Good afternoon. I think this is the last session. Maybe there's one more. But I'm Tim Arcuri, and I'm a semi and semi-equipment analyst here at UBS. And we're very pleased to have Sandisk with us. We have David Goeckeler, who is the CEO, and we have Luis Visoso, who is the CFO. So thanks to both of you.
Thanks. Great to be here. Thanks, Tim.
Okay, so first of all, let me just say congratulations. Because as you know, last time you and I talked before you split the company up, I was a bit skeptical of the value I'm talking about.
Oh, come on, Tim.
Obviously, I've been completely wrong. So.
Do you mind if I read the safe harbor?
Yes, please go ahead.
We can celebrate by reading the safe harbor.
Great.
We'll be making forward-looking statements in today's discussion based on management's current assumptions and expectations, including with respect to our product portfolio, business plans and performance, market trends and dynamics, and future financial results. These forward-looking statements are subject to risks and uncertainties. Please refer to our annual report on Form 10-K and other SEC filings for more information on the risks and uncertainties that could cause actual results to differ materially from expectations. We will also be making references to non-GAAP financials, and reconciliations of our GAAP to non-GAAP financials can be found on our website.
Thank you.
Great.
Appreciate it.
Perfect. So let's talk about just the business in general. And obviously, things are very tight. Pricing is going up. Some of this is because of demand, but some of this is due to the strategy where you're investing as if supply to meet mid- to high-teens bit growth, and unconstrained demand is more in the low- to mid-20s, possibly even higher right now. So you're sort of purposely undershipping and maximizing margins through allocation. It's a similar playbook to what's being done at WD as well and has remained the case. So can you kind of talk about this?
Yeah. So I think you got it. I think the way you framed it is right. I think this is a long-term kind of supply side. The way this has played out, if you look back to the major downturn in the market, there was a lot of value destruction in that downturn. I think we came out of that downturn. We certainly did, looking at our portfolio, realizing that we needed to be more specific on how we manage supply. I think that there was kind of a view in the NAND market that when the economics turned against you, you released a new node, and those new nodes drove better economics because they were lower cost, and it also expanded the TAM. And I think the downturn was really that strategy kind of ending spectacularly as a business strategy.
Because what has happened in the 3D era is you have these high layer count nodes, very CapEx intensive, maybe three times as CapEx intensive as the 2D era. And also, more importantly, the relationship between bits provided per node and cost downs fundamentally changed. It used to be more cost downs, fewer bits, and it's moved to more bits, fewer cost down. So this idea that you could release a new node and that it would change the economics of the market actually inverted. And what happened is you ended up flooding the market with supply, which drove even more oversupply situations. So I think we came out of that downturn, and we were pretty explicit about how we talked about this. We talked about the new era of NAND, how we were going to manage our portfolio.
And it's, we're going to be more prescriptive on how we manage supply. And we're going to manage supply for a mid-teens growth rate. This idea that the market was going to grow at 30% and cost downs were going to be 15%, like that was like ancient history at this point. So we're going to manage for mid-teens growth. We're not going to talk about cost downs anymore because they're variable and they're not what they used to be, and you can't just model them so easily. And that's what we're going to manage to. And I think that played out. You played that forward for several years. And we started seeing this probably in late 2024 and in 2025, started talking about the market was we thought the market was going to be undersupplied through the end of 2026.
We saw some wobbles in that because of the way the enterprise SSD market played out, which is a larger story. And I think what you've seen over the last quarter is essentially the end game of that whole strategy playing out, where inventory had been drained in the suppliers, inventory had been drained in the customers, and we're now just back to raw supply and demand. And we're in a very efficient, very liquid market. And markets are very good at equalizing. And they do that through price. And so I don't view the market. I personally don't subscribe to these terms like tightness and shortness and all this other kind of stuff. We have a market. And we have an amount of supply, and we have an amount of demand, and the market will equalize to what the value of that supply is.
That's happening at a very rapid pace.
In the past, the skeptic would say that that never really worked because there are enough NAND suppliers that one always wants to gain share. And so one is always going to invest to a demand trend line that's higher than the others. And so there would always be a disruptive force on the supply side. What's changed now? Is it because demand is so much higher than what everyone's managing supply to?
