All right, welcome back to the Barclays Global Tech Conference. I'm Tom O'Malley, Semi and Semi-cap analyst here. Very pleased to have David Goeckler and Luis Visoso from Sandisk. Thank you for joining.
Great to be here. Thank you, Tom.
Yeah. So why don't we start? How's the crowd?
I'm sorry.
Oh, please, disclosure.
I'll say it very quickly. He will 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 risk and uncertainties. Please refer to our annual report on Form 10-K and our other filings with the SEC for more information on the risk and uncertainties that could cause actual results to differ materially from expectations. We will also be making reference to non-GAAP financials, and a reconciliation of our GAAP and non-GAAP financial results can be found on our website.
Perfect. So how's the quarter progressed thus far? What are you seeing in terms of demand trends and customer engagement?
So, I think we came into this quarter a very dynamic market. I think things were changing very rapidly from this kind of moving into more of an allocation mode. And I think we've definitely seen that continue as we go throughout the quarter. Things have gotten stronger. The market continues to be very dynamic, move very fast. I would say the big picture, what we're finding that our demand side is very clearly in a position where they value supply over price. And so the market is working the way you would expect, and supply and demand is constantly being rebalanced in that equilibrium point. Wow, that was hard. I mean, it's early in the morning. That point is in a constant state of flux as the market continues to kind of absorb where we are as far as the supply-demand balance.
And I just wanted to interject because you saw the news last night, obviously an earthquake in Japan. You have facilities that are nearby. Any comment on impacting to the fab or any production facilities?
Yeah, no real comment on that. Team is dealing with it. We'll update more. No immediate impact into what's going on. They deal with these issues, and there are some tools that need to be restarted, and they're working through all that now, but we don't expect any near-term impact on output, and we'll update more as we work through the whole process.
Gotcha. Helpful. Thank you for going off script there.
Sure. I appreciate that.
In terms of the gross margin side, so you talked about demand trends. You talked about through cycle performance of about 35% at the Investor Day. Just given the current dynamics, could you talk about how that may change that outlook? Could you get there a little bit faster?
Yeah, I think what we're seeing in the market right now, and I'll ask Luis to comment on this as well because he's very, very close to this. One of the things I think we did when we set up Sandisk is we moved kind of all the pricing and allocation organization to Luis, which I think was a great move. So he's very close to this on a day-to-day basis. But I think what we're seeing is a market restructuring, right? We're not necessarily. I think we want to frame this in the context of the past. It's a cycle. It's a supercycle. But I think what's really happening is the market fundamentally changed coming out of the downturn in 2023.
There was a lot less investment in capital because there was just, we had to come off of this high of this idea that there was continuous 30% growth, 15% cost downs. That's not the reality. Growth is going to be more like mid-teens on an ongoing basis. Cost downs are not as much as they were in the past. So the market has been kind of resetting and restructuring. I think we're on this. We certainly see that the market really values our technology. I think one of the things I think one of the reasons we're part of this company is we really like the market, the big market, very diverse market, lots of demand drivers. NAND is used in almost every single useful technology device in the world.
We're also seeing this structural change where the data center market in 2026 will become the biggest market, biggest NAND consumer after, let's call it, 15 years of mobile being the biggest demand consumer. It doesn't mean mobile's going down. That continues to grow as well. PCs continue to grow. Devices continue to grow. But now we have this huge other market that is getting very, very large and will just grow more as we go forward. And I think the whole market is kind of trying to digest this in real time and figure out what is really the marginal value of this product. And I think kind of the answer to the first question, that equation is changing almost daily as we figure out kind of how this market is going to settle out.
Yeah, no, totally agree. And I think your question on can we get there faster. We guided to be above that now. So we're getting there as fast as possible. Now we guided 35 to be our through cycle model, right? And we have three quarters below 35, so we clearly need to average higher than that. And if you look at several years back, we've been clearly below that. But as David said, the market dynamics are changing, and we're frankly, 35 is what we aim to deliver, but we need to do better than that. We need to generate our return for our investors, and we believe that getting above that is super important.
