All right. Good morning, everybody. I'm Krish Sankar, the TD Cowen analyst covering Western Digital. We're very fortunate enough to have Dave Goeckeler, the CEO, and also Peter Andrew, head of IR for Western Digital. I'll pass it on to Peter to give some forward-looking statement disclaimer, and then we'll go to Dave for the meat and bone.
So first, thank you very much for having us today, Krish. Let me just read a quick statement, and then we'll get right back to Q&A. During today's conversation, we may be making forward-looking statements, based upon management's current assumptions and expectations. These forward-looking statements are subject to risks and uncertainties. Please refer to our most recent financial report on Form 10-K and/or other SEC filings for more information on the risks and uncertainties that could cause actual results to differ materially from our expectations. We'll also be making references to non-GAAP financial measures, and a reconciliation between the GAAP and the non-GAAP financial measures can be found on our website. So with that, let me turn it back to you.
Awesome. Thanks for that, Peter. Dave, again, thank you very much for your time.
Great to be here. Thank you.
So, you know, clearly it's exciting time for Western Digital. A lot of things happening. You're beginning to see, you know, a cyclical recovery in the business. You're beginning to see some products take traction. So obviously, some special situation event's gonna happen, things like that. So purely from the fundamental business side, I'm kinda curious, you know, you begin to see demand start improving. Pricing has been pretty good for a couple of quarters now. How do you envision that, you know, for the broader business, both NAND and hard drives? Then we'll get into the specifics. But when you look through the rest of this year into next year, you know, is there a way to think about demand versus price, or one step preceding the other?
So, yeah. I mean, we feel really good about where the business is. I mean, we can talk about the market we're in and kinda the, the cyclical dynamics of that. But it really starts with the product portfolio. And I think if you look across both businesses, our product portfolio is in the best shape it's ever been, and it's very resonating very, very strongly with the market. If you look at the HDD business, I think our ePMR, UltraSMR strategy, and the drives we're building are the lowest TCO, highest capacity products in the industry, you know, that being very well received by customers. That led the last quarter. You know, we now have the largest, fastest-growing, and most profitable drive business in the industry.
So that's a great place to be as we go into the second half and look forward to 2025 in the cyclical demand recovery we're seeing. We're seeing that demand come back into strength for our portfolio. Feel really good about the way we managed the downturn. It was, you know, kinda I don't know what the best word is, historic, but, you know, deep as we've seen in a long time, whatever you wanna call it. But, we were able to manage through that and also keep a firm eye on our investment and make sure we emerged with, a very, very strong portfolio. And on the flash business, you know, again, we're now launching and I think gonna lead the QLC transition and client SSDs.
We have an emerging enterprise SSD portfolio where we did a lot of work before the downturn that now as that market comes back, and I'm sure we'll talk about that in a little more detail. We think we're well positioned. Obviously, the consumer franchise and the SanDisk brand is still very strong. So we feel very good about where the portfolio is, and we're seeing a cyclical recovery in the business. I think for us, we're coming back into strength on the product side. We're coming back into a business where we've reset the cost structure, and we've taken some, you know, a lot of fixed costs out of the business. You're seeing that we're able to drive more profitability at lower revenue points.
So, we feel very good about going forward and looking into the second half where we continue to see good supply-demand dynamics in both businesses.
Mm-hmm.
You know, balance to tightness, I would say, in both businesses. And, you know, it's just in general, we feel good about the results we've delivered. I think we've delivered best-in-class results in both businesses, and, and we feel good about that going forward as well, about how we're positioned.
Got it. Got it. That's very interesting. If you look at, like, both the business over the last six months, you know, it looks like the hard drive business troughed around August, September, NAND pricing troughed around August timeframe. It seems like on the NAND side, you've seen a very strong velocity in pricing improvement.
Right.
While hard drives have been much more measured, I would say, since that cyclical trough. I understand forecasting this is very difficult into next year.
Yeah.
But if you had to take an educated guess, yeah, I mean, how do you expect the momentum for both those pricing to continue?
