Western Digital Corporation (WDC) Earnings Call Transcript & Summary

Okay. Good morning, good afternoon, everyone. I'm Sidney Ho. I cover semiconductor, semi cap equipment and IT hardware at Deutsche Bank. The next company we have is Western Digital. Western Digital is the leading supplier of both hard disk drives and flash memory. Today, we are very excited to have Siva Sivaram, WD's President of Technology and Strategy, with us. Welcome, Siva. Good to have you.

Thank you, Sidney. Good to see you.

So before we start, for those investors who are listening to the webcast portal, if you want to ask a question, there is a box on your screen where you can type in your questions. I will try to weave those questions in as we go through our discussion. [Operator Instructions] For now, I'm going to turn the mic over to Peter Andrew from WD for the safe harbor statement.

Okay. Thank you, Sidney, and welcome everyone to today's presentation. Before I turn the camera over to Siva, I will -- I want to alert everyone that we will be making forward-looking statements. And I ask that you please 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 of our GAAP and non-GAAP financials can be found on our website. With that, let me turn it back to you, Sidney.

Okay. Great. Thanks, Peter.

So Siva, I know you are the technology person, so I can go a little bit more technical here. But first of all, maybe just to step back. If I kind of look at your last quarter, fiscal Q4, the June quarter, how do you think from a technology execution and innovation perspective, how did you guys do?

So as you saw, Sidney, first quarter was this very, very good quarter with respect to innovation that leads to products to market and then that results in better margins and free cash flow generation. So if we look at the big picture, we're a technology company. And as a global technology company, all of our advances come from innovation and technology breakthroughs. So if you look at the hard drive side of it, rapid expansion of the energy assisted drives. So the energy assisted drives, we were fully commercialized the 18 terabyte, and we were shipping it across all the major titan -- cloud titans, almost triple the shipment of 18-terabyte drives. Capacity enterprise, HDD revenue shipments, both increased by over 50% quarter-on-quarter, substantive hits in HDD. On the flash side, you saw us talk about our new NVMe drives for the enterprise. We had again qualified into a couple of cloud titans, and a third one is underway. Shipment volume grew. And you saw the margin increase on both sides, both on the HDD side and on the flash side where it was clear quantum step-up in the gross margins. So -- and then leading to good -- almost $800 million in free cash flow. So you can see the basic technology coming into products and the products getting qualified to customers and the resultant financial performance that came with that.

Excellent. Thanks for that recap. Now WD is unique in the way that you have both HDD and then NAND together, unlike your competitors where it's either/or. What was deficient when the team acquired SanDisk in 2016? And how has things transpired since then against that mission? What will you say is the biggest upside and downside surprise during the past few years from this strategic move?

So I want to be a little bit, I mean, humble about it. Nobody knew what the future was going to hold. When the acquisition happened in 2016, there was a hard drive market with certain characteristics and flash markets with certain characteristics, and several markets were then just growing up. But in the 5 years hence, we have created the world's largest storage provider. The first full spectrum storage provider all the way from hard drive to flash. But what has happened is that the diversity of our portfolio actually reinforces the two to each other. The diversity helps, hard drives helping flash, and flash helping hard drive. So for instance, the clearest example was client SSD. SanDisk had a very, very good client SSD product portfolio. But our market share was in the 7%, 8%, 9% range. And we were still having trouble breaking through into the market. We significantly had a substantive presence in client HDD. All the major OEMs had been customers for 20 years. We knew what the procurement process was. We knew what the qualification process was, what the engineering requirements, what the quality requirements, how do we qualify et cetera. Put the two together, voila, today, we have close to 30% market share in client SSD. You could see how that works together. Same thing in the channel. HDD had this very well-developed distributors, value-added resellers. So you could reach far reaches of the world, every country in the world we could get to. Now you've taken and put the SSDs into flash, into that marketplace. Now you can see the channel really starting to bloom for flash. On the flip side, there was a big consumer marketplace for SanDisk. SanDisk was a very, very strong consumer player. Now we can go back and say, external hard drives and external SSD going into the same channel, coming back, we can shape the demand on both sides. So there is a lot of interplay that has worked. No where is it so pronounced as in the enterprise marketplace. Enterprise HDD, capacity enterprise got into its own in that 2016, '17 time period and has had this 35% a year CAGR growing dramatically, same customers buy NVMe capacity SSD with hard drives. We are able to work both from it. Now you can see the success we have on both sides in the enterprise, in the data center space, both in HDD and in SSD. So in the marketplace, we now are being viewed as a true full spectrum supplier of storage.