So if I look at memory in general, DRAM tends to be a little bit more straightforward for the average investor because you see the accelerator, you see how much DRAM and HBM is around that accelerator, and you can do some math around how much new or incremental demand is being brought to the market, and then you can do some math around supply-demand dynamics in DRAM. NAND is a little trickier where you don't necessarily see NAND or all of the NAND that's needed for all of these accelerators sitting in box. It sometimes sits at a box at the end of the row. There's some that sit in that box, but the math is just a little bit tougher to do.
So you have these two dynamics going on, right, where traditionally the largest exabyte contributor to the data center, the hard disk drive guys, have talked about not committing as much CapEx to the market. You're seeing them just kind of stay at a similar, slower trajectory of supply than historical. And then there's also this data center boom in which I'm sure that you're seeing some tailwinds for NAND as well. So how do you split out how much of this tightness is really related to the demand side or how much is related to tightness given one of the largest exabyte providers is slowing down?
There's a lot to unpack there. Let me start to work on it. First of all, I'm going to quibble with you a little bit. The words like tightness are not in my vocabulary. We have a market, there's supply and demand, and the market then rationalizes how those things equal out. Any market, you could increase supply and you would change, or you could decrease demand or whatever, and you would change that equation, right? Just because that equation of where it balances is changing doesn't mean it's "tight." It just means we're figuring out the balance point and the marginal value of our product, right, before we see excessive demand destruction or something like that. That's the starting point. Look, I think where we're at in this industry is rooted deeply in what happened in the downturn, right?
I think the industry kind of had a way, or I'll just talk about us. I'm not as comfortable talking about the industry because I don't run those other companies, but for our company, I think that there was this idea that we continue to drive the cost of our product down. It increases the TAM. That accelerates the growth rate. We're in industry. We're a business where we have a lot of R&D productivity, a lot of R&D productivity. We can introduce a new node, and that node will deliver an enormous amount of new product to market, and so we got to the point, especially where 3D technology as well. I mean, the NAND story from an R&D perspective is just spectacular, right?
This 3D technology, where we can continue to deliver more productivity to the market, and that just got kind of out of hand, right, where those new nodes got extremely expensive, three times the CapEx is what it was in the 2D era. They became much more productive, and the amount of cost downs you got out of those nodes was just not as much. So this idea that I could just continually fuel the market, drive the cost down, the market would absorb that, and my economics would be okay. That model spectacularly ended in 2023 because you got too many bits, not enough cost downs, too much capital. That is what that strategy led to, and that's how the industry had the worst downturn maybe ever, tens of billions of dollars of value destruction.
So now coming out of that, I think there's a different way of managing the industry, which is be much more kind of thoughtful on the supply side. Just because we can launch a new node doesn't mean we should, right? The market will not absorb all of this additional supply. The cost downs are not significant enough to rescue your economics in those deep downturns. So we have to be more thoughtful on supply. I think that's what we've been doing. And when we look around the industry, that's what's been happening. And we're still investing enormous amounts of capital to drive mid-teens % growth in the industry. That's a great industry. $100+ billion industry growing at a mid-teens rate. That's a great industry.
And so that's where all of that's kind of we're 2+ years into that way of running the business, and we're starting to see the results of that now. Literally, over the past 10 or 12 weeks, we're starting to see that shift in how people think about our technology, how they procure our technology, how do we have conversations with our customers about future investments. And that's where this is rooted, in my opinion. And then you have all kind of second and third-order things. Has data center demand gone up? Yes. I mean, it wasn't that long ago. We were projecting, call it low to mid-20s growth in data center exabytes in calendar year 2026. Now we're projecting 40s, right, something with a four handle on it. And we're projecting that market becomes the largest consumer of NAND.
That's fundamentally changing the way the dynamics of the industry work. It's new customers that are the biggest buyers. They have different ways of thinking about their procurement. Quite frankly, if you look at their demand numbers in like 2027, 2028 kind of timeframe, you project forward what's happening, the numbers are significant amounts of demand. I think they're thinking about it differently. And so I think all that is happening at once. The market is trying to digest that, and that's where we see this kind of world we're in right now is that gets sorted out. And I think it's going to take a while to sort that out. And I think it's going to have far-reaching impacts on the way the industry works.
