . Now in the meantime, we have further ongoing discussions with all our customers. We do get long-term visibility from our customers based on our deep customer engagements that we have and last week at our earnings call, we were actually very confident to say that we now expect greater than 25% exabyte growth over the next 3 to 5 years on a CAGR basis. maybe something trending that starts with a 3. And so what is driving that? First of all, the key growth is still there, right? More and more data is being generated by, I don't know, 7 billion or 8 billion people on the planet. And every business and enterprise on the company in the world now continues to store all their data instead of even not capturing it or throwing it away, all data is being stored. Now in addition to that, of course, we do have the AI kicker. And there is 3 main blocks there. First of all, Training of the multi-model LLM is still ongoing. A lot of training has been done, but all the big players continue to relearn and retrain their existing models making continuously improvements and which will result in the next generation of LLM. Some of them are also training very specific dedicated language models for specific tasks. -- and that will continue as well. And that requires a lot of data. But we've also moved already into inferencing, as you indicated. And I think all or about this year, 2/3 of the compute power that's installed will be used for inferencing. And what we learned now with influencing that it is actually really good for hard disk drive and data storage because every time you have an interaction with a chat box or even for more complex tasks by AgenticAI were agents now execute complex task. All that output is being stored and not just the output is being stored, but also a lot of data along the way, the reasoning, how they get to certain of those answers, all of that is being stored, so there is a history. And then third, what we also see now is the physical AI. And that's just beginning, physically, robots, humanoid, autonomous driving, autonomous cars. They use a lot of video, all those devices have video. They lose a lot of video capturing that's then being used as well to train and optimize the algorithms. There is actually not enough real live data to do that. And now they use AI to create synthetic data test being used to further optimize the models. And so when you put this all together, it's really compounding, right? More data for training, more better models, more inferencing that creates more data that's being fed into the loop and that all requires storage in the most reliable economical way, and that's hard to drive.

Great. Yes, there was a lot in there, and I think it leads into the second question pretty well. I think one of the things that people struggle with is modeling this business. It's not like there's a direct attached per GPU or when you're thinking about forecasting this exabyte growth, do you think about per gigawatt per accelerator, how are you trying to formulate what your growth rate can be relative to what you're seeing from the deployments from your customers?

Yes, it's -- you're absolutely right. There is -- it's hard for us to make a direct link to the number of gigawatts of compute that's being installed. For memory, which is closely attached to all of that, there is a more direct link as I explained it, right, hard disk drive storage and data storage follows different economics, right? It's actually -- so again, it's a compounding loop versus compute it's being reused, right? You use a couple of seconds of compute power and it generates something the GPU frees up and can be reused, but the output of your inferencing is being stored and will be stored forever, right? And so it's slightly different. And so that's why we -- the last 12, 18 months, have really established those deep customer engagements with all the large hyperscalers in the world, in the U.S., in Asia and other places. And we've really worked with our customers to get long-term visibility and ask them the right questions, what are their technology and product road maps, what are their road map in terms of data center build-out how much capacity or exabyte do they think they will need over the next 3, 5-plus years. And based on that, we are developing our own technology and product road maps to support that.

Okay. great. You're leading me right into the technology stuff here. You laid out this road map where you're going to ramp both the ePMR and HAMR up to 60 terrabytes. Talk to us a little bit about that decision and what that means for customers and your ability to supply multiple technologies, what are the benefits of that type of strategy?

Yes. We have an industry-leading technology and product road map that is, again, really developed in close relationship with all our customers. And for the last so the dominant recording technology was ePMR, right? And that's what we use today. initially, the industry was thinking that we could only get to maybe 30 terabytes on ePMR. While we've broken that logic, right? And today, we're already shipping in our latest generation of ePMR up to 32 terabytes and we found a way to continue to stretch the ePMR technology to 40 terabytes, 50 terabytes and 60 terabytes in the future. the 40 terabytes we already have. It's in qualification with 3 customers right now, and we are getting ready for a pretty steep ramp in the -- of the 40 terabytes ePMR in the second half of calendar year 2026. Now in parallel, for the last 10 years, we have been working on the next-generation technology, which is HAMR, and we are in a really good spot there. We have -- we're going to launch the product and start ramping it in calendar year 2027 with 44 terabyte solutions. That is currently in qualification with 4 customers. We are getting really good feedback from our customers. They're really happy with the performance, the quality, the reliability of the product and how it performs into their qualification systems. They're really happy with the aerial density, as I said, up to 44 terabyte per drive. And we still have a little bit more work to do, but we're not rushing it because we have our ePMR technology that goes in parallel to our HAMR technology that kind of derisks it for our customers and derisk it for ourselves, and we think that's the right way to play it.

