Western Digital Corporation (WDC) Earnings Call Transcript & Summary

Great. Good afternoon or good evening, everyone. I'm Joe Moore from Morgan Stanley. Happy to have with us today Peter Andrew and Siva Sivaram from Western Digital. So I think, Peter, you're going to read a safe harbor, and then we'll go straight into Q&A. And then if anyone has any questions, I think you guys probably know this by now, but there's an e-mail system, and I'll get those questions via email, and I'll relay those as they come through. Peter?

Okay. Thanks, Joe, and thanks for the Nasdaq team for hosting us. Before we kick it off, I do have to make a brief announcement. We will be making forward-looking statements, and I ask that you refer to our SEC filings for the risks associated with these statements. We will also be making references to non-GAAP financials and a reconciliation between the GAAP and non-GAAP, results can be found on our website. So with that, back to you, Joe.

Great. And just to kill you guys with all of the disclaimers, I also have to read for important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. And if you have any questions, please reach out to your Morgan Stanley sales rep.

So with that, you guys have talked in a couple of different forums about some of the bigger decisions that you've made. And in particular, you've highlighted the strategic benefits of having both hard disk drives and NAND under one roof. Can you give us some of the summary conclusions from that and talk about the benefits that you see there?

Yes. Thanks, Joe. This is a big decision we have made with 2 large markets. In NAND, we have both certain multiples and servicing the storage marketplace. What we have done is clearly organizationally made these 2 separate so that there is focus on both of these markets. So a senior set of reputed executives are focused on hard drive and flash separately. But given the demand that is very similar, especially with respect to the customer synergy, and go-to-market synergy, there is a lot of overlap. The best place I can illustrate this is in the channel. Hard drive has an extremely well-developed channel for getting them to the value-added resellers and to the mom-and-pop stores. And SanDisk originally from the Flash business had more than 400,000 points of sales that we were supporting. The fact that these 2 can cross sell each other and get into 1 integrated go-to-market is a huge advantage. Similarly, when it comes to the large hyperscalers and they're mining storage, it is a common procurement and engineering and development organization that integrates the storage. We work very, very closely with these hyperscalers on both the hard drive and on the Flash businesses, a large synergy. Back in manufacturing within our own corporation is very, very similar when it comes to assembly or test, both in hard drive and in flash. There's a lot of commonalities that we leverage. And then, of course, technology, we will be talking in the future a lot more about the kind of new things that have gone from flash drives to eMMC flash, we recently introduced this OptiNAND technology, which enhances our area density advantage even more in hard drive by the use of flash in the control plane of the hard drive. The ability to store metadata about the drive in the flash to quickly service that part of it. I'm not talking about the data path but the control plane of the hard drive, a remarkable increase in area density because of that. So you can see multiple points at which these 2 come together. Customer go-to-market channels, technology, manufacturing, a lot of places where there is synergy, but we have organizationally made sure that the 2 businesses get full, complete focus for next years.

I think that's a really important point in terms of the resegmentation. And I think, I guess, 2 years now since the CEO transition that you had, I've always come at the story from a long history of covering SanDisk, not WD. And I had felt like before that transition that there was maybe more focus on hard disk drive storage. It seems like you're now focusing on these as 2 kind of separate but equal businesses. How do you make sure that you're getting the right resources from like an R&D perspective to both of the businesses? And how do you -- and how do you make sure that you're running each of those as well -- because your competitors are not -- they don't have the benefit of the combination you have, but they are maybe more focused on either memory or hard disk drives. How do you balance this?