And I don't think we know what all of those are yet, but I think one thing we do know is we're producing a very valuable product that is a part of every part of the technology stack. And getting back to a little bit of your question, AI is just another part of that, right? Clearly, the models are getting bigger. The models are getting bigger. You need more storage. The context windows are getting bigger. The caches are getting bigger. And that architecture, and then you have the limited scalability of DRAM. And so that architecture has to extend now into NAND. NAND is the most scalable memory technology out there. It's maybe the only scalable memory technology out there. So it naturally has to become part of this equation as this architecture continues to mature and the demands just get bigger and bigger.
And that's happening, and that's just another part of this demand driver into this equation. And you're right. It's not as easy as saying there's this many GPUs, therefore I have this much HBM. It's more complicated than that. And we're all still sorting out what those demand drivers are. But when you add it all up, that's how you go from like a market that's as big as it is will be 35% plus of the entire NAND market at 2026, growing at something with a four handle on it is a pretty spectacular result. And that's what's happening underneath there. And I don't believe it's because it's a substitution for HDDs. I just don't. That doesn't make sense to me, quite frankly.
I think it kind of logically, like, yeah, if that were true, maybe that would explain what's going on, but I think there's a bunch of other reasons why that wouldn't be true, and I think these other demand drivers are more sustainable and more real.
So it sounds like a structural change in the way that the industry is viewing NAND and the way that you are positioned in this ecosystem. I guess one thing that comes with that change is conversations around long-term agreements with customers. Let's not use the word tightness, but let's talk about structural importance in the industry. If you are structurally important, customers are going to want to get their hands on your product earlier, and they want to have assured supply for longer. So maybe talk about what you're seeing in terms of LTAs or longer-term agreements with customers, and where do you think the right balance is? Is that a quarterly kind of conversation? Is that a half-year or a yearly? Anything there would be helpful.
Yeah, I would say it's a little bit premature to get into the details. We're in conversations. I think a few things I would say is some big customers have reached out to us. It's not the other way around, and what they are basically saying is they value certainty of supply, which is critical, right? Because that's their foundation over their business model is they require the NAND to create this billion dollars that they generate as a franchise, so more to come, but we're very encouraged with the conversations overall and where the industry is heading.
I think you're getting to the right way to think about it, which is, and I think that the changes on the industry could be potentially profound. I mean, this is an industry where essentially on the supply side, we're making 10-year capital commitments in fabs, in R&D, enormous amounts of ability to supply the market. And again, even in the current situation, we're still supplying the market 15% growth every year for as long as the eye can see. We talk about our capital plan to do that, our R&D plan to do that. We're working on new nodes. The demand side essentially shows up four times a year and kind of runs an auction on what they're willing to pay for that. So you have this massive, like we're making 10-year decisions.
The demand side is showing up essentially four times a year and deciding to make a commitment or not and at what price. And I just think as more people come into this market with different business models and they look at, and also very large demand drivers, they look at this situation and probably go, "I'm probably not willing to bet my franchise four times a year that I can get infinite supply to NAND. And maybe we should think about this a little differently." And we're completely open to that conversation. Exactly where it's going to land, I think we're going to find out, right? I think we'll find out.
All this leads to the conversation of if you are more critically important, if the ordering process doesn't look like we'll check in every quarter and we'll see how much we can get because it is more strategically important. It begs the question of if you have line of sight to a better demand profile over a multi-year period of time. It would beg the question of capacity and potentially increasing capacity. You guys have been pretty clear thus far around your plans for capacity, but maybe talk just briefly so that the listeners can understand what ability do you have to add capacity today in terms of existing space, and what are your feelings towards increasing that capacity? Would you need to have line of sight to some more firm orders? Is the market kind of dictating a different trend than you have historically?
How do you think about the addition of capacity and what you would do in this instance?
So we think about the capacity in that timeframe we talked about, right? I mean, we have to get, if you build a fab, you turn it on, and ideally you don't turn it off because it's a very expensive asset. And so we think about getting a return for that over a decade. And so what we do now is we look at the market and we say, "Well, what do we think the long-term growth rate is?" And we want to supply the market, right? We want to stay very close to our customers. And I don't want to be unfair to our customers either. Our customers give us commitments to say, "Look, this is what my demand is going to be for the next calendar year." But we actually don't lock in on supplying that at a price. We do that every single quarter.