I just think that it's a big market. I mean, maybe when the market's a $5 billion market, you can, like, I'm going to use my balance sheet to distort the market, and I'm going to do all these things to change the dynamics. But we're going to be close to a $90 billion TAM. This is a big market. So the idea that one person can influence it that directly, I just think that that is, again, the downturn showed that if you oversupply the market, everybody loses. I mean, we have one of the smaller shares in the market, and we performed the best. So that strategy would have said the opposite should have happened, and that didn't play out.
So I think that this is a market where my observation is somebody that's only been in the market for five and a half or six years and has managed a bunch of other technology franchises before this. There's all these rules of thumb about the way things work that are just not really true. And they're certainly not true anymore. And I think if you sit around waiting for that world to come back, it's going to be a long wait. I think the market's going to behave differently going forward than it did in the past.
So maybe on that front, can you talk about the decisions to add supply? What are the variables? I mean, I think maybe some of what's going on today is because the incremental clean room space is going to probably go to HBM. But if the margins keep going higher, then the margins in NAND are going to be higher even than in HBM. So for now, would you agree that maybe for now, the new clean room space, yes, you have a limiter on supply because people will add HBM capacity and not NAND capacity? But what are the variables when you think about adding capacity?
I'll start by saying we add capacity every year. We're investing billions of dollars to grow the market in the mid-teens %. We have absolutely zero commitment from people to buy that output. We're already investing an enormous amount of money to increase supply in a market that we think is, and we do that freely. We think it's a great market. I think this is a great business. It's why I spend my personal time here. I think it's why Luis spends his time here. We think this is an unbelievable market, and the economics can be fantastic. I mean, NAND is like a market that's extremely diversified. It's got tons of different kinds of buyers. It's obviously every mobile phone, every PC, every tablet, now every data center. It's a spectacularly attractive market.
We invest billions of dollars every year to grow that market. You're asking the question, I think, is, well, what would it take for you to invest for that market to grow faster? I think the way I think about that is when we invest, we think about a 10-year return. We're building a fab that costs billions of dollars. Once we build that fab, we need to run it 24 hours a day. The output's going to come, and we need to sell it. The market is structured in a way today that on the demand side, they can show up every quarter and decide if they want to buy something. Our time horizon is 10 years, and most of the buyers' time horizon is three months.
I think those two things need to converge a little bit more before we start talking about, well, things should grow faster. Perhaps the demand side should think about making commitments that are longer than three months at a time. And I think that behavioral change will happen because I think that's a very healthy thing for a market. I think the market is, I think that the market has been conditioned that I can show up at any time and buy NAND. And I think that it would be a healthier thing to say, well, if we're going to actually grow faster than that mid teens, which is a nice growth rate, a $90 billion market growing at 15% is an attractive place to spend your time, assuming the economics are right.
And we need to get the economics right to be able to continue to invest that money and not go through these huge episodic periods of losing money. And so when people ask me, well, what is it going to take for you to grow the market faster than what you're spending billions of dollars to grow it already, it's going to be more conviction that there's sustained demand. I can't do anything about providing more supply in the next two, three, four quarters. I have to think about making an investment that's going to pay off for many, many years. And so when I think we get the behavioral change in the market, that there's conviction all around that we're all willing to commit to that higher growth rate, then that would be a way to start to think about how to invest differently.
So really, you're talking about LTAs. And you have been talking more about that. How are those discussions going, and how many bits do you think can actually move in the near term under an LTA?
Yeah, it's a little bit premature. But what we've seen is a few, very few, but very large customers have approached us. And they're interested in having a conversation. And they're talking about 2026, 2027, potentially beyond that. But as I said, it's premature to talk about more details because those conversations are just starting.
I guess maybe you wouldn't even answer this question, but the parameters under which you would engage, based on what you said, because if you're going to go invest money under an LTA, you want to make sure you have guaranteed A, guaranteed offtake, and B, guaranteed price within a range.
