Good afternoon, everyone. Welcome to the next fireside chat here. I'm Samik Chatterjee, and I cover hardware and networking names at JP Morgan. With me, I have the pleasure of hosting Western Digital, and from Western Digital, Kris Sennesael, who's the Chief Financial Officer. Kris, I hope I pronounced that right, but please.
Close enough.
Apologies if I didn't. Let's talk about the industry first before we get into any Western Digital questions. Do you want to read out the safe harbor first?
Samik, thanks for hosting us here. A great conference so far. Before we go, I have to make some disclosures that today I will be making some forward-looking statements based on management's current assumptions and expectations, including to our product portfolio, business plans, and performance and market trends. These forward-looking statements are subject to risks and uncertainties. Please go read the 10-K and other SEC filings that provides more information on risk and uncertainties that could cause actual results to differ materially from expectations. We might also be talking about some non-GAAP financials and the reconciliation you can find on our investor relations website.
Okay. I saw that Ambrish just walked in and saw you reading the safe harbor, so he's probably happy as well. Let's start with the demand drivers for the industry. I think genuinely everyone's been surprised of the demand drivers and how much it has exceeded expectations consistently. Just talk about what are the biggest drivers you're seeing right now in terms of driving the exabyte growth and why we could potentially have this exceed expectations more consistently going forward as well.
Yeah. I mean, it's a great time to be in the hard disk drive business. When Irving, our CEO, or myself or anybody of the company goes and talk to the customers, they come back with good news and better news. It looks like every time we talk to them, the demand is stronger for longer, which is good. Just to put it back a little bit on, in February of 2025, we indicated that we expect the market to grow mid-teens on an exabyte compound annual growth rate over the next three to five years based on what we see in the cloud. We indicated that if and when AI kicks in, the growth could accelerate into the low 20s.
Fast-forward to today, and at our last earnings call, we've actually indicated that we now expect exabyte growth to be higher than 25%, compound annual growth rate over the next three to five years. What's driving that? First of all, the base is still there. The cloud, right? 8 billion people loading pictures and video up into the cloud, sharing it to social media and so on. That continues to be very strong. In addition to that, every enterprise on the planet, including WD, we understand the value of data, and we're not throwing away any more data. We store all the data we can capture in our manufacturing site, in our engineering operations, throughout the company, and I think every other company on the planet is basically doing that.
In addition to that, we definitely now get a better understanding of AI, and AI has different facets as well. First one, of course, is training of the multimodal models that is out there that is not only text-based, but also pictures and more and more video-based. The training is still ongoing because all the big players continue to develop their next-generation model and store a vast amount of data. They also work on retraining, relearning of the models and leverage all the data. In addition to that, we've definitely moved in inferencing, and we estimate that roughly 2/3 of the compute power in 2026 will actually be used for inferencing. How do we benefit from inferencing?
Well, what we noticed is that all the chatbots, all the way down to agentic AI that is using inferencing, the output of that is being stored, and the vast majority of that output is being stored on hard disk drives. They wanna know what's being generated, they wanna know the history, they wanna know as many characteristics around that. That's driving a lot of more data storage. In addition to that's the last one, that's still early days, you have the physical AI, right? Which is autonomous cars, robotics, and in the future, humanoids. What do those devices have in common? They have multiple cameras attached to it, they wanna try to capture video as much as they can and store that and feed that into their training models as well.
There is actually not enough data available there to train the models that they now use AI to create synthetic data to complement the real-world data to further optimize their algorithms. When you put this all together, we now are convinced that exabyte will grow more than 25% on a compound annual growth rate the next three to five years.
Got it. Got it. Very comprehensive. Thank you. Maybe then how is that driving the engagement with your customer to change? For example, you've talked about in the past, Long-Term Agreements, build to order. Like, what did those historically look like before you started to see the AI tailwinds? What do those look like now? How is AI and the demand that you're seeing changing the engagement with the customer in relation to these agreements?
Yeah. First of all, we definitely have changed our customer engagement model, and we do have a lot deeper customer engagements right now, especially with the hyperscalers, right? The relationship has changed from a supplier-customer relationship 3+ years ago to a really strategic partner. All our customers, the hyperscalers, fully understand the critical value of data and the critical part that hard disk drives and storage plays in their overall AI data center build-out. That is very important. We also have, again, much deeper customer engagements. They share their technology and product road maps. They share and give us much more visibility about their demand for storage multiple years out. In return, we have done the same, right? We share our technology and product road maps with them. We've invited them to our factories.
