Perfect. Good morning, everyone. Make sure it is still good morning. Hope you all are doing well. Really delighted to have with us Western Digital today, and we have David Goeckeler, the Chief Executive Officer, and Wissam Jabre, the Chief Financial Officer from Western Digital. We also have Peter over there sitting from the IR team to make sure we all stay honest. I guess, you know, before we kick it off, I'm going to turn it to Wissam to read some of the safe harbor statements.
Just one moment. Good morning, everyone. Hi, Ron, and great technology here. Thanks for having us. It's good to be here. Let me first start with, we will be making forward-looking statements based on current assumptions and expectations, and I ask you to refer to our most recent annual report on Form 10-K and our other filings with the SEC for more information on risks and uncertainties that could cause actual results to differ materially. We will also be making references to non-GAAP financials, and reconciliation of our GAAP and non-GAAP results can be found on our website.
Perfect. Thank you. All right, so, you know, Kirk and I, Kirk and I will just kick this off with a few questions. We'll stop in the middle if anyone has any questions in the audience, feel free to jump in at that point. I guess, you know, maybe before I kick off into all the questions, David, I know you can't talk much about the strategic review that you have underway.
Yeah.
But maybe just touch on, you know, the timeline one could expect to conclude this by, and more importantly, maybe just as you reflect back, what are the benefits of the JV? What does that enable you to do from a go-to-market base, for example, just talk about, like, what are the benefits there and maybe the timeline when it gets wrapped?
All right. First of all, thanks for having us. It's great to be here. We appreciate it. So first of all, two little different questions. The strategic review, you know, I've said last week or so, we're kind of in the home stretch at this point, so we expect to wrap it up by the end of the year. We're looking at different implementation scenarios, so in the issues that come with that, working through it. So that's the kind of timeline we're thinking about. As far as the JV, look, the JV is an enormously beneficial asset to the company. It's a great relationship. It's 23 years going now and going strong, and, you know, I think anything lasts that long because it's very valuable to both parties.
It allows us to produce at scale in a market where scale matters, so our technology roadmap is exactly the same.
Mm-hmm.
And we have a combined team that work on it every single day. It's like a hand-in-glove relationship. And we just announced BiCS8, which maybe we can talk about a little bit later. I think a major step forward. 16 nodes of NAND technology now throughout the lifetime of this, I think really just sustained excellence in innovation and engineering in the NAND market is what that JV is all about. The second step of that is manufacturing. We manufacture together, so supply comes out of the same fabs.
Mm-hmm.
We invest in those together, so we get the benefits of scale there. And then when those wafers come out of the fab, we each take our own, and we go our separate ways as far as how do we take those to market, what products do we build? Lots of different products you can build in NAND. We have different portfolios, different go-to-market. So the JV on that whole defining the NAND roadmap, the capital efficiency of that roadmap, the way we design it, what we're designing for, the manufacturing of the NAND itself, all part of the JV and all very beneficial to us and, kind of the foundation of the business, if you will.
Got it. Perfect. I know we'll talk about BiCS8 in a bit, but, you know, maybe just on a broader level, there's a lot going on from a macro perspective, and at best, it looks like it's a very uneven macro environment across geos, across product categories, if you may. You know, I'd love to get a perspective from you. You know, what are you seeing from a very high level on the demand side, the good and the bad, and then we can dig into some of the questions more specifically from there?
Yeah, I think one of the things I really enjoy about Western Digital is our visibility into the market. So we have everything from a consumer business, where we sell hundreds of millions of devices in every country in the world where it's legal to sell, and every e-tailer platform, all the way to selling products to the biggest technology companies in the world, and enormous amounts of products to those companies, and of course, distribution in the middle of that. So when we look across that business, you know, we kind of think about the downturn and where it started last year. Consumer business was kind of the first into that. We're at the point now where the consumer business is stable.
We think that business and the players in it are past their inventory correction. Business is behaving in a way we expect it to. We're heading into some seasonal strength in that business as we go into the end of the year. As Wissam has told me, no matter what happened in macro, we're not going to cancel Christmas. So, there's some good drivers in that business. It's behaving how we expect it to behave. From a macro perspective, the inventory correction was the big PC OEMs.
Mm-hmm.
We saw that last summer, went through that inventory correction. We think that market is now pretty much building to demand. Nobody is building a ton of inventory just yet, but building to demand. And then finally, the data center market was kind of the last into the inventory correction. I think we still got a little bit to go on that, right? We're still seeing some inventory correction there, some depressed buying patterns in both HDD and NAND. We think we're at the bottom of the HDD now, and things will get sequentially better throughout the year.
