Western Digital Corporation (WDC) Earnings Call Transcript & Summary

Okay. I think we are live now. So we have Western Digital. We have Siva Sivaram, who is the President, Technology and Strategy. We also have Peter Andrew, Vice President of IR. And I think, Peter, you're going to start with the safe harbor, and then we'll hop into Q&A. Welcome.

Yes, thank you very much, Wes, and thank you, everyone, for joining us today. Before we get into the Q&A, I do want to remind everyone that we will make -- or will potentially make forward-looking statements, and I ask that everyone please review our SEC filings for the risks and uncertainties associated with these statements. In addition, we might also make comments related to non-GAAP financial measures. And as everyone knows, there's a reconciliation on our website to reconcile the non-GAAP to the GAAP financial measures. So with that, let me turn the mic back over to Wes.

All right. Thanks, Peter. So this is an open Q&A. It's a fireside chat format. If you do have questions, feel free to submit them through the portal on the website, and I will take a look at those and try to get them in. I was hoping, Siva, to begin with, I know you came from the NAND side of the business, and you have a lot of experience there. So I would love to start with an overview of the joint venture. That's where you get all of your flash. You don't own your own fabs. Can you talk about, maybe at a high level, how the joint venture works? And why it was started and why you operate that way?

So the joint venture with Kioxia, originally, the joint venture started with Toshiba and then after with Toshiba Memory and now with Kioxia. It's over 20 years old. We have built 2 mega fab complexes, one at Yokkaichi and the second one now starting out at Kitakami. We develop technologies together. We transfer the technologies to manufacturing, and we manufacture together flash. We produce close to 800,000 wafers per month between us. And this development capability has gone on successfully for a very long time as we have moved from 2D NAND to 3D NAND. What it allows us to do is to pool the resources in development, get innovation from both sites to come into play, we get to have the scale of manufacturing we need together, and this allows us to be able to bring technologies to the market much quicker. It has worked very well for a long time. It is a lot of effort keeping the 2 together, so that the companies don't diverge in their respective ways. We both are committed to the JV, and this has worked very well over the long term. And we intend to continue to do this for the long term in the future.

Can you talk about how it's been so successful for 20 years? It's a really long time for a tech JV to operate.

It is true. It is true. I mean it is not an easy journey ever, as you would expect. I mean we are an old married couple, and we have our squabbles. But we do produce beautiful offspring, so we do end up staying together. And there is a lot of effort that goes from both sites to keep the JV active. Communication, as you -- in any marriage is the key. We spend about -- in pre-pandemic days about 30 or 40 engineers and senior managers travel over to Japan every month and the next month, 30 people or 40 people come over here and we sit face-to-face. When we do a design, we share resources. When they are short of layout resources, we'll send it to them. When we need verification help, they will come and give it to us in development. In manufacturing, we have manufacturing engineers from our side, thin film engineers, etch engineers, implant engineers in the fab together, so that we are watching yields and particles and volume and logistics together. So this is truly a very closely well-knit joint development and joint manufacturing.

Okay. Interesting. Can you help us understand the decision to build at a new site from -- moving from Yokkaichi to Kitakami? And why did you have to move that complex?

Yes. So it is a complex decision, and both sides took a lot of time deliberating before we chose to build the next site in Kitakami. We have had 6 fabs that have been built, I may be wrong on the count, but about there where we had the original fab 2, 3, 4, 5, and then 2a, 2 became new 2 and then 3 and so on and -- but we are running out of space in Yokkaichi, and we did want to have another site to get going. We wanted to make sure that we are able to leverage our scale effectively, we're able to develop in Yokkaichi and transfer to manufacturing very quickly. One of the biggest advantages we have is that the development and manufacturing happen in the same place in Yokkaichi. The development pilot line is embedded in the manufacturing side. So literally, there is nothing called a transfer time. It's instantaneous. When the technology is developed, the next day, you can have wafers starting the manufacturing. We wanted to maintain as much of that in going to a site close enough. So we don't lose that flexibility. We even now transfer equipment, wafers, people back and forth between the sites. As we run out of space in Yokkaichi, we grow in Kitakami. As the new BiCS3, 4, 5, 6, 7, 8 happen, we need more space, and we can maintain the development in Yokkaichi and still able to get the space we need in Kitakami. That's the reason that these 2 have worked out very well together.