We share our capacity and manufacturing road maps with them. They fully understand that it takes 52 weeks to produce a hard disk drive, nine months to produce the wafers that become the heads, and then three months to put it all together, including testing. We have told all our customers, or the larger customers, we expect 52 week order lead time for purchase orders because it takes 52 weeks to produce a hard disk drive. In addition to that, our customers also wanted to secure more supply further out, and that's why we have entered into LTAs with our larger customers for calendar year 2027, with some of them for 2028 and even less of that, 2029. For us, again, those LTAs are frameworks, right? Those are not take-or-pay, hard agreements.
They are there to create more visibility from the key customers to us and commitments that we make to them. Those LTAs have a volume component and a price component in it. They're slightly different by customer. Again, there is still some flexibility there. The most important thing, it's creating better predictability, better visibility on both sides.
Got it. Going to the 52-week lead time that you referenced, I mean, one of the questions, and I'm sure you've got this from investors as well, is you start investing in capacity today, you're shipping out 52 weeks later. What happens if the customer's demand profile changes, right? You're getting into these LTAs, but not necessarily that these are take-or-pay contracts. How do you get the confidence about adding capacity when you have lead times that extend to one year almost at this point? How do you get confidence around that you're not adding capacity at a peak?
Yeah. First of all, again, for us, it's very important to continue to get that visibility multiple years out, right? That helps us to plan our business. So far, as I just explained, the demand signal is stronger for longer, right? Which is good. When we look at that, do we have to add unit capacity to support +25% compound annual growth rate for the next three to five years? The answer is no. We know how to support the strong growth in exabytes of our customers, not by adding unit capacity, but through technology and product transitions, right?
As you probably know, currently, our average terabyte per unit that we ship in nearline is only about 23 TB per unit, despite the fact that we do have the highest capacity we have available today is 32 TB units, and we ship a lot of them already. We've indicated that we are currently in qualification for the next generation ePMR drives that go up to 40 TB, right? Qualifications are going really well, we are going to ramp volume production in the second half of calendar year 2026. In parallel, we have been working on HAMR for 10 years. We're now in qualification with 4 customers. Qualifications are going really well, that goes up to 44 TB drives, the ramp will happen in calendar year 2027.
In addition to that, we have a technology and product roadmap that will get us to 50, 60, 70, eventually plus 100 terabyte drives, right? We have plenty of room to, as we execute on our technology and product roadmaps, as we collaborate with our customers on qualification and adoption of those higher capacity drives, we can continue to execute and deliver and supply 25%+ more exabyte on a compound annual growth rate basis.
Got it. Got it. Maybe that's on the volume or the storage side, storage exabyte growth side. Let's talk about pricing, maybe let's start with a different angle, which is pricing structure has changed for the industry. It's a much more favorable pricing environment. Can you go into the competitive landscape of the industry structure today, how much do investors need to bank on rational competition between the players to see a favorable pricing environment continue in the industry?
Yeah. I can't speak for my competitors, but you know the competitive landscape. It's basically two large suppliers and a third smaller player out there. That's the competitive landscape. The way we go at pricing, it's a value-based pricing, right? As we execute on our technology and product roadmap, as we continue to innovate, right, and move to higher capacity drives, we provide more value to our customers. Higher capacity drives give them better rack density, lower real estate cost, better power consumption, and more exabytes, right? There is more value for our customers, and we wanna get paid for that, right? That's why we continuously increase our pricing per terabyte. If you look at last quarter.
our average, blended average ASP per terabyte was up 9% on a price per terabyte year-over-year. Again, as we continue to further deliver more value to our customers, we see further opportunity to increase the pricing on a price per terabyte basis.
Maybe let's talk about the technology approach you have relative to some of your competitors or your primary competitor. You have a multi-pronged approach on the technology front. You're doing, you're ramping HAMR, you're ramping ePMR, which is different from your competitor where there's a much more of a single focus on the technology front. Why the multi-pronged approach, and how does it impact eventually financials when you think about OpEx spending or R&D? How are you sort of thinking about the strategic rationale as well as the financial implications of it?