Okay. And maybe just digging into the HAMR a little bit, talk -- so there's still a little room to go positive feedback so far. But what are kind of some of the technical milestones that you think you need to hit? And what's the current view on kind of availability and revenue time line for some of the HAMR products?

Yes. Again, we're -- we've been working on it for 10 years. It's a new technology, right. It's heat-assisted magnetic recording that require some redesign of the head, including the integration of a laser it requires some modifications on the substrates and the media to do the magnetic recording. But again, we figured it out, and we're finally getting to a point here now that we are in qualification with 4 customers. and getting ready for the ramp in 2027. By the way, HAMR is not just about getting to 44 terabytes per drive. We believe that HAMR over time, we will get to 50, 60, 70 and eventually more than 100 terabytes per drive, right? And that is the technology. We invest in that. By the way, we're also investing already in the next-generation technology that will probably come to market 10 years down the road, which is called a HDMR, heat dot magnetic recording.

Okay. Yes. I hope I'm not doing this in 10 years to ask you about it but maybe I will -- we'll see. Some other new product developments you talked about addressing bandwidth and power the high-bandwidth drives and dual pivot actuators. Could you talk a little bit about those technologies time lines? Was this customer push? Was it or customer pull or [indiscernible] push, and it's obviously addressing some bottlenecks. So what do you the outlook or the opportunity for WD in these areas are?

Yes. If you look today, in data centers in the cloud, roughly 80% of the data is being stored on hard disk drive, right? And the remainder, 20%, is stored on SSDs or NAND or flash or whatever you want to call it. And so despite the fact that, that is a lot more expensive in terms of cost of acquisition and in terms of total cost of ownership, flash SSDs is a lot more expensive. But why is it not 100% on hard disk drive? Well, flash they have a performance advantage. They have faster in and out and higher throughput and better speed than hard disk test drives. But our customers, of course, they asked us, "Can you improve your speed and throughput. And so we have smart engineers who looked at it. And we found ways to drastically up to 8x faster, right, increase the speed and the throughput of the current hardest drives. It's something that we've announced on the Innovation Day. It's called high-bandwidth drives, right? Most of that is being done in firmware and software. And that, in combination with a dual pivot, which is a hardware solution, right, will get us to those solutions. Currently, we are shipping samples to 2 of our customers. So far, there again, the feedback is very positive. They really like what we see. And it's something, again, the customers want it, the customers need it. It's important to potentially for us to expand a little bit into the more warmer and/or how higher throughputs, but we also need it as we move to higher capacity drives, 50, 60, all the way up to 100 terabytes.

Okay. I want to get into some of the business type questions here. maybe just starting structurally here, HDDs are now vast majority, vast, vast majority going to the hyperscalers. Could you talk about how the relationships are changing, whether it's contract terms, prepayments guarantees, LTAs. Maybe just give us a sense of kind of how things are structurally changing because there's such tight supply and such high demand?