Yeah. So first of all, on the premise of your question, I mean, we in NAND, we were clearly, I mean, peak to trough, we were down over 50% on pricing. So, you know, when you're down that far, you gotta see 100% pricing increase just to get back to where you were. So it's not surprising we had a stronger snapback than maybe you would have seen in other cycles where you weren't down so low. And you're right. It's notoriously difficult to predict the cycle in this business. And the way I think about it, I'll get back to your question. But the way I think about it is what we're trying to do in the business is we're driving to have the lowest costs.
Mm-hmm.
In the market as well. Like, that's really, really important. In a market where it has commodity-like pricing dynamics, not commodity products, products that are very, very sophisticated, but kind of commodity-like pricing dynamics, making sure you have the lowest cost position is always a very, very strong strategic position. And I think we have that in both markets, and we can explore that in a little more detail. And then build enough agility into the company so that we can respond to the market at the pace that it's moving. And we're in a market that moves very, very quickly. You know, sequentially, we see significant price changes in our market. So it's important for us to be able to move faster than the market.
That's something we've done a lot of work on over the last three or four years, is reorganizing the company, putting that agility in the system, get our costs in the right spot, get the portfolio in the right spot where the products are very attractive, and then run that model and deliver the best through-cycle profitability no matter where we are in the cycle. And I think we saw that in the downturn. We performed very well in the downturn, and I think we're seeing us perform very well as we come out of that. Now, going in, you know, like, you, you kind of, prefaced your question. It's very difficult to predict, this market, and we don't like to talk about future pricing all that much. But it's, it's clear that the market is tight, right? We do see tightness across both markets.
I think in HDD, that's probably a more fundamental situation where we aggressively took fixed costs out of our business, in the downturn. So we have, you know, I think it's an industry that's been going through this client-to-cloud transition for well over a decade. And the client business is where, you know, we needed the industry or we needed to produce hundreds of millions of units to go into a market where you need to produce tens of millions of units. And that transition has taken a long time. And I think we're now, you know, the downturn finally took the costs out of the system. We're now ramping back, as demand returns, into a much more balanced environment where the ability to produce is more aligned with demand on an ongoing basis.
So I think that's where we want our business to be. We want that balance in the business. And I think we're seeing that as demand ramps back. And in the flash business, I think, you know, we've gone through several phases of the recovery. The first phase of the recovery was all the inventory on the supplier's balance sheets moving off. That happened relatively quickly. We're now in a phase where more fab utilization is coming ramping back up to meet demand. As we work through that, we're gonna get into the phase of more capacity's gonna re-require for more supply. And that's an area where, if you look at the amount of Capex that we've been putting in the business, given the downturn has been quite depressed for quite some time. I mean, we cut our Capex 30% in FY2023.
We cut it another 50% in FY2024. So, and that's where we're gonna need to see better long-term profitability for that investment to come back. And so I think we're gonna see how that plays out as we go through the second half and through 2025 about how this shapes up as a long-term demand situation to see how much when and if and when and how much CapEx will come back in to increase the supply.
Got it. Actually, I think, Dave, you've been very vocal about this, about, you know, a disciplined Capex, especially not just for you folks, but for the industry in general, given how deep the cyclical downturn was last year. And it seems like, so far, both you and your peers are being very disciplined in supply additions on the NAND side. How much of that do you think is some of your competitors so much focused on HBM that they're not investing in NAND, or do you think they're both mutually exclusive events?
You know, I can only manage our business. So I don't, you know, it's hard to talk about what the industry is doing. Other people are making different calculations for their business. But I think, you know, some of the dynamics, I think, we're seeing in the downturn in the NAND business. You know, NAND is an interesting business because, on one side, there's a lot of elasticity. And if you drive the cost down, you'll see more absorption and more demand. And then on our side of it, there's still, it's a, you know, a part of the semiconductor ecosystem where Moore's Law, if you wanna call it that, is alive and well. We can produce a new node, and we get 50, 60, 70% more supply for the same wafer.