Yeah, I mean, that's what Luis said. I mean, it's a little early to start talking about what the terms of that would look like. I think the fact that on the demand side, people are lifting their head up and looking further down the horizon and saying, hey, I've got a big investment. I have a big investment in my business, and it requires a big investment for you and your business. Perhaps we should have a conversation that's longer than a quarter at a time. And I don't mean that in a negative way. I mean, we have discussions with customers today all the time. Customers commit demand to us a year at a time, but that's conditional on us agreeing on the price every quarter. So it's kind of a structure that makes the market work.
The market works fine, but it's not a structure that says there's really a level of commitment to actually buy something. That actual commitment to buy something happens every quarter for the vast majority of customers. There's a small number of customers that do it on a yearly basis, but.
Got it. Maybe we can talk about China. That's another debate, obviously. Is the base case that China will be supplied domestically, but the rest of the world will not be supplied by China? I mean, it seems like Hynix and Micron are basically exiting that market by and large. So are there opportunities for you in China? What's your thinking around that market?
Obviously, there's a great company in China that's supplying China. They're an enviable company, and they're investing heavily, and I think that's a China for China story, and I expect that to continue. I still think there's incremental opportunity there, but it's clearly a market with an indigenous supplier that's doing a very good job of supplying that market, but that's not a supplier we see in the rest of the world.
And do you think that other suppliers are de-emphasizing China enough that it would create opportunities for you?
Look, I mean, it's hard for me to talk about what other people's strategies are. I just can talk about what our own strategy is, and our strategy is we certainly do have great customers in China. That's a market we do participate in. But we participate in markets all over the world. We obviously have a huge consumer franchise that's sold in China. So it's a very important market to us.
Let's talk about demand elasticity. I mean, prices seem like they're going higher for the foreseeable future. What does it mean to demand in your eyes, in particular for edge devices and for consumer devices? There's a lot of concern that demand is going to get destroyed by prices going up this much.
Yeah. I mean, it was really part of the first question we didn't get to because we were talking about the supply side. But in the midst of this long supply story going on, we're investing to this mid-teens growth rate and seeing that we're going to get better equilibrium of supply and demand. You also had the data center market over the last year coming in and continuing to up and up and up the amount of demand in that market. So you've ended up in a situation, as we look at 2026, where there is, if you looked at unconstrained demand, it would be significantly above what we think actual supply is going to be. So there's clearly going to be a situation where not everybody gets everything they want. And I think that's, I mean, that's a market. That's what a market is.
I think that's what's happening. It was what I said earlier. That's what's happening right now. You have a market. Markets are extremely efficient. We're in a very liquid market. We're in a very big market. And that market is very good at matching supply and demand on a real-time basis. And that will continue to happen throughout next year. As long as you have this situation where you have demand over supply, you're going to have that situation. And there's going to be markets where it's just not economic to sell into that market.
So, would you say, would it be fair to say maybe, yeah, maybe that happens, which is what you're saying, yeah, maybe it happens? But maybe it's a little premature to worry about that because prices are really, if you look at year-over-year prices, they're up, but not, I mean, not a ton. They're up a lot in the last six months, but in the last year, they're not up that much. So even if that were to happen, would you say it's a premature debate?
I mean, again, I think that's well said. I mean, we're very optimistic about where the business is going. I mean, the business is very strong. I mean, on a real-time basis, business is very, very, very strong. But I mean, the quarter we printed last quarter was 29.9% gross margin. Our through cycle target is 35%. So we need to see some continued progress on the economics of the industry. And I think we'll continue to see that. I don't know. Luis, do you have any?
Yeah, I totally agree. We're making progress. We're very confident. We're going to continue to perform well. But there is still a lot to do.
Great. Let's talk about data centers. So the WDC, the split knife cuts both ways because obviously the combination helped the old Sandisk GainShare and data center. But there are some synergies from splitting the company up, which seems like maybe it's in data centers. So can you just talk about that? And I remember back in the 2022 Analyst Day slides, the combined company showed a 16% share goal in 2026, up from 8% share in cloud. So can you just talk about that?
Yeah. I'll maybe give a little longer historical context on what I think of Sandisk being in the company. I think when you look at, you think of kind of the pre-Western Digital Sandisk. I mean, just a fantastic company, a really great consumer brand, great IP monetization. I think looking back on the chapter of Western Digital was really the rise of the client in the Sandisk portfolio. I mean, we're like 25% client share. That's not surprising because we had the product that we were replacing in the client. The client started as the PC was hard drives that got replaced by a client SSD. It's not surprising that a company that had both franchises took advantage of that because you knew the customer, you knew the use case, you knew how to test it, you knew all the features, and so that worked really, really well.