Mm.
NAND is going... You know, we can talk a little bit more about the NAND environment. I think NAND's going through a couple quarters of stabilization here. Clearly, we've been under shipping demand for a while now, and that industry is coming back into balance.
Right. Perfect. You know, and it's just a longer-term dynamic here. You know, at the Flash Memory Summit, I think Western Digital, you know, you folks had a keynote, and it was an interesting dynamic. You talked about the challenges of balancing bit growth with sustainable cost downs as you go forward, right? I thought it was a very interesting kind of perspective. So I'm curious, you know, what is the optimal long-term—let's just say, a margin-neutral industry bit growth rate look like in your view? And what are the sustainable cost downs in that paradigm?
So this goes to the heart of what's really important about the JV and the way we think about how we develop flash. And, you know, we can talk all we want about the flash, different end markets...
Mm-hmm.
Client SSD, enterprise SSD, what's happening in the cloud, consumer. It all starts with the fundamental NAND technology and what, how well is that built, what's the capital efficiency of it? The way the JV has thought about this for a very long time is we want to get this balance right of driving bit growth. So obviously, NAND is one of the areas where, if you will, Moore's Law is alive and well. Every node, we get more productivity per wafer, and you balance that productivity against what are the cost downs of that. And that, that goes directly into the design philosophy of the JV. We're designing for a specific growth rate, specific cost down. For us, that cost down is 15%. We've, we've met that. We're talking about bit growth rates in the, let's call it, mid- to high 20s in the industry.
So getting that calculus correct and making sure the designs deliver both, not just more bits, but the right kind of cost down. Of course, the storage market, you want to drive the cost down. That expands the TAM, always kind of this evergreen market, finding new applications, and I think this has been the hallmark of the JV. The question you started with is really staying focused...
Mm-hmm.
On that balance with each successive node of the BiCS roadmap. And if... You know, one of the things the JV really focuses on is what is the capital required to get that incremental next bit out of the, the roadmap, and, and how do we optimize that? And that's something I think we've done very well on over the years. Like I said, we've had 23 years of sustained excellence in innovation, and, you know, we went back-- Peter, you mentioned earlier, went back and looked at the numbers for us. Over the last five years, the amount of CapEx we've had to put into our to get those extra bits out is a third lower than the industry.
Wow!
That's really what that conversation-
Mm
that you're referring to talks about, is how do you get that balance right? It's not just a race to who can build the most layers. It's not just a race to, How do I get more bits out all the time? It's like getting this balance correct. I got to get the supply, got to get the costs down, and I have to do it in a very capital efficient way. That's the way the JV is focused, and we've had really years and years of results of being able to drive that BiCS roadmap in a very capital efficient way to make that equation work for the economics.
Once you get that right, then you can start building all the products on top of it, and you have a very solid foundation of being able to get the supply, high-quality product at the right amount of CapEx investment...
Right
To continue to drive the growth of supply.
Being able to get that capacity or bits at a third of the CapEx is actually phenomenal. I, I don't think I ever appreciated that dynamic, but that's impressive. You know, I guess, maybe on the SSD side a little bit, you know, last quarter, I think you folks talked about sort of return to growth, at least on the client and consumer side, that's starting to happen, on the flash side, along with you talked about, like, normalizing inventory, increasing content per device kind of stuff. And you talked about pricing kind of decline, moderating price declines, right? Can you just talk about the sustainability of these trends, and how do you think about the current downstream inventory management?
Yeah, I think this is a reflection of the... those parts of the market are through their inventory correction, are more building to demand. You know, you talked about PCs. We're clearly coming off of a pandemic era-
Mm-hmm
Of, you know, I think a year ago, or a year and a half ago, we were talking about different people and players in the industry. We were talking about 1 million units per day.
Right.
We're kind of coming off that now, back to kind of pre-pandemic levels. So I think the industry has reset to that and is shipping to that demand. And to your point, the, you know, we're seeing the NAND market work in the sense where elasticity starts to kick in. You know, across our client portfolio, we're seeing 20% more capacity per unit than we did a year ago. Across the consumer portfolio-
Mm
On portable SSDs, it's 40%. Across the entire consumer portfolio, it's 35% year-over-year growth in the amount of capacity per unit. And so, those markets, they're kind of readjusted to a post-pandemic reality of what the, what this-
Mm
What the demand is, and now we're seeing that we've kind of reset, and now we're starting to see some year-over-year growth again with that elasticity kicking in.