So, you know, maybe in the 2D era, you could just continue to roll the nodes forward, and the market would just absorb. There was enough elasticity to just continue to absorb all of that additional supply. In the 3D era, I think maybe one of the lessons that we take from the downturn is, you know, it's too easy to oversupply the market. You can ramp these big nodes forward. They take a lot of Capex, but they also produce—you know, the scale of the industry is much larger now—so they produce an enormous amount of additional supply. So how we think about that equation.
Right.
Has to change. You can't just, the way we think about it is changing. You can't just rely on this elasticity question is gonna eventually absorb everything that you produce. So the way we're looking at the market is trying to really understand that long-term demand a little better, much more bottoms-up kind of analysis of all the big markets. Go through the smartphone market, the PC market, the data center market, the gaming market, the IoT market, automotive. Look at all of it. Look at what demand's gonna be in the future. Model that, predict it, and have that guide how you're gonna invest more in this business.
Got it.
And you know, not just roll the nodes forward because you can.
Yeah. And, you know, and I do remember, I think, you've spoken about your technology roadmap, like BiCS6 on the NANDs are going to BiCS8. And if I go back, look at, like, I think a couple of years ago, when you did the analyst day, at the time, one of the goals was to increase eSSD share.
Right.
It seems like you're progressing in the right direction. I don't know whether you still hit the target of 20% or something. Maybe it's a longer-term target. But how much of that is predicated on BiCS8 versus how much of it is more driven by the end-market, you know, cycles and things like that?
It's not dependent on the nodes, right? So we qualified our enterprise NVMe enterprise SSDs at multiple web-scale providers just in time for the market to go into a downturn, quite frankly, and they stopped ordering. Now, the good news is, as we go into the second half, we're seeing that demand return, which is good. It's kind of the last market, I think, where the inventory correction is coming to an end. As you said, we saw it in capacity enterprise HDD. Now we're, you know, enterprise SSD has been a couple quarters behind that. But we're starting as we go into the second half to see that demand return. That's gonna be a good tailwind for us, as that portfolio still is qualified to those customers and will have a chance to start to ramp those sales.
We also have a new product on the market, which is more, we think, compute-focused PCIe Gen 5.
Mm-hmm.
Enterprise SSD that was built for one of the web-scale players, but we're finding much broader adoption for that now as the AI, there's much more AI demand-driven driving enterprise SSD. We can talk about that in a little more detail. But as we look into the second half, we see enterprise SSD being driven by, you know, a couple things. One is this return of data center demand that has been in this inventory correction for quite some time. And on top of that, we see this, you know, AI training infrastructure that is really pulling high-capacity and high-performance enterprise SSDs, really starting to drive the business as well and drive demand.
Mm-hmm. Got it. And I think, Dave, you've, I mean, you touched upon this on the earnings call. You know, obviously, AI is such an exciting area. And, you know, if I look at both your businesses, you know, it makes a lot of sense that both nearline hard drives and eSSDs should see improving demand or increased demand because of AI. The pushback I get from some investors is that, you know, it's the same nearline hard drive, the same eSSD, but they're used for AI or non-AI applications. So how do you know exactly whether it's being segmented for AI or not? So I'm kinda curious to know your thoughts on that.
So I think so first of all, in the general big picture, I think that, you know, AI we're in a cyclical recovery, and we have this big secular tailwind now. We're all trying to figure out, to your point, try to decompose what's the cyclical recovery, what's the secular tailwind, 'cause the secular piece you wanna understand does it how much does it drive sustained demand. I think on, on, capacity enterprise hard drives, you know, I think it's primarily a cyclical recovery at, at this point.
Mm-hmm.
It's hard to tease apart, but, you know, there's just not enough deployment yet of the models to really drive an enormous amount of output from them. And we're more in the training phase of that than the inference phase on a broad scale. But we have a pretty strong belief that, you know, there's two big-picture dynamics. One is, you know, these large language models are just another way to take all the data you have stored and figure out a new way to monetize it and make it productive and use it for, you know, productivity enhancements or whatever it's going to be. And so the value of storing more data has gone up. And then you're gonna once the models get widely deployed, you're essentially automating the creation of data. And most of that data is gonna be stored on hard drives.