So we have views of what they need for, let's say, a year. And then we do our own analysis of where we think the market is going, and then we put a capital number behind that to supply the market. And as I said, clearly, pre-downturn, there was too much capital going into the business and too much R&D productivity going into the business. There wasn't sustainable profitability. And so you're asking a question of, "Well, what would it take to increase from mid-teens growth rate to something higher?" And it's very clear you would have to get conviction that not demand next quarter or two quarters from now or even next year is higher. You would have to get conviction that demand on a sustained basis for the next decade is going to be higher.
And certainly, I think the demand side can help with that by giving more visibility into what that demand looks like and what is the commitment to that level of demand. And I think one of the things that's very important right now is stay very close to our customers. We have great relationships with all of our customers. We want to be a very trusted partner. We are a very trusted partner to all of our customers and kind of understand what that equation looks like. And that's exactly what we're doing.
Another trend we're hearing in the market today is with bits needing to be allocated to certain areas, particularly the data center. There's concern that just given the supply footprint today that customers in PC, smartphones, consumer electronics could get shorted into next year. And you're starting to hear people talk about potential limitations to output. Do you guys have any thoughts around the industry's ability to meet demand trends into next year? Obviously, right now, I think the general expectation is for low single-digit growth across PCs and smartphones, but anything that you guys would comment on in terms of those other end markets and their ability to grow?
Yeah. I mean, overall, we're committed to the three end markets that we have, right? We have a very nice position on the consumer market, which is great on a downturn. It happens to be the most profitable market there is. We have close relationships on the edge with PCs, mobile gaming companies, and that's something we definitely value, and we're growing in data centers. So I think overall our strategy is to maximize value over the short, medium, and long term, and therefore optionality in the three segments is very important. So we're not going to dramatically move from one end market to another. We want to penetrate more the data center, but we will continue to be committed to the other end markets.
Why don't we switch to the technology side? So I think last year we were up here and talking about the transition to BiCS8. I think you pointed to 40%-50% of your portfolio by the end of the fiscal year. Can you talk about where you are today in that transition? Are you on track? And then in this environment, right, where clearly the structural demand trends are a bit different and very much in your favor, does that change your thoughts on the technology transition at all? Does it accelerate them?
So let's go to the first part of the question. Yeah, we're very excited about BiCS8. It's a great technology, wafer bonding, bringing it to market. The QLC performance that we're seeing, the power efficiency, it's just a really good node. And this is the foundation of the NAND business. We can talk about everything else, right? We can talk about what markets we're in or what controllers you're building for SSDs and all that kind of stuff, whether client or enterprise. If you don't have a great node, then it's really hard because you're building on top of something that's not sound. And this is really the magic of the JV. Between us, we're the largest provider of NAND in the world. And that means we can invest more than anybody else in our R&D.
And it shows up in both the quality of the product and the capital efficiency. We're able to get that incremental bit in the most capital-efficient way because of this R&D prowess that we have. And so BiCS8 is a fantastic node. We ended last quarter 15% of the portfolio. We've talked about at the end of this fiscal year, it'll be the predominant node, which is 40%-50%. That ramp is locked in when we talk about the CapEx plans that we have. We have to order tools and all that stuff way in advance. So that plan is kind of there, and we look forward to ramping that throughout the year.
And then I think you alluded to it earlier, just the idea that NAND, in terms of the crossover, let's just start historically, you've seen a lot of NAND crossover products, and there hasn't been much success. You're seeing now with accelerators the leaning on HBM. So you've effectively seen HBM become such a pivotal point in the market that when you have shortages, there's no other area in which companies have been able to grow. So you see NAND, you see products that you can introduce in NAND that can alleviate that stress. You've talked about HBF in the past, right? Is HBF something that is going to be in the market in the next 12 months, in the next 24 months? People have pointed to it as potentially alleviating the stress of HBM.
Why is this product going to be the one that actually works when historically you've seen some of these crossover products kind of come to market and then go away?