Yeah. First of all, our customers, rightfully so, are very careful to store your data, my data, everybody's data. They're very conservative about that. For them, quality, reliability, they don't want anything to go wrong with your data. They are somewhat conservative. We listen to them, we work with them. Let me go back a little bit. I think there was maybe a misconception in the industry that ePMR drives will max out at 28, 30 terabytes.
per unit. We clearly proved that was a misconception, that was wrong. We are shipping in high volume 32 terabytes right now. We are in qualification with the 40 terabyte that will ramp in production in the second half of calendar year 2026, and we believe we can further improve ePMR all the way up to 50 or 60 terabytes in the near future. For now, we think it will cap out there. Who knows? Maybe engineers will find ways to further stretch that out. In parallel, for the last 10 years, we have been working on our HAMR technology. It's in really good shape. We are, as I said before, in qualification with four customers.
That comes up to 44 terabyte per drive when we launch that in volume production in 2027. Of course, we will continue to expand the capacity per drive into the 50, 60, 70, all the way up to a 100 plus terabyte there as well. That's what we do from a technology transition. As you probably know as well, ePMR and HAMR comes in two flavors. We have the CMR and the UltraSMR flavor. UltraSMR gives us 20% more exabyte versus CMR, mostly done in firmware, software adjustments. Our customers want more exabytes. Two large customers are fully adopted UltraSMR. The third one is now ramping UltraSMR. We have been talking to all our other customers and say, "You want more exabytes?
Let's collaborate on the adoption of UltraSMR. We have a plan now to convert almost all our customers or the larger customers to UltraSMR, with qualifications finalized by the end of 2027.
Okay. I mean, maybe just trying to think about the financial implications here. You migrate most of your customers to UltraSMR. You have a cost benefit there. On the flip side of it, how should we think about R&D expenses to support these multiple platforms that you're working on?
Yeah. Of course, if we would only focus on one platform, that would be a little bit result in lower operating expense. We choose to support both of them because, again, we believe there is more value for our customers there. They love the ePMR proven technology, which has been out there for 10 year. They understand the quality, the reliability. We know how to scale it in a very condensed way. Yes, we have to deal with some higher operating expense right now. By the way, we also not only investing in higher capacity drives, we also investing in improving the overall performance of the drive, increasing the bandwidth as well as the speed of which the data comes in and out of the hard disk drive.
We're getting really good feedback from our customers on that with our High Bandwidth Drives as well as the Dual Pivot technology. Of course, that requires some operating expense.
Yeah
as well to support that. We believe again that the value creation we can realize by investing, focusing on innovation, is the right strategy to do that. Operating expenses going forward, there's definitely leverage in the model, right? I mean, revenue is going to grow a lot faster than our operating expense. We don't need much more operating expense. Yeah, there is some inflation and merit and some other adjustments, the operating expense will continue to come down as a percent of revenue at a good pace.
Okay. Okay. Maybe, let's move to focusing a bit on ePMR. Even as you're ramping HAMR, ePMR clearly is a differentiation for you because when investors look at gross margins, your gross margins are actually ahead of your peer despite ramping HAMR more recently compared to your peer company. Why can't your competitors replicate what you're doing with ePMR? Like what does that competitive moat look like, and what prevents competitors from copying or replicating what you've done?
I think it's a, it's an architectural and strategic product roadmap and technology roadmap decision, right? Our competitor decided not to continue to invest in ePMR and invest on HAMR, and they, that's their decision. Again, there is the, it's only two and a third smaller player, right? We don't see new entrants. We don't see new investments into the hardest drive business. We have our technology roadmap. Again, we believe that is a really smart, good decision. HAMR is a great technology, and we're all in on HAMR, right? For us, HAMR is the path to plus 100 terabyte drives. HAMR does come with some incremental cost. You have to add a laser to each and every head.
Okay.
There's some incremental cost. You have some incremental cost on the substrate. Again, it's the right thing to do. We're all in on HAMR. We're making good progress. We're ahead of schedule. We do believe that there is still a little bit of longer life with ePMR in a very cost-effective way, again, a product that our customers knows very well, scalable, reliable, and we will continue to execute in parallel ePMR and HAMR.
Okay, okay. No, that's good. That's a good segue. When you talk about the 100 TB drives, you outline the roadmap to it from a technology perspective, which will be driven by HAMR. What are you looking at in terms of internal milestones to track progress to it? What are the hurdles that you think you need to overcome if you, in terms of the path to that 100 terabyte drive?