Got it. Perfect. Cody, I'll give you-
Yeah, so maybe just double-clicking on the NAND supply-demand environment. You know, current ASP levels are widely considered to be unsustainable. One of your competitors actually recently indicated that NAND pricing is bottoming right about now. You know, just curious how you're thinking about the timing of an inflection and the linearity toward more sustainable pricing levels.
It's, I think, you know, that we've been... I'll talk about our business. You know, we've been under shipping or underproducing demand for a while now to get the market-
Mm-hmm
Back in balance. We've talked about this, of utilization of the fab. The market is—you know, if you look at the last four quarters of our cost, of our price declines, 22, 20, 10, 6. So, you know, we're kind of moderating the ASP declines. I think we're in a period now of a couple of quarters of stabilization of the market. There's different markets inside of NAND, everything from wafer markets to extremely large customers that buy an enormous amount every quarter, those aren't all gonna move in tandem. But clearly, we're at a point where we're finding the equilibrium point in there before things start to inflect up, where you know, this kind of underproducing demand kicks in, and the market stabilizes.
Right.
So I think we're in that period. I think it's gonna be a couple of quarters, and then I think we'll, as we go into 2024, certainly, if not before, we'll have a, a better pricing environment. Wouldn't, wouldn't be surprised to see a little bit of lumpiness here as we all kind of search for that point, in the industry.
Makes sense. Maybe just changing gears a little bit, wanted to touch on SSD underutilization. I believe in January, you indicated a roughly 30% reduction in wafer starts. We understand that you continue to manage dynamically, from an utilization perspective, but can you just give us a sense maybe of, you know, where we sit roughly today, and also sort of the path and cadence to zeroing out those underutilization charges over time?
Yeah. So, yeah, as you said, we continue to manage dynamically, because we are always trying to balance our supply with where we see the demand for our products. And so since January, we've underutilized the fab on the NAND side. We continue to do so in this current environment, and the expectation probably, as we talked, is probably another quarter of this, at least based on what we see today. Of course, as I said in my at the beginning, this is dynamic, and it depends on where the supply-demand balance is for our products.
Yep. Great. And then maybe just, you know, as a derivative, you know, to that, on the cost down side, things really remain challenged near term. You know, can you maybe discuss some of the puts and takes around returning to that mid-teens year-over-year cost down target, including sort of a shallower node ramp at BiCS6, and also the timing around BiCS8?
Yes, sure. So, you know, the way I would talk about cost downs is I would strip out the underutilization effect because it really distorts what the technology does and how the cost downs really happen. So in fiscal 2023, we came very close to that sort of 15% cost down, maybe a tad bit below that. What we're projecting also for this year is to be within the same range, again, excluding the underutilization cost. And all this assumes and reflects all the node transitions that could potentially be happening, let's say, in going, the start of the next BiCS6.
As we go forward, you know, cost downs are equally important with respect to looking at capital efficiency and how the technology is being developed. So there's no reason for us not to continue to target similar type of cost downs. But it all depends also on how, when we see that supply-demand balance and when the market starts to turn up because that will influence the transition, sort of the nodal transitions on the flash side.
Sure. Yep, makes a lot of sense. I think before we continue, maybe we'll turn it over to the audience, see if there are any questions.
A little bit about the competition in China.
Maybe I'll, I'll just repeat that. I guess the question's really around competition in China and what you're seeing from that perspective.
Competition for?
NAND.
Yeah, I'm not-- I mean, we, it, it... I'm struggling with exactly how to answer that question. I mean, we compete at a lot of-- I mean, the NAND market, there's a number of suppliers. I guess the answer is we don't see anything in particular that would, that would call out, say, the competition in China is any different than competition in any other supplier we deal with on a daily basis in the NAND market.
On the flooding like, market, the past couple... They overbuild probably in the last few years.
China overbuild in the last few years?
I'm asking them. I'm not saying.
I mean, look, I mean, the NAND industry is oversupplied. I don't know if that's a particular country issue or just a general issue that we're... You know, we're certainly focused on our, as Wissam just talked about, our utilization, our fabs, to make sure that we're producing what makes sense...
Mm-hmm
For our, what we see as our demand, and that we can keep our inventory under control. I think this is something we've done quite well. Actually, during the downturn, is being able to still generate market-leading profit off the portfolio and not let the inventory get out of control. So, I think we've done a good job of balancing that, but clearly, the market is oversupplied right now and coming back into balance, as I said, over the next couple of quarters.