Okay. That makes sense. I do want to come back to this base topic in a minute, but maybe we can do that as we're talking about some of the flash road map. One question that comes up a lot is Kioxia may be doing an IPO this year or next year. Does that have any impact on the joint venture?

If anything, it has got a very positive impact on the joint venture. We actually like the fact that Kioxia will be a publicly traded company with the standard pressures and responsibilities of a public company working with us together. So in effect, this has all been working out for the better for both of us.

Okay. All right. Let's talk -- let's spend a minute on the road map for flash. It's been a pretty interesting journey from 2D to 3D, and I know that Western Digital and Toshiba, Kioxia have been leaders. As you go to a 112-layer, can you talk about the complexity to continue sort of doing this migration? Does -- do migration time lines start to slow down at all? Are there compelling road maps over the next 2 or 3 years that would keep the layer count continuing at the same pace? Maybe explain some of it.

Yes. It's an interesting interplay of 3 or 4 factors. People always talk about layer counts. I'm 112, somebody else is 128, that is not the -- it's an important factor, but not the most important factor. In the end, what you're looking for is continuous flash scaling. In the end, you want the lowest cost per bit. And so when you scale it, you look at it from 3 or 4 perspectives. How do I get X and Y scaling? And then when you multiply by the Z, you get truly the cubic effect of it. That's number one. So you want to make sure you are balancing those at the same time. If I end up having a die size that's bigger, adding layers does not help. So you want to make sure the base die size stays, Xs and Ys scales nicely and then you add Z. The second factor is something called capital intensity. Capital intensity is when you move from one generation to next generation. To produce the next additional terabit, how much capital do you need to add? You want to minimize that because for all the obvious reasons; you're looking for cash flow; you're looking to make sure that the cost per bit goes down. We want to make sure we are not overloading on CapEx for the companies. That's an important factor. So what you do when you look at it is, what's your installed base? What kind of tools do you already have? How do I leverage most of them for the next-generation? And then you choose an optimum on this. When people talk just about layers, that's a very simplistic way of looking at it. It's not the way we would look at it. Now having said that, we do have one of those times in technology development where you actually see a road map far enough in the future. You were around then, Wes, when we were in the 19-nanometer 2D or the one -- there's always a cliff right in front of us. We don't know when 2D NAND was going to end. We are not in that place. We now are in a good place where we sort of have a predictability in what we need to do. But the more interesting things are going to be, how do I increase 4-bits-per-cell? How do we go past that? What new techniques that you can involve? There are a lot of breakthrough idea that both sides, both Kioxia and us are trying out. When they are ready for public consumption, when we think we are ready for manufacturing, we'll introduce them. But the only message I have for you is the road map is robust. We can expect to have about a 15% cost reduction year after year in flash and that scaling is alive and healthy.

Got it. That's helpful. Can you give us some clues as to some of the tricks or technologies that you're using to -- because I know that dimensional scaling doesn't really happen in the same way with 3D NAND. So as you're adding layer count, you're adding process time. And it feels like it's relatively challenging situation, we've kind of moved away from the traditional exponential scaling. So how do you manage that? What are some of the tricks that you use?