Yeah. Again, we're feeling really good about that. When you think about how do we continue to increase the capacity per drive, there's two ways of doing that. One, which is the most important one, is to increase the areal density, right? Find ways to store more terabytes per platter, right? Currently, we are approaching 4 TB per platter, but we have a path to get that to 5 TB, 6 TB, 7 TB, eventually 10 TB per platter, right? We know in the lab how to do 10 TB. It's still gonna take us multiple years to ramp that in high volume, but we know how to do that.
Okay.
Right? Areal density is the most important factor. In addition to that, we know how to add more platters per unit, right? That's not an excuse for areal density. Areal density has to be at the industry leading levels, four moving all the way up to 10 and more eventually. Adding more platters per unit also gives you a cost advantage as well, we will continue to find ways to add more platters as well. Now how are we getting to better areal density? You need to make investments on your head technology, right? The wafer nano manufacturing and head technology that does the read and write functionality, as well on the media, right? Media is substrate. Do you use aluminum or glass, right? Currently in ePMR, we use aluminum.
HAMR, we will use glass substrates. How do you condition your substrates and turn it into media for magnetic recording and work on the recipe there? There's a lot of science, chemistry, magic that happens there, but we have very smart engineers, which again, have figured out most of that. Now it's just about making sure we continue to execute on it, deliver on it. In many cases, you just need time to prove it out, make the improvements over time, make sure the quality, reliability is sustainable, of course, you have to work on manufacturability in your factories as well. All of that is being done in parallel. It's really hard to do that, right?
Again, that's points to the high barriers to enter in the hard drive space. We know how to do it and keep focusing on execution.
Got it. Got it. You have these two paths, ePMR, HAMR. You have the 40 TB ePMR being qualified. You have the HAMR being qualified. When you go to a customer with both, what are you likely to see in terms of customer allocation of share? How do they pick and choose between ePMR and HAMR? What do you think likely will happen when you go to a customer with both of those options?
Customers don't think about recording technology. They want scalable, reliable exabytes, right? However you can get that to them, right? Obviously, with ePMR, again, it's a technology they have been using for more than 10 years. They understand the quality, the reliability. We have gone through multiple ramps of previous and next generation of ePMR. They feel very comfortable around that. HAMR is a new technology. Again, they don't have any experience with HAMR drives that are five or six years old in their fleet, right? They don't. Nobody has, right? They are probably a little bit more careful there. They wanna go make sure as you go through qualifications, they spend a little bit more time making sure they understand the quality, the reliability, how the drives behave.
That's why it takes a little bit more time. Again, to answer your question, they don't care about the recording technology. By the way, again, back to the question of the LTAs, the LTAs we have are exabyte based.
Exabyte based.
They don't specify the recording technology. They, so to speak, couldn't care less. They want high quality, high reliable exabytes.
Got it. Okay. Let me just pause here and see if anyone in the audience has any questions.
There is one here.
Hi, thank you. You said earlier that you feel confident you don't have to add unit capacity to support the 25% plus CAGR in exabytes. At what level would you feel like you do need to add unit capacity in order to support that?
Yeah, that's a great question. We've indicated that we expect exabyte to grow now at +25%. Now the question is what is +25%? I leave that up to you. I'm comfortable with the +25% that we don't have to add unit capacity. Again, I'll give you some examples, right? Today, we ship 23 TB. That's the average, right? If I can get to 44 TB , right, that's almost 44 over 23, almost 100% more exabyte per unit, right? Now again, we need to qualify the 44s. We need to ramp the 44s. There is a certain adoption curve. You have to move customers over and over.
In 44, right, shortly after that, we will get to the 50s and the 60s and the 70s and the 100+. If I look ahead over the next three, five plus years, and I look at my technology and product roadmap, I don't see a need to add more unit capacity and still support very strong exabyte growth year-over-year.
Great. Just a quick follow-up, if I may. Whatever that number is, much higher than 25%, is it fair to say that the hyperscalers haven't shown you that level of stronger for longer yet?
No. Every time we talk to them, they come back with stronger and longer, right? Again, it was like a year ago, they gave us a certain view of the demand for the next two, three years, and it was like, "Yeah, but we don't know exactly year four or five," right? As we had multiple discussions on a continuous basis over the last 12, 18 months, they now signal a much stronger demand each and every time we talk to them, and they are willing to almost commit further out, 2031, 2032.