Got it. Perfect. Yeah, I guess, you know, you, you talked about BiCS8 earlier, so, you know, it's, it's the upcoming NAND, you know...
Right
Node. And I think it has, like, 218 layers, and without getting... I'm not the technical guy, but a lot of your peers would say: Well, we have, I don't know, 222, 238 layers. We have a roadmap going to 300 layers. I guess the assumption is more layers is better. But I would love to kind of get your perspective, you know, how is your process differentiated? Maybe you can talk about the cell density optimization, and how do you see this strategy evolving beyond BiCS8, really?
Yeah, this is something I think that deserves a fair amount of attention. And what you said there is sometimes I don't know, I think some people maybe do believe that more layers is better. More layers means more CapEx. So when I talked earlier about our technology roadmap is more CapEx efficient-
Mm-hmm.
These are the reasons why. So there's multiple layers of architectural scaling in NAND. You first have architectural scaling, like Circuit Under Array, wafer bonding, these kinds of architecture. You have lateral scaling, XY dimension, how big is the die? And then you have vertical scaling...
Mm-hmm.
or layers, and then you have logical scaling, which is how many bits per cell? Like, is it TLC, QLC? So you got literally this four-dimensional problem, and you're trying to optimize that for the variables we talked about earlier.
Right.
How do I get the right supply, at the right cost, with the least amount of CapEx?
Mm-hmm.
That's the objective. And so that's essentially the problem that's given to the engineers in the JV and say: "Okay, go solve this, and come back with something that gives us the supply we need, the quality of product that our customers expect, the right level of cost downs." And, if you can do that in fewer layers, that just means you've built a better mousetrap, right?
Right.
You've built a better product. I can do that with fewer layers, requires less CapEx to build that stack than...
Which appears...
More layers.
Right.
This has always fascinated me, that we really haven't heard this conversation in the last year.
Mm-hmm.
Because, like, having a higher stack and basically running the flag up the flagpole that says, "I have a higher stack and I spend more CapEx" is not really a great thing in a down market, so nobody talks about this anymore. But we have always been very focused on, you know, if you, you know, just maybe a very simple way to think about it, why have we been more CapEx efficient for a decade plus? Is because our engineers are able to build a product that requires fewer layers to accomplish the same thing. And so another question maybe somebody had asked, well, is your product good?
Right.
Well, our product is in all the premier products in the world, right? So it's a world-class piece of technology. There's no doubt about that. So, look at BiCS8, so that gets us to BiCS8. So, you know, I think the team went through that calculus. How do we? You know, we're going from BiCS6 to BiCS8. We need to deliver more bits.
Mm-hmm.
We need to do it the most cost-efficient way possible, and we need to deliver the cost down. So how are we gonna design a package to do that? So on the architectural scaling, let's go to array bonding or wafer bonding.
Right.
Let's move away from Circuit Under Array, where you build the CMOS, and then you start stacking the NAND on top of it in the same manufacturing process. When you do that, you start to degrade the CMOS under it. It's a more complicated process. What if we could separate these two, send one wafer through that builds all the CMOS, send another wafer through that builds the NAND stack, and then bond them together at the end?
Right.
Like, kind of sounds ambitious. Quite frankly, 2 years ago or 2.5 years ago, when the team was talking to me about this, I said, "Wow, that sounds very, very ambitious." Here we are. They accomplished it, right? So that first dimension, logical scaling...
Mm-hmm
Major innovation on wafer bonding. So what does that mean? That means I can get... First of all, the CMOS I send through is much more pristine because I'm not building NAND on top of it in the manufacturing process. So interface speeds are faster, much better product on that, that part of it. Much, much more fidelity in the product. I can also send each wafer through separately and drive the yields higher, faster.
Right.
Right? So I get faster yield ramps. Faster yield ramps means more capital efficiency, more cost downs, right? I get up those yields faster-
Mm-hmm.
I get to higher yields. So that's the first thing in BiCS8: go through this major architectural change. If we can get through this architectural change, we got a better product. We did it. It's something I think everybody in the industry is eventually gonna have to go through. We've now commercialized it at scale. Lateral scaling, XY dimension.
Right.
How do I think about the die size, XY dimension? I won't go into all the details here, but I would encourage people to look at the patents that are there and the work that's been done. Very, very clever work about how to get the XY dimension, pack more into, into the same die. Okay-
Yeah
You got both of those solved now. Now, we're gonna start talking about stacking.
All right.