That makes a lot of sense. I remember 2 months ago, you gave a talk and you talked about the existence of both hard drive and SSD, how that is a competitive advantage. And you also talk about synergies in terms of market, internal operations, technology and customers. What are you most excited about? And in your view, what are the areas that you think that are least understood by investors?

Good thought process that went into this, Sidney. It's obvious when you come back and say, "Hey, there's a hard drive, there is a flash. Oh, the technologies are very dissimilar, why are we keeping them together?" You look at it one step below, then you start to see the power of them both being together. Clearly -- and from the outside, you can see products that are coming up -- that are learning from one to the other. We just introduced OptiNAND. OptiNAND is flash enhanced hard drives that are very different than the all the hybrid drives when the 2 companies are separate. This is something that is on the control plane of the hard drive that improves areal density, performance, reliability. Last 2 quarters ago, you heard us introduce ZNS, zone named space drives, in the flash narrative. That was a direct derivative of SMR drives to come back and say the same kind of technology when you take an SMR and HDD and apply into flash, you develop ZNS drives. We now talk about, in technology, new things like MRAM, where MRAM material state is the same as HDD, but the product is a solid state drive, it's on the same device. So you take the material set from one, apply it to the other. So these are in products and technologies. You can see visibly from outside. What you don't see is inside the company. Flash back-end manufacturing and HDD drive manufacturing are remarkably similar: automation, procurement, factory operations. Like, for example, when COVID hit, how do I transport people in Penang or in Thailand and making sure that the systems are the same so we can protect our employees. You could see common systems developed within the company in our own operations. And then of course, in markets and customers, I was talking earlier. As the -- our customer base is consolidated. The customers are big buyers of both flash and hardware, especially in the cloud space and the OEM space. When I have a combined revenue of close to $1 billion from one cloud titan from hard drive, in fact, we are able to access relationships at such high levels within the company all the way down because $1 million dollars of spending on one customer is substantive, how -- whatever the size of the customer is. So the customer relationships have gotten a lot more intimate, especially these large OEMs and large cloud titans. So you can start to see whether it is in the market -- oh, just as an example. The intelligence we get from the channel in flash is very cyclical. You want to have advanced intelligence as to what's happening. You gather that from the channel that gives us an idea of where the market is trending. When Chia happened, we could react fast because we had such a strong presence in the channel. We could quickly react, "Hey, this is what it means for flash. This is what it means for hard drive. This is what it means for inventory. We can work through them." So markets, customers, internal operations and manufacturing, technology products, strong synergy across the company. We have come together very well over the last 5 years.

Very good. I want to double-click on OptiNAND technology that you announced last week. You mentioned just now that it is very different than hybrid drives. I think a lot of people still think it's some sort of hybrid drive. If you can give us a little bit of color, maybe what segments they are intended to address. And for investors, do they carry higher prices and maybe margins when compared to comparable hard drives without the NAND flash in them.