Yeah. One more question in the front here.
Thanks. Can you just talk about some of the innovation on the high bandwidth technology? Just interested kinda how that shifts hard disk drives potentially in the data center and whether it can expand the, kind of, share of the total storage spending.
Yeah. Today, like, if you look at data centers, right, roughly 80% of the data is stored on hard disk drives, and about 20% is stored on SSDs. Despite the fact that SSDs is a lot more expensive in terms of cost of acquisition and even in terms of total cost of ownership. That was the case 12 months ago, and today that's even a lot worse, right? Their cost of acquisition has gone up a lot. Their total cost of ownership has gone up a lot. Our customers don't like that, right? There is a reason why about 20% is stored on SSDs, because they have a performance advantage in terms of throughput and speed, right? Our customers came to us and say, "Can we work on this?
Can we collaborate on this? Can you help us? Can you improve the performance of your hard disk drives? We have smart engineers that looked at it and studied it and found solutions. We've basically found a combination of two solutions. One is High Bandwidth Drives, which is mostly a firmware, software solution to create better performance. Then in combination with the Dual Pivot, which is a hardware solution, and when you combine both of that, you will have more throughput, or a bigger pipe, more bandwidth that will over time increase the performance 2x, 4x, and eventually 8x compared to what we have today. Customers are very excited about that. We are already sampling with two customers the High Bandwidth Drives.
We see a lot of traction there. As we execute there, right, there is a possibility for us to, so to speak, encroach into that SSD segment. We never gonna wipe that out completely, but it creates an opportunity for us to move into a warmer layer of data.
Let's move to, for the last few minutes, discussing HAMR a bit more, and really trying to focus on HAMR yields. In the past, you mentioned when you were ramping 18 terabyte or 20 terabyte platforms that you start with about 60% yields that ramps when you get to a more mature phase that ramps to about 80%-90%. Is that a good trajectory to think about HAMR that you ramp starting with 60% and you get to a maturity of 80%-90%? How much time before you get to that mature level of yield?
Yeah, no, I think the story is slightly different. I mean, a couple years ago, overall yields were in the 60s, right? We've really focused on automation in the factories. We captured a lot of data and initially used machine learning applications on top of that, and now we use more advanced machine learning, call it AI applications, that has really helped us to drastically increase the overall yield from somewhere in the low 60s to somewhere now in the low 90s. Now, of course, every time you go through a technology transition, you go a little bit through a learning curve in terms of yield. Although if it's moving from one generation ePMR to the next generation of ePMR, that has gone really smoothly.
It's a little bit more challenging if you move from one, recording technology, ePMR-
Okay
to the next generation HAMR of recording technology. That is not easy to do. That is challenging, we know how to do that. Again, the good news for us, based on our strategy and technology and product transition, we're not in a rush. We wanna take our time. We wanna make sure, again, that the customers are comfortable with the quality, the reliability of the product, and ourselves are comfortable with that. We are going to ramp it in the factories, not feeling rushed, because we have the 40 TB ePMR, right? We will take our time, go through the learning curve, and make sure that if and when we launch HAMR right, it is kinda neutral to accretive from a gross margin point of view.
Got it. Got it. Okay. I was gonna follow up on that, but let me just wrap it up with a question in terms of pricing then. You did report 9% pricing, as you said. When you think about the go forward, obviously you reprice these LTAs, you get that pricing benefit. Why shouldn't we expect pricing to accelerate from here on, given that generally a waterfall would indicate you have more LTAs coming up with better pricing over time, why shouldn't pricing accelerate from the 9% that you've seen last quarter?
Again, from a pricing point of view, we feel good, right? The demand is strong, the supply is increasing, but we think over the next three to five years, demand supply will remain tight. It's up to us to execute on our technology and product roadmap, provide more value to our customers by moving to higher capacity drives, and then eventually as well, increasing the overall performance through the High Bandwidth Drives and the Dual Pivot. As we provide more value to our customers, get paid for that, right? We're not going to give it away. We wanna continue to increase the price on the price per terabyte, while we do that, provide more value to our customers, potentially lower their total cost of ownership.
It's a win-win for them, it's a win-win for us, and it's a win-win for our shareholders. That's how we approach it.
Great. I'll wrap it up there, but thank you to the audience, and thank you to you as well for coming to the conference. Thank you.
Yeah. Thanks for having us.
Thank you.