So you look at all that stuff underneath, and you say: "Okay, how many layers do I need now to deliver the number of bits I need?" And we came up with 218. Right, so we're operating... And of course, you got the logical scaling, which is how many bits per cell. Is it two? Is it three? Is it four?
Right.
And clearly, we're going from TLC to QLC. So when you look at that whole equation, that's how we think about NAND architecture in BiCS8. Like, how are we gonna think through each of those dimensions of architectural change to deliver the best product? Team was able to do that, and they were able to build a world-class product with 218 layers. Right? Now-
Mm
... that's fewer layers than everybody else, and that just tells you that the innovation in the lateral scaling, the innovation in architectural scaling is outstanding, and we don't need that many layers to deliver the same product. That leads to the capital efficiency, and that's why people sometimes ask us: How can you stay competitive? 'Cause you consistently spend less CapEx than everybody else in the industry. This is exactly what I just described is exactly why. Because we have this JV, and we've got 23 years and 16 nodes of sustained excellence in innovation to build a really world-class product.
Right.
And so we're very excited about BiCS8. That's all in front of us. As we continue to move the portfolio, we're going to move a number of products directly from BiCS5 to BiCS8. We'll move other products through BiCS6 because we need the capacity of BiCS6-
Right.
Some of the capabilities. But we have all of this in front of us of ramping onto a really, really stellar node in the roadmap.
Got it. You know, you mentioned a lot of, really good stats there, but, you know, when you talked about sort of lateral scaling and, you know, you sort of said, "Hey, listen, our yields are much better than everyone else, and others have to go through this journey as well." If you think about it, like, what sort of advantage do you have from a time basis, if you may, right? Are you think you're 12 months ahead of your peer, or is it two years? It's-
I'm not, I'm not going to handicap that way. You can talk to our technical guys, and maybe they can handicap it. There's a lot of smart people in the world, right? I'm not saying we're we have a corner of the market on that. But we've we clearly have now commercialized a technology in BiCS8 of wafer bonding, and then you put all the... I kind of went through all the rest of the stuff-
Right.
because you started the question on stacking, right? As if stacking is the only issue. And, you know, one thing I would, again, just to encourage people, you cannot understand NAND architecture by just thinking about stacking. It is way, way more complex than that. So look, we, it's a long project, you know, NAND, these nodes, although they come out, let's say, every 18 months, they're in the R&D cycle for years-
Mm-hmm.
Leading up to that. So it's a significant process, and, you know, we feel very good about we've commercialized something that is going to allow us to scale the portfolio for years to come, quite frankly.
Fair enough. I guess maybe moving to a less intellectual topic, but more important perhaps, is SSD gross margin targets. I think you folks have a target of, like, 35%-37% through the cycle. Just talk about how—you know, what's the path to achieve that? How—what does mix look like, especially with eSSDs ramping up? Just talk about how—the journey to get there, what do you need, and the mix that works into that?
Yes. So, look, when those targets were put together, there was a lot of thought going through to basically get to these targets. And the way we think about these targets is obviously in a through-cycle manner. So, we still have the same targets as we laid out at our Investors Day a year or so ago, a year and a half ago. And the way to get there is very much focused on the portfolio mix.
How do we sort of mix more towards the enterprise SSD from our pretty much big presence today or big content today of consumer business, but also the nodal transitions and the continued efficiency and the ability for us to continue to transition to more efficient nodes, like, for instance, going from that BiCS5, BiCS6 to BiCS8, as well as the ability to scale. These fundamentals haven't changed for us because the long-term growth of the business hasn't necessarily changed.
Mm-hmm.
What we're going through, obviously, now doesn't necessarily translate into where the long term of the business is expected to be.
Got it. Fair enough. So those, those targets are very much alive. You just need some of these challenges, let's just say, in the NAND market to abate and-
Well, the biggest... Remember, the biggest thing that we're—that is impacting, that impacts also the gross margins, is pricing, and pricing depends on the supply and demand imbalance-
Right
Or balance, so.
Fair enough. Maybe move to the other side of the business, the HDD side of it, if you may. Let me just talk about, you know, what, what are you seeing, so, you know, especially around the nearline HDD demand trends right now, and perhaps you can talk about, you know, what you're seeing in hyperscale versus enterprise, U.S. versus China. Just anything on the drive side would be helpful.
Sure. I mean, if you, we're clearly at a point of depressed demand in HDD as well. I will say just to start, we expect sequential improvement throughout the fiscal year from here. You know, if you look at—you ask hyperscale versus enterprise, the enterprise business is pretty much stabilized, right?
Okay.