Throughout that, the company was working on making progress in enterprise, which is a very big prize. It's also very difficult. It's the most difficult market, most IP intensive from a controller point of view. We've been working on that market very diligently. I think you're going to see the chapter going forward is where we really start to make progress in that market. We've spent the last three-plus years building a new controller for the storage class enterprise SSD. That's what we've called the Stargate program. That program is just going in the qualification with the first two hyperscalers. We have a third hyperscaler that wants to start a qualification in 2026. Those are long processes. There are many quarters to get through that. But the prize on the other side of that is significant consumption.
So we feel very good about where the technology is. We feel very good about where the customer engagement is. And I think we'll see that as we go through 2026. I think we'll see that story continue to get better and better.
And does the qualification of BiCS8, the timing of the qualification, I mean, these are the highest density, best performance products by some metrics in the market. So does the timing of that, the call cycle there, also make you optimistic to gain share in?
I mean, that's well said. I mean, we think we've got a lot. You got to get a lot of stuff right. And it starts with the NAND itself. If you don't have a great node, it's hard to make up for it with the controller. And so we've got a great node in BiCS8 UltraQLC. QLC performance is extremely good. Power efficiency is very good. We're building a brand new controller on top of that. And so we feel very good about the way the whole picture is coming together. On top of that, we have two terabit die. If you're going to build a high-capacity drive, the bigger die you have, you need fewer of them so you can build a bigger capacity in the same footprint.
So a lot of things have come together that took years of development to all arrive at the same time. The node, the die, the controller, and then the customer engagement. And all that is happening right now as we speak. We're ramping BiCS8. We're starting the qualifications of the controllers after many, many years of building it. And it's a great product. And so we're super optimistic about it.
Great. Let's talk a little bit about your roadmap. You and your partner historically have a little higher bit density, and you don't stack layers as much as the peers do. And that strategy has worked out pretty well. It's been a pretty capital-light strategy. And customers seem, as you said, pretty happy with BiCS8. Does that approach run out of runway at some point, and you have to start to join the layer race?
I think we know how to add more layers. I think that our strategy is you want to add as few layers as possible because more layers means more CapEx. And so I think this is really the story of the joint venture. With Kioxia, we're the largest provider of NAND. That means we have the largest R&D team, and we've been doing this for decades. And as you do this for decades and you have people that have been doing it and you can invest a lot, we can both, if you will, punch above our weight because we have the same roadmap. So we can invest more than we could individually. And so you roll that forward for 20 years or 25 years, you have this roadmap that we've very focused on capital efficiency.
How do I get the most output for as little capital as possible? And to your point, the way you can scale NAND without getting into it too deeply because I'm not a NAND designer myself, but there's lots of ways you can scale NAND. You can increase the die size, memory hole density, which is what you referred to, if you can get the holes tighter, that means you can put more of them in the same footprint, which means you need fewer layers. Our actual IP on the cell itself is very, very strong and very good, which packs them tighter. So basically, you're just continuing to iterate on that, the materials you're using, the way you're doing it. And that team is just really expert at that, and they've done a fantastic job.
And even before the downturn, when everybody was trying to, the more layers you have means you're ahead. I've never believed that. I've never bought into that. It's all about CapEx efficiency, delivering the most bits you can in the most economical way. And I think our teams between Kioxia and Sandisk are extremely good at that. And we're working on multiple generations into the future on driving that roadmap. So we feel the fundamental technology, which is the basis of the business, if you don't have a strong foundation, the actual NAND node itself is the foundation of the business, that is very, very solid. We feel very good about that.
Great. I wanted to ask about High Bandwidth Flash. Inference is making it potentially more interesting. You were the first to present it, but now Hynix is more vocal about it as well. You have a partnership that you talked about last call. Can you talk about the use case and the advantages? And most importantly to me is how do you get around the power being high and the write speeds being low?