Yes. So first thing about this hybrid drive. In the early 2001, 2012, 2013, there was all this hype about, hey, take flash, take hard drive, put them together in the data path, you'll have a strong buffer of cash that's available that can improve the IOPS for the hard drive, and I can get you that. It was a disaster. Never went anywhere. It was not a successful product. Trying to have the characteristics of both hard drive and flash, you end up getting the worst of both worlds. What we are doing is very, very, very different. For instance, on a 20-terabyte drive, having a 64 gigabyte of flash is not going to improve that overall density, right? Is 64-gigabyte versus 20-terabyte, we are talking hundreds of magnitude different. So it's not something that is introduced in the host to the device data flow. It does not touch that data. It sits on the control side of the drive. So what we are trying to do is, let's say that drive is increasing, we want to increase the number of tracks for each, so I can get a higher areal density. When you bring the tracks closer, there's the adjacent tracking difference, you'll have the remap, what is going on. That kind of information, the drive produces gigabytes of those, making sure and the spindling runs at a certain sector, the spindle runout is a little up or down. The factory stores that kind of information. Instead of keeping that in the hard drive itself and then buffering it in DRAM, which is only about 512 megabytes, we have now introduced a 64 gigabyte of flash in that drive that allows for substantive improvement in the areal density. I can bring these tracks closer. I can actually improve the performance of the drive. 60%, 80% improvement in right cash disabled performance of these drives. Reliability, even the power goes up, what do I do? So these are big -- measurable big advantages of doing putting the flash in this. Because it can't be done by any -- you can't just take flash and put it into the drive. That's not the way this is. This is in the infrastructure on the drive, making sure that the firmware talks to each other, does this in the background, on the control of the drive. That's why this is not a hybrid drive. This is actually enhancing HDD with control being performed through flash.

Okay. That makes perfect sense. Now you also talk about earlier that there are other products that you guys are developing, whether it's MRAM, and I think in your talk, you also talked about data center solutions like computational storage and storage accelerators, those type of things. How significant are these opportunities over time? And what is the realistic time line for them to be a meaningful contributor?

Yes. Before I get there, I just want to finish one thought for us on the OptiNAND. As we were saying, the OptiNAND itself is a technology that sits on top of all of our regular products that we're going to be making. So we're introducing a 20-terabyte drive with only 9 lists in it. So you can see the advantage of having a 2.2 terabyte per disk on a drive. So that's where the cost advantage comes. The cost advantage comes from the fact that we use less media, less heads and getting to the same density, while at the same time, improving the performance and reliabilities, so the customer value is substantially better. So hard drive, the economics are compelling when we put the OptiNAND on top of it. Now on to the question about MRAM and computational storage and accelerators. So in many of these, these may be products that we work with customers. MRAM, as you can see, is a new storage class memory that we can deliver. It's a very big market. This is in the space between a DRAM and a flash, a large gap exists and you want to go address this. This is a multi $10 billion kind of a marketplace over time. But it is not going to be instantaneous. Even if the technology is developed, the customer has to develop use cases for it. There's going to be software rewrites that need to be done. So it's going to be a slow progression to it. So we need to develop all our controllers, firmware. Customers have to develop a method of accommodating these persistent memories in their products. So I would expect that it's not a revenue producing in the next 3 years. It's in the 3 to 5 years is where the product will start to be in the marketplace. Digital storage on the other hand, it's not. Computational storage is a broad scheme of things where you come back and say, what can I do to minimize moving huge amount of data up and down the network back to a central processor for minimal task? For example, encryption, decryption, compression, database acceleration kind of activities. These can be done now, so we actually have these kind of products along with the ZNS, we are thinking about very, very quickly. You work closely with customers to make that happen. Accelerators is sort of in between. The CXL bus is just starting to come up and getting [indiscernible]. When the second edition of CXL comes, we want to be right there with their acceleration, our video compression products that can sit across specific accelerators for specific products that we are working on right now.

Excellent. So let's bring it back home a little bit closer for the current technology development. So on the HDD side, there seems to be room to continue to increase the capacity in the near term through your energy assist technology. Is there a theoretical limit as to how much energy assist can go? And when and how does HAMR technology start to impact your road map? What are the trade-offs to consider if and when you switch over to HAMR?