I have to be ultracareful as to what we disclose, what we don't, we still are in a competitive marketplace. I will talk only to those that we have publicly disclosed. So we went at the right time to do 2 layer, 2 stack memory. So we wanted to make sure that if you just etched it one way all the way through the 112 layers, even a small angle, 1 or 1.5 degree change in the verticality ends up blowing up a die size. You can get -- you can do it in one step, which is better on the throughput of the process, et cetera, but it is bad because you end up making the hole size bigger. So we have publicly stated that we have gone into 2 stacks on this. That's -- and that is going to be needed for the long term. And so to get that under our belt and make sure that we are able to align the 2 layers and produce it in high volume, which we have done in BiCS4 generation, that's a major advancement. Similarly, we have talked about how do you minimize the dummy layers between the holes? So remember, in flash, you have to reach from the side to etch out things to produce the interconnects, the wordline planes. The farther out you can go, the higher the density on it. We have done some very interesting work. The third is on staircases. We need to make contacts to all of them. There's a very unique, innovative way we have done in the staircases, whether the staircases, like in a mind pyramid go all straight down or they make great angle turns, those things are very, very -- I mean it's hard to visualize easily. You should go look at a cross section, and they are very good work that we have done. These are the kind of it. Then, of course, 4-bits-per-cell, then what happens after 4-bits-per-cell, right? These are good ways to improve scaling.

That's helpful. That's helpful to understand the road map. And then just kind of following up on the road map idea, I know you said layer count doesn't really matter, but like why are most people going to 128, Western Digital is at 112? Does it imply any sort of advantage or disadvantage versus others?

So I did not say layer count does not matter. Layer count is one of the things, right? Now having said that, getting the same cost reduction or better cost reduction and scaling, meaning getting more bits on the wafer, if I can do it to the lower layer count, that's the best thing. That's what we are pretty proud of. We can get the same scaling, better bits per wafer and lower cost with the 112 layers. So I think we are actually a little bit ahead of the world in that thinking...

That's helpful.

That I don't need to go to 128 layers. I don't need to get the equipment to be spent on it, and I don't have to have to cycle time it. I don't have to throughput it, I don't have the additional consumable cost hit, and I can still get the cost reduction and scaling on 112 layers. That's a unique advantage.

Got it. Is there some sort of limitation in terms of the number of layers that we should think about? And then maybe we see the industry move toward more QLC or maybe 5-bits-per-cell?

They both have happened simultaneously. So you need to have QLC. So I think in the BiCS5 generation, QLC is starting to hit its stride. QLC was not that useful ahead of it. Even if you could make it, it was not the best thing. About the BiCS5, BiCS6 generation, QLC is coming in -- coming on to its own. And we'll do both, layer counts and 3 and 4 layers -- 4-bits-per-cell, and we'll keep pushing to see if we can go farther than that. As far as the limit of the layer count, we see visibility for the next 3 or 4 generations, very well, well over 200 layers. So I think we are okay for the next 4, 5 years.

Got it. Okay. Good. And I have coming from an investor a follow-up question on the Kioxia IPO. It would be good to come back and hit real quickly before we move on, and the question was essentially, you mentioned that it would be good if they did an IPO potentially for Western Digital. Can you elaborate on that? Is there any direct impact related to whatever potential for valuation this company gets...

Clearly, I mean, for us to have a pure-play flash compound there, that's one of the better things. I mean we currently are one of the few -- the only company in the world that has got both flash and hard drives. We are the storage leader because anything storage, we have the products. But there is no pure-play flash for us to compare against. For us to have that is of immense benefit to us. That's number one. Number two, the decision-making processes for us and for them to be on the same level playing field because we are both public companies trying to optimize shareholder value the right way, that's probably the best thing that can happen to both of us. We play by the same rules, disclosures are the same. So we are not at a dissimilarity because they are not a public company. It is not materially different because of the same people making the same decisions, but the rules become uninformed.

Okay. And maybe, Peter, this is maybe for you, but is there an equity implication in terms of Western Digital's equity stake in the joint venture?

No. There shouldn't be any change with regards to the way we'd be reporting at or disclosing the joint venture participation. So no, I would not anticipate there's any dramatic change there.

Okay. All right. Good. I want to spend a couple of more minutes on flash and then spend some time on the HDD road map because I think it's really interesting. The new applications for flash, we're seeing the gaming market now moving for -- at least for the consoles into flash. Is that a performance reason or is it an economic reason? Do you have a view on this new market and maybe how that would translate into other market opportunities?