Yes. I mean, again, our business has changed drastically for many reasons, right? Even if you look back a couple of years ago, -- more than 50% of the business was still client consumer. And then we started moving into the cloud, with the hyperscalers and some of the enterprise cloud providers as well. Fast forward to today, 90% of our business is with the -- is cloud mostly with the hyperscalers. We still have 5% of our revenue in consumer and 5% of our revenue in client with the PC business. But the vast majority is cloud. We have the hyperscalers. And we've completely changed the business engagement model, right? We I mean, it used to be in the past that we were -- it was turns business. We were taking orders for shipments in the same quarter. That is no longer the case. We went to, again, our major customers we explained them our technology and product road map. We invited them to our factories, and we showed them. It actually takes almost a year to produce a hard disk drive, 9 months for the wafers that eventually turns into heads and then 3 months to put it all together and test it. And so we Again, we've completely changed the business engagement model. We get a lot more information from our customers. We provide a lot more information to them. And we've learned from each other that creating better visibility in the long-term benefit both of us, right? And so now we do have better visibility from our customers, in some cases, all the way stretching out 3, 4, 5 years, right? And so while we're having those discussions with our customers, -- they -- we made it clear that, first of all, you have to place your orders 52 weeks in advance. So we have purchase orders for the next 52 weeks, right? Of course, that has volume price and the exact SKU that we're going to skip to them are fixed, but they also say we have to secure our covering all of calendar year '27. With some of them, we actually covered calendar year '28 and with some of them all the way going into calendar year '29. Those RTAs, they have a volume and a price component in baked in. But the volume there, of course, there is some flexibility there, right? In some cases, they might need a little bit more, and we will try to get them more at potentially different pricing. Also, in some cases, maybe some customers need a little less. Well, the sooner you tell us, the better maybe we can move that supply to another customer. We might -- of course, there might be some different pricing as a result of that. And so that's all baked in those relationships. I think for me, LTAs, it's a big word. For me, the most important thing is, again, better visibility, partnership, collaborating with technology and product road maps, creating more supply because for us, back to maybe your other question, right, we see very, very strong demand. We believe the next 3 and 5 years, we will be able to supply that strong land growth, not by hiring unit capacity. But by working on our technology and product road map moving to those higher capacity drives and strong collaboration with our customers, right, because they have to adopt those new technology and they have to adopt those higher capacity drives. We're moving more and more customers to Ultra SMR, which gives from CMR, which gives them a 20% boost. And that needs to all happen together in collaboration and that's more important to me than a piece of paper, which is called LTA.

Right just to dig into it a little bit, 2 follow-ups. One, I think you talked about this year mid to high single-digit year-over-year ASP increase each quarter. Is it safe to assume when you're looking at the LTAs that go to '27, '28, '29, we're factoring in similar-ish type of -- or ASP increases, number one? And number two, I know you made the point before, and I agree, is very different than NAND DRAM. But obviously, those prices are going up much more dramatically. So maybe just bridge that for us? And does that mean that at some point there is even more opportunity on the pricing side because some of the other related semiconductor areas are seeing more?

Yes. So we obviously feel really good about the pricing environment. But we do play corporately different than the memory side, which is more of a commodity and a commodity trade, right? We play the mid- to long-term relationship with our customers and our partners. And our pricing is based on value, right? The more value we provide to our customers, the more we want to get paid for that, right? And how do we provide more value? Well, by moving to higher capacity drives, right? and moving to higher capacity drives have all kind of benefits for our customers, better rack density, lower real estate cost, lower power consumption. And so we can actually charge them a little bit more on a price per terabyte basis and still delivering a lower total cost of ownership, which is a win-win for all of us. Obviously, operating in a very strong demand environment with supply that comes online as we execute our technology and product road map. That obviously helps. And so currently, we are seeing price per turbine increases in the mid- to high single digits. Last quarter, it was 9% up year-over-year. At innovation, they have indicated that I do expect for all 4 quarters of calendar year '26. ASP per terabyte to be up mid- to high single digits. And again, we are executing to that. And yes, I do believe that, longer term, there is further opportunity to improve the pricing as we deliver more value to our customers.

Okay. We did have a question. I'm assuming it's related. Can we get the microphone there, please?

And I just have 2 follow-up questions on the LTA. And some customers are willing to -- are eager to sign LTAs if another group of customers, they don't sign LTAs then what would happen to those customers who don't want to sign LTAs?

Nothing. Again, we have a set of -- it's a limited set of customers, right? You have 4 or 5, maybe 6 large hyperscalers in the U.S. You have a couple of large hyperscalers in Asia. And then you have some smaller and upcoming cloud players out there. That's the set of customers. And we have relationships with all of them. So we don't actually care too much who is a winner and who grows faster than the other 1 because we deal with all of them. Our market share with each of those players is on or about the same. And so we want to make sure, of course, that we have supply allocated to each and all of those players. Even if you have an LTA, if you don't have an LTA, it doesn't really matter, right? And for me, at is piece of paper. We have customer relationship, customer engagements. That's the most important thing.

My second question is -- if 1 customer -- because the company mentioned that for capacity increase is really for a product upgrade. And if 1 customer is willing to say co-invest with Western Digital, say, put a 50% CapEx down on the table and then just physically to increase production capacity wish you and we would be willing to take that offer and to physically increase production capacity or the factories to push up?

So the answer is no. Again, we're -- in memory, we start seeing that, but memory is a totally different business than hard disk drives, right? Memory has really high CapEx requirements. They need to go build new factories to go support the incremental demand. In hard disk drive, the CapEx is very low. it's 4% to 6% to revenue, right? So -- and we don't need to add more unit capacity. We can increase our supply somewhat in line with the strong demand that we see by moving up to higher capacity drives, and that's the answer.