This is a very important point that you're making. As you said, it's been 17, 18 months since we had our new CEO come over. And there has been a substantive change in the senior management as much by the way in the Board as well. And the 2 businesses are now led by seasoned business veterans in Rob Soderbery and Ashley Gorakhpurwalla, both are them are very, very well seasoned senior executives. I mean focused entirely on these 2 businesses separately. They have engineering organizations directly under them. They have access to customers. They have access to the road map technology that are specific to them. They run their businesses very well in a execution-centric manner. That may have been a little bit lacking earlier because of the commonality of running both of them at the same time. Now lot more focused on the road map, product portfolio, execution and directly working with the customers. So there is a -- they have the resources and the structure to allow them to work. It has made a big difference, you can see in how fast we introduced the 18 terabyte. How quickly we ramp the 18 terabyte. We are now -- a vast majority of our shipments in capacity enterprise is now in 18 terabyte. Similarly on the flash side, how quickly we introduced the NVMe drives into the enterprise marketplace. We have qualified multiple customers and ramping fast. You can see the results of this focus very, very quickly.

Great. So maybe we could look at end markets separately, starting with drives. And before we get into the road map, you just talked about 18 terabyte and the progress there. Can you talk about how broad are the customer qualifications at this point? It sounds like it's the mainstream product, but are there still customer qualifications ahead of you? And separately, where are you with the cost structure of that platform?

Yes. So as you know, 18 terabyte is the first energy enhanced drive in the marketplace. It is a major change in our long-term road map. And that qualification is now essentially complete. We have qualified across all the cloud patents, many, many, many OEMs in the channel, 18 now is our bread-and-butter product. It has reached maturity with respect to yields, with respect the manufacturing efficiency, we are shipping a lot of volume to it. SMR on top of it actually gives you an SMR 20 terabyte on that platform. After that, we have now introduced the OptiNAND. OptiNAND for a platform application, meaning OptiNAND is not a point-in-time technology, it can be applied to all the future products. We are starting with sample the 20 terabyte. We already gotten that out. So in general, the competitive advantage that we have had because of our areal density advantage, we are translating that technology advantage into a product and delivering to customer advantage consistently in the hardware space.

Yes. Just to follow that up a little bit more. You talked about costs. And if you take a look at the HDD business, if you take a look at the gross margin that we've been able to generate, if you look last quarter, we had a gross margin of about 30.9% and that included about 200 basis points headwind due to COVID. A year ago, we were around 25.6%. So there's been a dramatic improvement in the gross margin of our HDD business, as we started to execute on some of the things Siva mentioned.

Yes. That's been good progress on the margin side. Maybe -- and I know Dr. Sivaram, you're not the finance guy, but I just would be curious on your perspective on hard drive market share, I kind of had a view that as 18 terabyte became mature that you would regain some of the share you've lost recently, and you did. The last couple of quarters, you had picked up share. But if I look at the variance in guidance between the 2 companies in December, it looks like you're giving some of that share back. So again, I'm not looking for a forecast. But like generally, is my thesis right that you guys should be in a little bit of a market are recovery? And how do you feel about that on a more forward-looking basis?

So as a business, that's the -- one philosophical adjustment is that we look for profitable business, we focus on the gross margin revenue, top line and bottom line, a lot more than coming back and say, "Hey, I want to keep that market share." And that's just a philosophical shift in that business to make sure we are being careful about running this business, not just run on market share. Market share is very important, don't get me wrong. But much more important are top and bottom lines of that business. And that's number one. Number two, any short-term changes between us and our brethren are always because of some single customer going in and out of their own ordering patterns. In this particular quarter, we had a very large customer with whom we have a very, very large market share. I happened to have some supply chain issues with respect to availability of other components that they could not build out their system. And that it is not a traditional demand. It is demand that just gets postponed a little later. What you do see, however, is our LTAs are getting much more pronounced, with respect to a percentage of the business that is running with an LTA. That is a good indication of our own strength of relationship with these very large customers in the hyperscale business. And you see up and down here and there. But however, I think we would continue to be the largest hyperscale supplier in the industry.

Great. And then you mentioned OptiNAND with respect to your hard drive business, there have been a number of kind of hybrid drives introduced over the last several years where you're sort of getting some of the latency benefits of NAND and cost benefits of hard drive. Why is the time right now for this to be more of a mainstream product than what we've seen historically?