Yes. Wes, you and I have seen this over time. As flash prices come down, just completely new markets open up. And we have seen this maybe play out for the last 20-plus years when -- into cameras and films to CDs and DVDs and audio players and client SSDs for computers and mobile phones. Every time flash gets a certain threshold on cost, on normal scaling, it comes down, completely new markets open up. One of the obvious ones was gaming. Gaming was there for it to happen. It was just the right time when the cost structure works perfectly for the speed that they need, the reliability, robustness they need and the cost of flash. And when all that converge, you know it's an obvious place where flash full go into. Video is a place where we expect a lot more movement into flash. So these are things that we expect as we see logical extension of technologies into new space -- new adjacencies.

Okay. And then I don't know if you spend as much time on this kind of spearheading the R&D side and the technology road maps. But can you update us on your outlook for just overall supply and demand for flash this year and then sort of moving forward? And then any comments on the cloud digestion that's anticipated for the back half would be helpful.

Yes. So with respect to supply and demand on flash, we are at about that near level area that you expect. We are in the close to equivalents in supply and demand, and we are in that range. Now it's hard to predict given the global pandemic and economic downturns what's going to happen very quickly. But the major consumers of flash continue to be buying. There is still a lot of demand for it. So we are in the right space. And equally importantly, we are watching from their public announcements and in the marketplace what the other competitors are doing. There is rational behavior out there. So we are watching the supply demand to be in reasonable balance in the short near future. Of course, nobody can predict the longer term. But in our view, supply/demand seems to be imbalance. With respect to cloud absorption, it's sort of a -- given the big cloud players, there's always one that is inventory collecting, the other one is starting to buy, the next one is inventory collecting. I think, yes, there is some absorption that always goes off, especially because early part of the pandemic when everybody was trying to get ahead of the curve. But I think over time, the big 4, 5, I5, I4 players will start to be sort of slightly out of cycle that it ends up evening things out. We'll always go in through the absorption that is going to be tough of what happened in the past as to how fast they bought in calendar Q1 and Q2. But I think over the longer term, that will also work itself out.

Okay. That makes sense. And just a little bit on the supply side, it feels like some companies are ramping up CapEx a little bit. Does that land any concern? And related to that, Yangtze Memory starting to ramp up pretty aggressively. I think they'll probably be a meaningful contributor to supply next year I know with kind of a lagging etch tech, but I would love your thoughts on those.

Yes. Very, very important thought process to follow along. Among the established players, we modeled CapEx investments, what they need to, when they go, so when does a fab or a technology wear off, when they need to do a refresh? There is nothing systematic that somebody is increasing or not. It is just timing on when they buy, et cetera. But YMTC is a new entrant, and they still lag in technology, but there are markets for those kinds of wafers. Scale is not quite there. They will have to come up to speed. So we don't expect much of an impact either in this calendar year or next calendar year. In calendar year '20 and calendar year '21, they -- we don't see them being a major impact. We do need to see how well they come along the technology curve afterwards. We are watching that, of course, very, very carefully. And it's something that we all have to adapt ourselves to based on how well they exhibit. The only thing we can do is to keep moving this leadership position and technology and keep there well ahead of it. So our cost structure and wafer output and bit output stay ahead of the curve.

Okay. That's helpful. Let's spend some time on the hard disk drive segment as well. First, I asked you just before the presentation, but how strange is it for you to be in charge of hard disk drive R&D coming from the flash side?

It is -- you said it was surreal to start with initially. But then as an engineer, it is -- nothing is more fun than going back into one of the most advanced mechanical engineering marvels with the other passion of my life with material science thrown in. Every known atom in the universe is being used somewhere or the other in the media or in the head and mechanically to go from single actuate to double hinged to triple hinged, it is amazing how far that technology has come, that a rotating media has been able to deliver 18 terabyte in a little 1-inch form factor is just incredible what has happened. So I am having the time of my life trying to learn and come along with further items in that technology development.

Great. Okay. Well, maybe you can help us understand some of the technology changes because I know there are some big ones coming down the pipeline with energy-assist. How should investors think about that? And then is there -- it's an area where over the next few years, you really see that areal density road map accelerate?