So first of all, I think we'll start with the premise that we're in one of the most exciting times in many generations from a technology point of view. And it's early. It's like super early in this whole idea of introducing AI. It's just like magical what's happened and what's happening before our eyes. And one of the things I learned in my last job where I ran the technology for the internet was it takes a long time for technology to penetrate the world. The world's a big place, right? So we've got a long way to go on this. And this is like just a complete incredible opportunity for innovation, right? You don't just take the architecture that you have today and just linearly extrapolate it for the next 20 years. This is an opportunity for people to come in and how do we rethink this architecture?
How do we bring new technology to bear that's either more scalable? We can move faster. How do we reimagine some of the way things work? And that's what HBF is really about. And I think we're seeing this in the architecture today where there's lots of approaches to this. There's the high IOPS enterprise SSD because that SSD is becoming more part of the AI architecture. It needs to be faster. We need faster access to that data. We can't store it all in HBM. DRAM does not have the enviable scaling principles that we talked about earlier of NAND. So you've got to bring NAND into this equation. That's happening in real time. HBF is a way to look at the inference side of it and say, "Hey, inference is a more deterministic type. It's not training.
It's more deterministic kind of read-based application." NAND is a very dense technology, as we say, very scalable 3D technology. Models are getting larger, so the amount of data is growing. We have a technology that supports a lot of data in a small footprint, so I think the insight was clearly a couple of years ago, "Hey, if you're a NAND designer, you've been told your whole life, figure out how to get more density into your product. Just give me more density. Give me more density at a lower cost," right? That's been your job for literally like 25 years. Like two, three years ago, some very clever people in the organization went to those same NAND designers and said, "Hey, build me something that is higher bandwidth.
“How would you redesign this technology where I can get higher bandwidth out of it?” I'm not going to create and nobody's going to create DRAM. We're not creating a plug and play. But can I change the characteristics of my technology that now starts to fit this new application space? That's exactly what we're doing. And that shows a lot of promise. We're talking to customers about how they would integrate that technology into their roadmaps. This is not a component play. We're not just going to build a component and put it out in the market, and then somebody's going to plug it in. This is a systems play. You got to work with everybody else in the ecosystem and say, “Hey, I'm going to build you something. What if I—” you go to the person designing the device or the data centers.
What if I give you a technology that did all these kinds of things, right? Gave you way more density, more storage than you possibly even thought you could get at these kind of specs as far as bandwidth, endurance, all these other kinds of - what could you do with that?" "Oh, if you give me that, well, I can design my system this way, and now I can deliver this great device or this much more scalable or more cost-efficient solution to the market." That's what HBF is at. And we're in that conversation with customers about how they would integrate that in. And that's an iterative process, right? You're not just going to design it and give it to you. You're going to have to go back and forth and iterate. How do we get this? How do we change the specs?
We're in that process right now with customers. We'll have the memory die in late 2026, in the second half of 2026. Obviously, you need a controller on top of that memory die that's going to tie into the system. We're designing that. That will be available in early 2027. We have a relationship with Hynix about how we do the specification of that controller integrates into the system. That's how you create a market. It's hard to create a market by yourself as a supplier, but when you get other suppliers that come to the table and say, "Hey, that's a good idea. Let's work on that together." That helps build momentum, and that's where we're at.
And as we progress in this process and we get more certainty around that, we'll have more to say about more specifics around the questions on timing and TAM and business case and all those kinds of things. But we think the technology is very exciting, and we think that the market we're playing in is just incredible. And we got a long way to go on this, and it's wide open for innovation. Obviously, the innovation that's going on in those markets right now is spectacular. And I think bringing the most scalable memory semiconductor technology deeper into that architecture is a welcome development.
So I think the highlight of this conversation has really been a change in the industry dynamic, a change in the importance of NAND more broadly. Maybe just to conclude things here, Luis, in terms of your capital return strategy and priorities, does this change anything just given the business trends in the broader market?
No, I mean, as we said before, we want to be very prudent, right? We're in a transition in the industry. We've been paying down our TLB. We started with $2 billion. Last quarter, we closed at $1.3 billion. We continue to make progress. So we're very encouraged. And we'll be very prudent and take our time so that we make the right decisions here.
Very helpful. Well, I appreciate you both being here. Thank you so much.
Thanks, Tom. Appreciate it.