If you look at it on a year-over-year basis, it's really about hyperscale and demand there. And we have... You know, we've talked about this a bit. We have some customers that are significantly depressed buying, maybe not buying anything in a quarter for several quarters. We have other customers that are buying at a lower level than usual. And then we have some other customers that kind of move in and out of the market quarter by quarter, depending on what their inventory level is. So, as I said earlier, in the data center market, we're still getting to the end of that inventory digestion phase.
Mm-hmm.
Now, we have incrementally better visibility because we're more of a build-to-order type model now, so we're looking for more visibility from our customers. We can talk about a little bit. We've taken a lot of cost out of the system on HDD. We had a lot of capacity in the system for building client units, quite frankly, that we've now just removed from the system over the last year, year plus. So we're getting that supply, we're getting the demand, supply infrastructure more aligned, so we're asking our customers for more visibility into what they're going to buy, so that gives us some confidence that we're going to now start to work our, our way out of this.
But that—it's really that hyperscale market that is still significantly depressed in the nearline market, but we expect some improvement from here.
Got it. And Bob, is the depressed hyperscale market just a reflection of, hey, they bought a lot of stuff and they got to digest it, or are they changing how long they want to store my emails on Gmail, for example, or some other factors?
I think there—like anything in life where you have something that just—that is, you know, that, that is this consequential, it's a culmination of a number of things. I'll give you my point of view. One, I think there is, you know, it's extraordinary time coming out of a pandemic-
Mm-hmm.
And now going into a lot of macro uncertainty. So managing the dynamic between those two has been a sharp inventory correction in all the markets we've seen. As we talked earlier, we saw that in the consumer market, where people that have consumer-facing businesses, you know, it took them a while to work through the excess inventory they had accumulated. In the PC market, we're talking about going from a significantly elevated number of units now to a more, you know, kind of more pandem-
Mm-hmm
Pre-pandemic level, number of units. You can imagine the transition between the amount of inventory you were holding for the higher to get down to the lower, you have to go way below true demand for a while. In the data center, clearly, the same thing happened.
Mm-hmm.
There was some over-purchasing of or oversupplying, whatever way you want to look at it, so there was a lot of inventory. Of course, these customers are kind of markets of one, very limited. They're not all the same, so there was differences among them. That's a big impact, and that was the last market to start to correct, so it's not surprising we're still kind of working through, we're working through it here. On top of that, in that period, of course... Well, on top of that, you had the cost of capital environment has changed on us.
Right.
Right? It's like working capital is more expensive now than it used to be. I think every company in the world is managing OpEx a lot more tightly, maybe in ways they haven't before. And so when you just have that much more cost visibility or cost discipline in your business, that's going to impact how much inventory you're willing to hold, how fast you're willing to accumulate it, those kinds of things. On top of that, you've got macro uncertainty, like what's going to happen in the future? Is now the time to take a big inventory position? Like, what's going to happen with, you know, what's going to happen with different economies? Are we going to have a soft landing? Is there going to be a recession? Is there not? What's going to happen to major economies around the world?
Are they coming back as fast as they used to?
Mm-hmm.
A lot of uncertainty, another reason maybe to be more conservative. Then I think you've got a priority question a little bit, you know-
Right
Which I'm sure will come up, where a lot of people spending on building out, infrastructure for, AI, which is extremely important. I mean, I think that's great news for us. We're kind of the second wave beneficiary of that.
Yep.
And I think this, I think this is the brilliant thing about the cloud, that the cloud can deploy really exciting new technology very quickly and make it available to literally billions of people, and then they can start using it, and that's going to generate more storage. So I think you put all of these things together, and then the utilization is part of that.
Mm-hmm.
You know, when I'm in an environment where I'm more concerned about how much I'm spending, how much I'm going to be willing to hold, of course, if there's optimizations I can make, I'm going to choose to make those.
Right.
That's going to push the recovery out a little bit. So I think all of these things are going on at once. We're working through it. It's a longer-than-usual digestion period, if you will-
Mm-hmm
Because I think there's probably a little bit more going on than just digest-
Right
Digestion. But, you know, the good news in all of that is, one, we're, we're getting through it. Two, I, I don't believe in any way possible the long-term thesis for storage has changed, right? Storage is 29%-30% of the TAM in the data center.
Mm-hmm.
Don't think that's going to change. I think that we're on the precipice of a major new technology revolution in AI, that is going to enable us all with tools to kind of automate the creation of data, if you will.
Right.