So you have this kind of virtuous cycle that's going on. So we think it's gonna be a good secular tailwind for, for the drive business. It's hard to decompose. If you asked me how much.
Mm-hmm.
Does it change the exabyte growth, we don't know just yet. But I think right now, predominantly, what we're seeing is a cyclical recovery in.
Mm-hmm.
In demand. I think on the flash side, enterprise SSD side, I think it is becoming clear. We are seeing demand from customers that are building training-focused AI model training-focused infrastructure. And there's a process for how that works. And at certain points of the process, you need different kind of enterprise SSDs and different capacity points and different interface speeds. And, you know, we're seeing demand for that as we go into the second half.
Mm-hmm.
It's, you know, we need to get a little deeper into it to be able to fully quantify exactly what how much it's gonna change the market.
Mm-hmm.
From a TAM perspective. But it's very easy to see that there's a new use case that the world has developed for enterprise SSDs, and people are building infrastructure to that use case. That use case is training models.
Got it. And then, you know, there's obviously another part of the debate, which is kinda like, you know, over time, will, you know, eSSDs cannibalize Nearline hard drives? or do you think there is enough piece of the pie for both of them to, like, you know, swim in their own lanes and be happy?
Yeah. I think that that term you use is a term we use inside the company, swim lanes. Like, you're trying to figure out what the use case is that you're driving. And, you know, we don't see a cannibalization story. We see a complementary story. They're used for different parts of the data storage ecosystem. And, you know, data that's required in things like training a model or data that requires very fast access speeds, you're gonna put it on enterprise SSDs. Data that doesn't require that is going to be on HDDs. It's just an economic equation. As long as we can continue to drive better TCO in both technologies, I don't see that fundamental architecture changing.
Mm-hmm.
If you're storing an enormous amount of data, it's extremely important for you to put the data on the right piece of equipment that's the best economic value proposition for how you're going to use that data. And so we do not see cannibalization of enterprise SSD across capacity enterprise HDD. We do see them both growing.
Mm-hmm.
We do see enterprise SSD probably growing faster because we have these new use cases, like I just talked about.
Mm-hmm.
AI model training. Nobody used to do AI model training on HDDs, right? It's a new use case. It's enterprise SSDs. The data that those models turn out in the inference phases, I think, primarily goes back onto HDDs that's stored in mass-scale data centers.
Got it. Very clear. That's very helpful. The other thing is also the fact that, you know, I mean, like you mentioned earlier, you know, you're beginning to see this nice cyclical recovery. There's probably a structural aspect to it. Your margins are improving pretty nicely. You've been very disciplined on capacity. What are, like, kinda the leading indicators you would look forward to before adding capacity?
Well, I mean, look, I think it's just very early to be having this conversation. We've just come through a downturn that lasted many, many quarters, right? 5, 6, 7 quarters. We've had one quarter now where we're starting to get back and generate free cash flow in the business. It's not the time to be thinking about how we start investing a whole bunch more CapEx. So, you know, our model is very clear: 35%-37% through cycle, on flash.
Mm-hmm.
We can get the HDD. But on flash, $35-$37, we've been below that for quite some time. We're gonna need to be above it for quite some time to be able to average out to $37. So when we get above it, we've lived in that world for a while. We've had the ability to recover from this significant downturn. Then we get conviction, as I talked about earlier, looking across all the markets we're looking at, that this is sustained demand.
Mm-hmm.
For years to come. Like, we can't invest for something that may be 1 quarter, 2 quarter, 3 quarter. It's gotta be sustained demand to go out and, you know, put a bunch of equipment in a fab. These are huge capital investments. If we get conviction that there's sustained demand for that additional supply, then we'll start to think about that. But we are far, far, far away from that.
Got it. Got it. All right. I wanna I wanna ask about CapEx increase. Let me ask about OpEx then.
Okay.