Yes. Both of us have seen that many, many times. This is not a hybrid drive. OptiNAND is not a hybrid drive. We are not playing with latency differences. You know what happens with it. And there's -- for a hyperscaler who has a large massive hard drive and the large mass of SSDs, they would like to determine how they apportion the data flow than us in a single drive. That's not what we are trying to do here at all. In this case, what we are trying to do is each drive generates a lot of meta data, for example, how fast we may need to refresh, just like you think in the flash case, a garbage collection or remapping. How do I make sure that my adjacent tracking differences are accounted for [indiscernible]? There's a lot of that information that needs to be stored and updated. Instead of putting them all back on the hard drive and reading back, which is very slow, we tend to do -- store a lot more of this in flash on the drive itself. That allows for the drive to operate better, higher reliability, higher areal density, that's where the advantage comes from. Not in a latency. We are not trying to buffer data with that, but the size of this does not matter. This is a small NAND that is sitting on top of the control plane on the of the hard drive that makes sure that drives areal density is better because of these metadata uses for the drive itself. That's why this is very different, and this is a platform that is going to be there for the long term for us.

Great. That's helpful. And so maybe to the extent that you look at the interaction between the 2 businesses, one of the bare cases for hard drives is that cost reduction has been slower than NAND. And now we have cheaper QLC NAND entering the market. What do you think is the role of hard drives relative to NAND over the longer term?

Yes. So all of us have conditioned to think based on the client market. In the client case in a PC, given amount of budget in a small amount of data that needs to be stored, it was clear over time that flash would slowly replace, a replacement business, which is not the case in the data center where the demand for data is growing [indiscernible]. And the temperature of the data is also being segmented very clearly from very cold to cold to cool to warm to hot to super hot data that is sitting there. The amount of data that is growing so fast is very cost sensitive. We need to make sure that there is a sensitivity to how that segment is serviced. We see both of them hard drives and NAND coming down in cost, NAND is coming down at about, for us, at least, about 15% a year. Hard drive is also coming down in cost, not quite 15% but not very far away. What we see is that for the amounted capacity we say, we jokingly say by 2060 or so they'll cross over. That's the kind of -- so the 2 are going to coexist in the marketplace for a very, very long time, servicing different segments of that huge inflow of data that's coming into the data center. We don't see the need for hard drives going away in the next decade.

Great. So maybe shifting to NAND. Again, the CEO transition here, I think, is important because I feel like one of the differences I've seen is a real focus on the JV, and maybe a more positive relationship between you and your JV partner. And that just seems to me to be really important because separately, you guys have less market share than your biggest competitor together, it's very similar to the biggest competitor. So if that JV is operating smoothly, it seems like you have all the scale advantages that anybody could have. Can you talk a little bit about the importance of that relationship?

Yes, you summarized it very, very nicely. There are 2 major advantages that come to be. First is in R&D. When we pull together our R&D resources to go make technologies, clearly, we do so much better, because 2 sets of eyes coming in from different segments of the market. We put our eyes together, pull our resources together and we develop the technology. And then, of course, manufacturing scale. As you said, together, we are probably the largest flash manufacturer in the world, and that gives each of us the scale that is needed to go make that happen. This relationship, as you know, is over 20 years long. Very, very strong. Now the time that we went through some tough times in 2017, as you rightfully pointed out, with our current management structure, we are extremely, extremely well coordinated between the 2 sites. At the top level from the management, all the way to an engineering level, this is doing very well. The JV is a crucial part of our strategy that gives us the -- as you said, the scale advantage, and we are paying the right level of attention to the JV so that the 2 companies can benefit from these advantages that we just talked about.

Great. And maybe looking across both you and the JV, I feel like for the last 20 years, there's been this kind of bear case that your capital spending is lower than the competition. And I mean, I feel like over 20 years, if that's always the bear case and you're holding market share, it should, at some point, become the bull case. Can you talk about how you're so capital efficient in this market? And what is it that allows you to spend less than the competition and kind of maintain share, which should implicitly mean a better ROIC over time?