That's one way to look at it if you're in the storage business, like all of us are being enabled with a set of tools where now we can generate data even faster than we could before. I think that that is going to drive demand for storage. I think we're going to get through this inventory correction. We're kind of resetting the business for what we see as that continued long-term growth. We'll get through that, and I think we're now going to start working our way out of it. You know, we're going to have both businesses with the same demand drivers in place, and kind of better structural cost dynamics-
Right
Inside those industries, inside those businesses.
Yep. Totally fair. You know, a few minutes ago, you talked about this build-to-order dynamic, and you kind of mentioned it, but I think on the last earnings call, you and one of the other only HDD peer talked about this as well, to implement more of a build-to-order strategy for me, right? Can you just talk about how does that impact, how do you see that impacting your utilization rates, your pricing going forward? And then, how is that different from perhaps the LTA stuff that y'all, we all did through the pandemic?
Well, yeah. So, when you think of what we've done over the past year in the hard drive business and the manufacturing footprint, we've taken down our capacity. We've taken capacity really out of the system. In many ways, we continue to do so. And so, we've basically now resized the business to where we see the long-term demand going forward. And so for that to be able, for us to be able to manufacture and deliver product in a very efficient way for our customers, we do need that visibility to be able to.
Mm-hmm
Because the manufacturing cycle is quite long. We do need that visibility to be able to provide the product on time. In the past decade plus, there was quite a bit of capacity that sort of was more targeted towards the client side of the business that was used to manufacturing for the products for the capacity enterprise on the nearline. And that capacity now has been taken out of the system. So we resized it based on where we see the demand going forward. That would allow us to be much more disciplined as well in our capital investment...
Mm-hmm
And, to be able to sort of, maintain the, supply and demand for our products, in, in a good balance.
Got it. Perfect. Great. So I think we'd probably be remiss not to hit on, you know, the flash versus HDD debate, at least for a question or two.
You'd be remiss.
So, you know, would like to sort of discuss this from a few angles, perhaps. You know, a flash array vendor has suggested that HDDs will be extinct in a few years as QLC flash gains momentum.
That's right
And cannibalizes near-term, nearline demand. You know, Western Digital is in a unique perspective to, you know, speak to this with HDD and flash in the portfolio. So would love to know, you know, your perspective both on the current market backdrop and how you sort of see that also evolving over time, given AI proliferation and, you know, how that could potentially impact flash versus HDD use cases, and also sort of the total cost of ownership consideration.
So, look, I'll just say, we don't see any change in the architectural structure of the data center. And I'm talking, like, the mass-scale data center, the biggest data center operators in the world, this thing where we sell 30% of the storage into that, or, storage is 30% of the total TAM of that. That market is roughly 85%-86% HDD.
Mm-hmm.
The rest is flash. The way that architecture works is it's use case driven. It's not like you just have this big sea of storage, and it's, like, all interchangeable. You have different storage for different types of things, like you have a big set of storage for boot drives, for example, for all your servers, right?
Mm-hmm.
You're going to store all the images for your servers, like, that's going to be stored on flash. All of your object storage, files, videos, the kind of the mass amount of storage, that's all stored on HDDs. So it's not a question of the simple substitution of price or cost. It's that, that's part of it. It's about access times, it's about when do you need it? And HDD, for the vast majority of those use cases, provides a just much more economic way to store.
Mm-hmm.
And that's not going to change. Both technologies have a very robust roadmap that is going to continue to drive the cost per bit of storage down. As long as that happens, there will be minimal substitution between the two. And again, there's a lot of conflation that happens here because on the device, on your laptop, on your mobile device, they're pure substitutes, right? You used to have a hard drive in your laptop. That market is diminishing quickly,
Right.
In fact, apparently faster than people expect. You know, the first iPod that came out had a tiny little hard drive in it.
Yep.
That doesn't happen anymore.
Mm-hmm.
All that has been substitution of NAND. In the data center, it's a very different equation, where there are highly complementary technologies, where customers have defined their architecture in a way they use the right product in the right use case. And so this whole debate is this like, kind of, it's a great, pithy little statement, but it's just not tied to reality. And, you know, I think if you go talk to customers, they're going to give you a very different view of this. Now, clearly, like, we're a big fan of flash. Don't get me wrong. Like, we think flash is great. There's going to be more flash in the future, and clever people are going to think of new use cases to deploy flash.
Right.