You know, now that, you know, things are getting better, obviously, variable costs up, you know, might get added. So how do you think about Opex in a, you know, going forward?
Yeah. So we've done a lot of work over the last four years. I think when I first came in the business, we were over $800 million a quarter on opex.
Mm-hmm.
And now we're at the low 600s. So we've taken a significant amount of Opex out of the business just to be and at the same time, we've built the best portfolio we've ever had. So I think our level of being able to put the right investment in the right place in the business at the right time to deliver the best products and drive profitability is exactly what we're going for. And again, this is one of the things I feel very good about. The company is just much better, much more efficient. Any investment we put in the business, we have high degree of confidence what the return is going to be for that additional investment. So right now, you're starting to see us kinda rebound a little bit off the lows because there was no variable comp in the system.
I mean, this is a statement of how deep the downturn was. I mean, we, we not paying employees bonuses for 18 months, you know, significant amount of cutbacks in the business. So you can't run a business like that on a sustained basis. So now the business is performing better. Variable comp pieces are coming back. You know, we expect that to be a sustained thing. We don't wanna go back to where we were. But we're gonna be very disciplined about this.
Mm-hmm.
We're not gonna grow Opex faster than our revenue grows. We're gonna, you know, the most important thing to me is to have confidence that when I make an investment in the business, I know what return I'm gonna get from that. And like I said, over the last three or four years, there's been a complete sea change in the way we've organized the company and the way we track this and the way we're able to measure it in the precision that we understand what is the value of the marginal investment we put in the company, and what's the value of the marginal investment we take out of the company, and make sure we understand that extremely well. And then, you know, we'll set the Opex at a level where we think that we get the best long-term profitability of the business.
We are in growing markets. We have opportunities.
Right.
Right? We have, I mean, I think one of the really strong, impressive things about Western Digital is the R&D team. We have an incredibly enviable patent portfolio that a lot of people pay us money for to use. Our HDD team, I think we're seeing the strength of our HDD team in the market right now and the decisions they've made over the last 10 or 15 years of how to navigate an extraordinarily difficult technology transition and do it extraordinarily well. So, we're in growing markets. We have opportunities. We're gonna invest in those. But we're not going to, we're gonna keep a very strong eye on making sure we are getting the right returns for the investments we make. And it's the best, you know, short, mid, and long-term decision for the company.
I hope one thing that everybody has is a lot more confident that as the management team makes those decisions, there's much more precision in our ability to generate long-term, value creation for any investment we do put in the business.
Got it. Very helpful. Let me just pause to see if anyone has any questions. Oh, go ahead, Erica.
Just a few comments here that, you know, seeing more training demand being pulled in from enterprises, and inference is likely more eventually. But I guess NVIDIA has already said that they, they think, you know, 40% of the compute being done on their GPUs, is, is from inference. So I'm just wondering if that's something you haven't seen yet. And just, you know, obviously, if you think about the data that's, you know, going to be refined and put back, onto the model and hence onto storage, there might be less of that going back, for inference, and stored, and maybe there's a, a timing difference as, as people sort of figure out what data they do wanna store.
Yeah. I think, you know, and that's so I think the comment is very well made. I mean, I'm not saying there's not any going on. I'm sure there's a fair amount happening. I think we're, like, just getting to this phase, though, of it. I mean, we're in these huge markets, right? So we're just now talking about the AI PC and what it is and how it gets deployed. I saw something last night where somebody is now talking about the smartphone being called the IntelliPhone because it's gonna be ready for AI and machine and artificial intel, you know, machine learning type thing. So that's when I talk about the impact of it, I'm talking about very broad impact on huge markets where we can start to see it penetrating very, very deep into the ecosystem.
I think that although there's a lot going on, I think that phase is still in front of us where we're gonna be able to see what the impact is on these very, very large markets.
All right. I think we are kinda out of time. So but thank you for that day. Very helpful. Super helpful. And thanks a lot, Peter. Really appreciate it.
All right. Thank you. Appreciate your time. Thanks, folks.