Maybe I'll see if there's any other before while the mic is out. Any there's 1 over here.

This is maybe a question on potential new upcoming competitors. This is, of course, by history, a very, let's say, all of holistic industry. We've seen some competitors moving over the years in the DRAM and NAND space in China. Can you share something maybe a bit, say, from your side of the business, what you're seeing out there?

Yes. So this -- today, in the hot desk drive, there's basically 3 players, 2 large players, including Western Digital and then a third smaller player. We don't see other new entrants in that business because it's really hard and difficult business with really high barriers to enter, right? I mean you need to know how to make heads that does right and read, right? It's a very complex specialized technology that goes through a 9-month wafer fabrication process, very complex and hard and difficult to do. It's not your typical CMOS or other semiconductor thing. In addition to that, you need to know about magnetic recording, the substrates, the media and how you condition that and then being able to bring that all together in high-volume production. Again, we know a little bit about how hard disk drives because we've been doing it for 55 years. it still takes us 10 years to develop the next-generation technology. So we been working 10 years on HAMR now, HAMR is ready. We start working on HDMR, which will probably take us another 10 years. And so we don't see new entrants trying to get into this space because it's really hard.

Any other -- all right. I'll go back to the list here. I want to talk a little bit about gross margin. You kind of touched on some peripheral areas now. I'd stop covering the top I think stock 5 years ago, it was in the low 20s, and now you're more than -- well more than double that. So maybe just talk a little bit about gross margin sustainability. Obviously, upping your exabyte growth rates helps. Talk a little bit about introducing HAMR into the mix. Not necessarily adding physical capacity, adding higher products. So maybe just break us down all the moving parts as you look at the growth in sustainability over the next few years here.

Yes. Yes. I feel really good about the gross margin. Again, -- this is a totally different business than it was 3 or 5 or 10 years ago, where margins for a very long period of time were 20%, right? Here, we're mean the value of data has increased a lot, and we add more and more value to our customers by moving to those higher capacity drives. And you look at it, gross margin has moved from 20s to 30s to 40s. Last quarter, we were the first 1 crossing into something starting with a 5 handle at 50.5%. We guided to 51%, 52%. For this quarter. And we do see further gross margin expansion as ASP per terabyte continue to go up. And as the cost per terabyte continues to go down both driven by moving to higher capacity drives, which as already explained, we will move from 30s to 40s to 50s to 60s, eventually 200 terabytes, right? HAMR is an important part of that because, again, with ePMR, we think we can only get to 50, 60 terabytes. So we need to get HAMR going because that gives us the path to get to more than 100 terabytes over time. However, again, we are not rushing into it because we have this dual track with ePMR in HAMR, HAMR by itself. There's some structural difference there that adds a little bit of cost, right, because you need to add a laser per head and you have a little bit more expensive media to deal with as well. But you get the cost benefit by moving to higher capacities per drive. And combine all of that, again, I feel good about where gross margin is and further progression and expansion of the gross margins over time.

Okay. SP1 Yes, it could be all that. It could be the hardware guys stopped covering in the semi guy. I started covering you and everything is better in life. I get that a lot. You mentioned lasers and HAMR last earnings call, you did talk about acquiring some IP and talent for HAMR internal development. So could you talk a little bit about that decision? You mentioned it's obviously an incremental cost item and complexity. So maybe talk about how you think that what -- how big of a role will WD take in that vertical integration relative to merchant and what do you think that could mean for margin structure and quality of product and everything else? .

Yes. No, it is important because, again, HAMR is heat assisted magnetic recording. And basically, what we do is a laser that's attached to the head that before you flip the bid on was 0, you hit the media, and that helps to drive further aerial density. We do have an external source of the laser but we thought it is a critical component in our overall HAMR technology road map. And so we wanted to do it internally as well. So we acquired some IP and some know-how and some smart people. And so now we're working in parallel on working with our external source on an internal source that will further help us to improve the aerial density in HAMR that will also eventually allow us to more closely integrate the laser into the head itself, which will also then create some more space and help us to put more platters into the box as well. And so there's a lot of performance advantages, integration advantages. Eventually, that will also then result in cost advantages by doing it in-house and integrated.

And what would the time line for this type of integration be?