That's not going to come at the expense of HDD. And then, you know, the final thing I'll say is just if... like, just do a very simple spreadsheet, and you will find that the economics of this, I guess, assertion simply don't make any sense, right? I've actually done that. I'll let you go do... It's an exercise left to the reader. But let's say the HDD industry is going to ship 800-900 exabytes this year. That's going to grow at 25% CAGR. The entire flash industry, everybody and every fab in the world, is going to maybe ship 600 exabytes, okay? And that's going to grow at 25% CAGR.
Mm-hmm.
So go put that in a spreadsheet, and then figure out in five years how many zettabytes of storage have to be shipped, and the amount of fabs that would have to be built in the world to support that, and you...
Right
will find it is a preposterous assertion. And the only people I'm going to believe on this is if a customer tells me that, or if somebody that's actually investing the capital to build the NAND
Mm-hmm
Tells me that.
Yeah.
There's not a single data point from either one of those two that would tell you that's true.
That is...
The rest of it's just a big echo chamber of people talking to each other about, like, what they think about, like, "I think this is good. I think that is good." It's just not grounded in the reality...
Right
of the market.
David, to your point, the vendor that's making that statement, who you know well, ships 17 exabytes a year, so of the 800, for the drives in total, so-
Okay, well...
For them to ramp that up, 17-800...
There's a long way to go.
It's a long way to go to your point.
Look, I mean, there's a lot of smart people in the world. I know, you know, and there's a lot of very good products in the world, and I didn't actually. That's an interesting number. When you have a market that's 17 exabytes, you can grow to 30 exabytes, right? That's good. Like, that's.
Right
All, you know, good. You can grow to 50 exabytes. That's very different than a market that's a zettabyte.
Right.
Hundreds, and hundreds, and hundreds of exabytes, and the people investing to deliver hundreds, and hundreds, and hundreds of exabytes. That's the business we're in.
Yeah.
These are different things.
You know, we're almost up on our time, so maybe I'll ask you the last one here. You know, I think prior to the current downturn, Western Digital talked about, like, $6-$6.5 billion kind of gross debt target number for you folks to re-engage with, you know, capital returns, if you may, right? I'm curious, is that still the right framework, or has the depth of this correction changed that or prompted you to reevaluate it? And, you know, I guess in theory, when you do elect to re-engage in capital returns, have you considered what method would be the most optimal one?
Look, on capital allocation, when we put these targets out there, those were also based on an analysis and thought about how the business would perform through cycle over several years. We haven't moved from these targets. Our capital allocation framework is still intact. The way we think about capital allocation is still similar to what we've laid out, which is investing in our business, paying down debt, and then returning capital to our shareholders. You know, we are in a very civil, severe, sorry-
Mm-hmm
Cyclical downturn, and so, yeah, the realities today are different. However, the, the targets and the capital allocation framework have not changed.
Got it. We're up on our time here, so maybe I'll pass the mic back to you folks and see if there are any closing remarks, anything we did not touch on that you want to make sure gets fleshed out.
Yeah, look, we appreciate the time, and we appreciate the discussion. A lot of really great topics. You know, I'll just reflect on we've... You know, Western Digital has changed a lot in the last three years.
Mm-hmm.
We've made a lot of changes in the business, a lot of changes in the portfolio. The downturn is, you know, it's a severe downturn, but I think it is showing that the franchises we have and all the changes we've made...
Mm-hmm
Are performing very, very well. In the drive business, we're gaining profitable share. That's something you always want to do. It's hard to do in a technology business. In the NAND business, our business is performing from our ability to extract gross margin out of the business. Although it's negative, it's still, I think, 30 points ahead of where nearest competitor is.
Right.
So both businesses are performing very, very well. On top of that, we're making structural changes in the business on HDD, to take cost out, to resize the amount of fixed costs. It's the lowest it's been in well over a decade. That will support profitability as volume comes back, which it will. We've talked a lot about the NAND business and how focused we are there with our JV partner, making sure we have foundational technology that is the most capital-effective, world-class product in the market. On top of that, we build a very robust portfolio where we can sell to literally every consumer in the world, and we do, all the way through to the largest hyperscale players in the world, and have a portfolio that spans that. And so we feel very good about where the business is.
We think as the market recovers, the business is going to perform very well. On top of that, we have an active strategic review going. We know it's been going a long time. We're looking very, very closely. We'll have some more to say about that over the next several months as we work our way to the end of the year, but if there's a way. We do think the business is undervalued. We're looking for a way to unlock that value. When we develop conviction around that, we will implement it, and I think that, you know, just in general, I feel very good about, about where the franchises are, where the business is, and our ability to, to create value in this business.
Perfect. Thanks a lot for your time. Really appreciate it.
All right.
Thank you.
Thank you.