Yeah. So this is an interesting story. Internally, when the teams came to us and we started talking about this, and I think this was a really, really smart insight. Again, the folks have been doing this for a long, long, long time and are experts in this. And I think the insight a number of years ago was NAND designers have always been focused on density. How do I get the most density? Which is a great thing to be focused on, right? Because we have a very big market and people need a lot of bits. But I think the insight a couple of years ago was, well, what if we focused on bandwidth instead of just density, just raw density? Let's assume we got the density roadmap. That's there. We're going to pay attention to it. We're not going to lose focus on that.
But what if you started to think about how do I design NAND for better durability, better bandwidth, especially when you start looking at specific use cases like AI inference, which is a lot of just loading this huge model into a processor. And so in some sense, it's almost deterministic the reads you're going to do. So how would you redesign the NAND die to optimize for that? And it turns out when you get a bunch of really smart NAND designers and you have them focus on that problem as opposed to just more density, they solve that problem and come up with some really interesting ways to do that. And that's kind of what turned into High Bandwidth Flash, which is if you look at the inference problem and you say, how do I models are getting bigger. There's no doubt about that.
How do I get more memory around this problem of inference? And the way to do that is to bring NAND to the table. And we're seeing this in the enterprise SSD market. You're seeing that models are getting bigger, caches are getting bigger, context windows are getting bigger, and you're having to move up into the NAND layer of the memory architecture. That's one of the things that's driving this data center demand. And so High Bandwidth Flash is another way of looking at that equation and saying, how can we bring this highly scalable technology to this new use case and solve it in a unique way? And that's what the team is doing. We thought when we went to our Investor Day, we were asking people to invest in the company. They should know what we're working on. And we announced it.
Hynix reached out to us and thought it was a good idea. They wanted to collaborate on the specification of the system level, not of the NAND design and all that kind of stuff, but just how would you build the system with this? It's been a good collaboration. There's still heavy lifting going on there, working with customers. It's not a plug and play for some other component of the architecture. It's rethinking the architecture in a way that's more scalable to introduce this technology, which has a lot of great benefits, which you can bring an enormous amount of capacity in a very small footprint. We continue to work with customers on how would they integrate that into their product design, whether it's a device or the cloud.
We continue to work on designing the actual die itself and building the controller on top of it. As we go throughout, we've talked about the dates for having the memory by the end of this year and having an initial system by early 2027 that we can put in people's hands and start to make some of those use cases come alive.
Luis, I wanted to ask you, you paid down $500 million of your Term Loan B. So really, I have two questions. A, what's next? And most importantly, you're obviously generating a lot of free cash flow. So at what point do you expect to start to return cash to investors?
Yeah. That was just last quarter, right? Remember in February, at the end of February, we started with a $2 billion TOB and $1.3 billion in cash, right? So we had, call it $700 million in net debt. And now last quarter, we reported $91 million in cash, right? So we've made a ton of progress, and we feel very proud about that. I think going forward, I believe that the way you create value is by generating cash. So we're very focused on continuing to generate cash, and that also would be through gross margin all the way to cash. And we're being very prudent on what we do with that cash. We want to take enough time to really assess our options. But at the end, we'll return the cash to investors, right? When exactly and in what form? We need more time to evaluate our options.
Maybe just last thing. How do you sort of talked about the eSSD rollout for high-speed TLC and high-capacity QLC? Can you sort of update us on that?
Yeah. No, this is a big part of the product strategy. We launched a product last year on the compute side, what we call the compute side of things, which is the TLC product, very fast interface. That product is qualified at multiple hyperscalers. When you see our numbers, when we talk about 26% sequential growth in enterprise SSD, it's that product carrying a big piece of that as it scales out at those places where we're qualified. And then we have the storage class product, which is QLC, what we call the Stargate program. That's just going into customers' hands now and starting qualification. And as we get through that, we'll see that start to kick in and add to the growth as we go through the second half of 2026.
When you add that all up, we feel very good about that on top of what's happening on the demand side, where the use case continues to evolve and continues to accelerate. We feel very good about where that part of the business is. It's going to give us just more optionality of where we ship our bits on a quarter-by-quarter basis to get the optimal return.
Great. We've run out of time, but thank you to you both.
Thank you.
Thank you. Thanks for your time, Tim. Appreciate it.