Yes. So I want to lay down the landscape slightly differently than the story that we've always heard. Clearly, Western digital is the areal density leader. The areal density leadership in the end is what drives that road map. However, how do you achieve that areal density? If you are a semiconductor guy, you can think in terms of, do I just shrink the device? Do I get it through the interconnect? Do I get through the packaging? Do I get through the system? The same kind of concepts apply in hard drive as well. We always focus on the heads and media, which is the equivalent of a device where you come back and say, "Okay, did you shrink that fastener? Did you get energy-assist? Did you do HAMR? Did you get MAMR?" They are important things, no question. But for you to maintain that steady cost reduction that you need for the long term and to scale, there are 2 other pillars, along with the heads and media that are important, which is mechanical recording, subsystems and the firmware. All 3 legs have to go hand-in-hand. What you do in firmware with recording algorithms give you a lot of freedom on the reliability specs and the performance specs for the heads and media. What you do in mechanical, whether you go from a single actuator to a double actuator or a triple actuator to go do it, allow you to increase the number of tracks per inch. In the end, all of these play a role in giving the finished drive with the density and the cost structure that you need. The biggest strength that Western Digital has is that we lead in all 3 fronts. We lead in heads and media in our ability to do energy-assisted and get there, which is important; equally, importantly, the mechanical subsystems and the recording subsystems; and then in the firmware, we needed to pull this all together in the drive. This is what has allowed us to introduce 18-terabyte, which we are ramping in volume right now. This is what we think will allow us to get all the way to -- on the current platforms with the kind of heads and media that get up to 30 terabytes. We have a road map that is right in front of us that we can reliably deliver. That's what's exciting about the hard drive is that the scaling in hard drive is not slowing down. For the longest time, we have been talking about the hard drive as a shrinking marketplace, it's actually a growing marketplace. With the capacity enterprise growth and the cloud growth, it's a growing marketplace to which there is an underlying growing technology scaling that goes with, together makes this market very interesting.

Okay. That is interesting. But let's spend a minute on the growth. Is that -- are you talking about growth in terms of bits of storage because of the expanding areal density, which makes sense, but -- or is it growth in terms of expanding revenue opportunity with just all the usage and data centers?

Clearly both. The capacity enterprise bit growth is fueled by the cloud growth. The consistent growth in demand for that bit growth drives the industry. What is fascinating is the ability of the technology to scale, as you know, just like DRAM leveled off in scaling. So any additional demand is just by -- added by more wafer starts or more capital spending, whereas hard drive is still allowing us to do cost reductions and bit growth in the same form factor. In that 1-inch form factor, we've gone all the way from 10 to 12 to 14 to 16 to 18 to 20. And then you add things like SMR, which is like going 4-bits-per-cell? SMR gives you that additional 15% density from the same form factor, many ways that the technology is scaling more bits on the same form factor to satisfy a hugely growing market demand for storage.

Okay. Can you help us understand how you expect that to translate into value for Western Digital? And loosely, it's a gross margin question. But broadly, with all of the technology that you just described, it seems like really only 2 main competitors. It seems like the margin profile should improve over the next few years?

Which is precisely what we think would happen also. Given the way the technology is scaling, the entry barriers are very high. And our importance to the cloud players has become very high. We have become an enabling technology. And given this marketplace, we do expect our technology leadership will lead to value growth. In the end, that will set us apart in our ability to produce that leading edge product that command the higher price at the right place at the right time period in right volume. Our manufacturing scale, our technology leadership, and our customer relationships are well poised for us to get the value out of that marketplace.

Got it. And I guess, in a roundabout way then you would expect technology leadership to create enough separation from your competition to create some of that value?

I want to say it in terms of satisfying the growing customer needs at the appropriate time. That allows us to be there ahead of everybody else.

All right. Well, I just realized we ran up against our time limit here. So with that, I think we should wrap up. Siva, this has been really interesting. I really appreciate your participation. And Peter, thank you so much for setting this up. And thank you.

Thank you, everybody.

All right.

Thank you, everyone.