Yes. I mean this is something that we are working on in right now, but that is -- I mean, we will continue to make it smaller and further integrate it into the heads over the next 2, 3, 5 years. It's a long-term project.

Okay. Follow-up on lasers. -- you spoke about vertically emitting lasers with a shorter-term form factor, allowing you to compress the space between platters and ultimately more inside of drive. Talk a little bit about the advantages there and the time line there and how that helps scale capacity.

Yes. Again, for us, the most important thing is aerial density. We're really focused on increasing the aerial density per platter. And there's a lot that needs to happen. We need to get better heads for that. We need to get better media for that. And -- but we are on track. I mean, today, we figure out how to get to 4 terabytes per platter and we have a path to get to 5, 6, 7 and eventually 10 terabytes per platter, right? So -- that is the most important thing. Now in addition to that, we also look at how -- is there a possibility to put more platters in a box without changing the physical structure of the box, right? And we and are industry leading there. We already have 11 platters in a box. We know how to get to 12 and eventually to 14 and potentially more after that. But that's the 2 items that we focus on to continue to drive to higher capacity drives.

I have any -- just want to give 1 more chance for anybody. Yes, 1 more.

I'm just asking question again sort of from the angle, what maybe disrupts, let's say, this very optimistic future in a way. And if we look at a logic or the discussions around DPU today, there are new the parts access coming to the market, which maybe deters more efficient, hence, more -- less computing power to -- but on the storage side, you mentioned sort of the amount of data which needs to be powered us there. Is there -- if you would have to put a black head on, where -- where is the vulnerability in a way of this, let's say, positive projection going forward? The fundability or the weak spot.

To be honest, I don't see it. I mean we are the most efficient, scalable, reliable with very strong economics for our customers. Very strong economics for our customers, right? A way of storing massive amounts of data. And we don't see any hesitation from our customers. I mean, AI is a data-driven economy, right? And it all is based on massive, massive amounts of data and data storage. And we provide that to our customers in a very lucrative economical, scalable way to them. And I don't see that based on all the conversations I have with my customers, I don't see that. I mean in the past, there was this idea about like all SSDs are going to replace hard disk drives in the cloud, right? I mean, -- if you go back a couple of years ago, that was the case in the PC, right, where hard disk in PC we're going to be see did get replaced by SSDs in the PC fair. All right, right? And maybe 1 or 2 years ago, some folks believe that, that was going to happen in the cloud as well. I don't see that I don't see that. Just based on total cost of ownership differential cost of acquisition differential and just the overall performance of and technology and product innovation that we continue to do not only by going to higher capacity drives, but also improving the overall performance, throughput and speed of our hard disk drivers. I don't see anything else that could disrupt that.

And if people say that a lot of inference is going to the ads or on the device. Is that a negative or not necessary.

No. I mean eventually, data is being stored and the vast majority of data is being stored in the cloud. even if there is some local compute or local inferencing on the site, in many cases, that got uploaded into the cloud as well. .

When -- we'll try to be quick with these 2.

And if China decided it was a national strategic priority to disrupt your orders place you on a 10-year view, they don't have a chance either through everything at it. .

They've already tried it twice and couldn't get it done and we don't see them trying again, but maybe tomorrow, they can and then maybe 10 years down the road, it might succeed in that app or not.

Yes, in part linked to that, just what are the biggest obstacles to you increasing supply, what are the physical limitations you have in your own supply chain and your inability to build as you compare to the memory example being, obviously, heavy CapEx Pat build-out. What's the equivalent for you in that regard? And I guess linked to that, we're not wholly into that as what therefore, prevents that these levels of situation and margins, someone else saying actually we can also do that. What's the biggest barriers to that?

Right. And so I don't need to increase my capacity or I don't need to increase my unit capacity. I don't need to produce more hardest drives. I can supply the very strong demand growth by moving to higher capacity units, right? So that's why my CapEx will continue to stay in the 4% to 6%

And the biggest area base would you want to be?

Again, I mean, making hats is a very complex, hard thing to do, right? You need years and years and years of experience. You can go to TSMC and say, make me some hats. They -- it's a very specialized unique technology. And then the substrate and media magnetic recording is something that requires years and years and years of technology and innovation, being able then to bring that all together in a box, it's just hard to do.

Okay. I think we are up on time. Kris, really appreciate it. Thank you. very much. Thank you, everybody for joining.

Thanks.