Yes. The proof of the pudding, as you see, Joe, is our cost reduction. And we published that cost per bit and you can run a number very easily. And you see that consistently, we are the best in the business with respect to cost profit reduction year over year over year. And it comes from that fact you just said. We are very attentive to the capital intensity of those bids. The brute force method is add layers. This go say, "OKay? [Indiscernible]." You can do whatever you want. That's the -- as I said, not the thinking man's view of how you do it. You want to do it, that's one lever. You want to make sure you use that lever and what do you multiply that lever with, we focus on that as well. And we have consistently come back and said, we can accomplish the same mid-growth, same cost reduction, if only you could shrink a little bit laterally. We do that very well. We have made that a point of pride for us. And we accomplished what somebody else does at 128 layers on [ firmwares ], somebody does in 176 is 162 layers. We have consistently done that, and we do a tremendous amount of [indiscernible]. This is another point that gets lost is -- The scale of our operations in both Kioxia and now starting in Kitakami, is big enough that the amount of reuse of capital equipment is very, very high. We are not starting greenfields every time. So we end up having technology waterfalls and focus on technology for capital intensity and we continue to be careful in making sure that we are not over investing and getting ahead of the market. All 3 are helping us to be the most the capital efficient among our competitors.

Great. And then maybe moving to the specific, BiCS5, which is your 112-layer technology, with bit crossover kind of soon. You've talked about this being the most capital-efficient node in the 3D era. Can you talk a little bit about that and what -- and your progress in terms of ramping and qualifying that product?

Yes. So let me start there first. I think aided in days and weeks right now out of the fab, we crossover to more than 50% of our base coming on BiCS5. We've already qualified in OEMs retail channel, we are starting to ship. We are shipping across multiple product lines, client, mobile, gaming, retail, we are starting to ship BiCS5. The yields have been very, very good. This is one of our fastest ramps we have done in a while. And this is a classic case of a tick-tock node where the 96 layer was a tick node, this is the talk node, you end up optimizing for, as we said, capital efficiency and ability to ramp. This will be a broad-based node. We'll have products across most segments in this. And this node is -- will be the volume runner for calendar '22 to probably early March '23.

Great. And then can you talk generally about capital intensity around vertical scaling. I mean I think -- and this is less of a WD question and more for the industry, but I think there was a thought that we would convert the industry from planar to 3D and capital intensity would rise, and then we would move to vertical scaling and it would decline again. And yet capital spending stayed pretty high. It's an element of the bear case for NAND, I think, is that capital spending is too high. But it doesn't seem like we've seen any kind of surge in supply. So it seems like vertical scaling continues to be a relatively expensive way to add bits relative to what we've seen historically around planar? Can you just comment generally on what you see as capital intensity trends in the CapEx?

Yes. So people know the [ DRAM ] [indiscernible], right? Yes. You know what's happening in DRAM. DRAM, there is no technology based capital scale, right? If you want more bits, add more capital. Marginal amount of cost reduction, but you just brute-force it. You go back and say, every time I need capacity addition I just need -- add more greenfield or additional capacity. In the olden days, Joe, when we were looking at 2D NAND, we used to get 30%, 40% cost reduction from node to node. We went from the 28-nanometer node to 19-nanometer node. 3D NAND is somewhere in between. It's not as bad as DRAM. It is not as good as to DRAM. However, the road map is still stable for the next 4, 5 years. Very clear road map in front of us as we go from the 128 layers to the 162 layers over 200 layers and getting more up to 300 layers in the next 4 or 5 years, we see a clear roadmap. In every case, we have to offset the cost permit of a brute-force by additional things, 3-bits-per-cell to 4-bits-per-cell, ability to do shrink laterally as opposed to just vertically. Segmentation in the marketplace, making sure that every bit under [indiscernible] used, those kinds of things, we will have to get more manufacturing efficiency to make sure that the capital intensity does not get out of hand. I think that path is still reasonably okay for the next 5, 6, 7 years.

Great. So I have a couple more questions, and then there's 1 from the audience. Maybe I'll go to the audience question first. I don't know if this is for Peter or Siva. But the question just generally -- the state of customer inventories and to the extent that -- I'm not sure if this is intended to be a drive question or a NAND question. But I'd be curious on both. You have these supply constraints in the industry, people are building inventory, some things not others. Can you just generally talk to that?

Yes. So we have been watching inventory, especially our finished good inventory is probably at the low for us right now. We are doing very well. We are watching the entire supply chain all the way through receive driving pockets of accumulation. We don't see it. We are always worried that maybe some of our customers are over ordering products so that when they get some part that they are short of them they can go [Indiscernible]. Both in NAND and in HDD, we do not see it. Clearly, there are supply chain disruptions everywhere. We pointed out last quarter that we left some revenue on the JV because we didn't have enough controllers. And that still versus that there is strong demand, wet supply chain disruptions, do cost -- causes some amount of revenue shortfall. The only place where we are looking at a little bit more carefully with respect to inventories in hard drives in China. There are some big customers new to government regulations. We do not ship. We cannot say the same thing about all of our competitors. There may be some piling up of inventory in certain places in China.

Great. And then I just have a couple of more questions on NAND. Maybe first, talk about BiCS6. And what we should expect around that transition and kind of the timing of that transition?

Yes. BiCS6 is our next 162 layer technology, which is the circuits under the array technology, meaning this is a case where we put the parking lot underneath the skyscraper technology that we are ramping just now. The early samples are coming out where we are starting to give internal samples out and we'll be starting to ramp in the early part of next year. So it becomes meaningful production volumes in the very late part of 2022 and early 2023. So the volume, the money node is still BiCS5 and BiCS6 is starting to come out of the gates. They're starting to sample, develop the products, make sure the products are qualified, so that towards the second half of 2022, we can start to ramp in the marketplace.

Great. And then I guess just going back to the BiCS5 transition, it seems like whenever you guys ramp a new technology, there's always a qualification window where it takes some of your customers longer to qualify than others and you end up with a little bit of a negative mix effect around that. Can you just talk to that? And I mean, you saw some of that in the September quarter. Where are you in terms of getting qualified with BiCS5, with enterprise and SSD-type customers?

Yes. This is a universal problem, right? When a new node comes in, different technologies, different customers take longer to qualify. So for example, for us, retail is instantly available. As soon as the technology is ready, we ship in retail, whether it is USBs or SDs or micro SDs or compact flash, and those kind of things. And there is -- the next highest thing is external SSDs. Then we can turn -- then goes in the channel, the channel green and then channel products. Then they go to OEM, then they go to the hyperscalers in that sequence. And so the reason you saw the choppiness last quarter is not because of anything else other than the fact that pricing in those nonqualified channels just happen to be softer than qualified channels. There are times when this is exact opposite. If you looked at it exactly a year ago, our retail was even much more strong compared to OEM. That's a function of seasonality and where COVID hit harder or less hard, et cetera. Normally, we have that unique advantage of being able to have all of these channels available to us. We ship instantaneously into retail. So that gives us that nobody can see it, we can balance the qualified and unqualified this easily. And that's always going to be the case. BiCS5 is now getting qualified across -- a lot of SSDs. We are in the middle of OEM calls on [ trial ] immediately, afterwards, enterprise will come, et cetera. So we don't see a particular problem with nonqualified with some BiCS5 right now. we have sufficient demand on our qualified customers.

Great. Well, thank you very much for your time. I think that takes us to the end of our allotted time. But thanks, Peter and Siva, for all the information today. Appreciate it.

No, thank you. I appreciate the opportunity. Thank you.