Yes. So energy enhanced is a broad umbrella. HAMR is an energy-enhanced device. MAMR is an energy-enhanced device. What we are currently doing as energy enhanced is where we have the DC [ FX ]. So these are progressive steps. What we are seeing in our energy-enhanced drives is, a, as we develop technologies and as we grow out the reliability, we'll continue to add this bucket of energy-enhanced PMR. ePMR is energy-enhanced PMR, but particularly magnetic recording with additional factors that are coming in. And as you would expect, you could talk about a home-run and then say, jump all the way through, we. Don't believe in that. We think that there are many steps that we can continue to take before we go into a full HAMR in this method. So I think there's plenty of room, continue to cost reduce, continue to push the areal density, continue to get the gains from all sides, whether it is mechanical, in the recording material science, whether it is in analytics, whether it is in the firmware, whether it is in the control, everywhere to find gain and cost reduction, so we can continue to move this road map from 20 to 22 to 24 to 30 to 50 over time.

Okay. That makes sense. One of the bear cases for hard drives is that hard drive cost reduction has been slower than flash memory and that the cheaper QLC NAND is already starting to replace some of the hard drives in some workloads. What is your response to that?

Look, this has been a miss for a longtime. Even the client space, it was clearly a replacement case. In the smaller density, 1, 2, 4 terabytes, you could see that flash as the cost came down because of replacement for client HDD. In the enterprise space, it's a very different ball game. We are talking about densities that are very high going from 20, 24, 30, 50 terabyte drives growing at a 35% bit CAGR. Flash is growing similarly at the same kind of 30% to 35% growth. We have 2 very rapidly growing marketplaces where one is already up here and it is cost reducing in the low teens. And this one is growing -- cost reducing in the mid-teens. These 2 are going to be going in parallel for the next decade together.

Got it. Got it. Well, will you switch over to the SSD technology for a second? BiCS5, a 12 layer is already ramping and is expected to reach bit crossover later this year. In the past, you talked about BiCS5 being the most capital-efficient node in the 3D era. How do you think about the cost benefits and gross margin profile as BiCS5 continues to ramp up?

Yes. So this is a conscious choice that we made in our technology development decision, that you can [ drew ] forces by just adding layers. So bit growth by just adding layers is a linear growth. So you add more capital, you get more layers, you get more bits. We thought and we still firmly believe that the best way is to use the number of layers as the lever but the basic multiplier is how much we get shrink on the x and y directions. So when you get a good shrink on the x and y direction and then multiply it by the z direction, you get that huge lever in cost reduction. This is how we managed to do BiCS5 to be the lowest capital-intensive node in the 3D era. We get the cost reduction, we get the bit growth, but don't quite spend that much amount on capital. Going forward, that philosophy will still be true. We will continue to push lateral shrink even as we add layers. We'll probably be about 10% lower than the competition in the number of layers, even though if we get the same cost reduction and bit growth, that would be to our advantage. So what that translates to us is in 2 places. You can see, we are really conservative about our CapEx spending. Second, we can consistently deliver the 15 plus/minus in annual cost reduction as we ramp BiCS5 and then later on into BiCS6.

Excellent. So you just answered the next question I have is on BiCS6. You talk about some of these improvements. You talk about having fewer layers and competitors, but probably same kind of density improvement. What are the -- what is the timing for BiCS6 from here?

BiCS5, as we said, we'll probably exit the year shipping more BiCS5 than we do BiCS4. So we have, as we said, in December quarter, we would have done the bit crossover. And then the big job is to proliferate BiCS6 into as many products as possible. So you take a horizontal technology and then put it into all the verticals, whether it is client, enterprise, mobile into consumer, into hard drives everywhere they're going to start putting those. Lot of effort will go there. In parallel, we already announced a BiCS6 technology as we just mentioned. We will be starting to transfer that to production about now where we're starting to go. But we won't reach meaningful volume. That bit crossover is usually in the 15 to 18 months between nodes. So the same thing the crossover will happen 15 to 18 months hence. Because that ramp is one thing to produce in the fab. It's another thing to get it qualified as products and put it into customers' hand. So we balance those 2 as we go out. And the technology [ repeat ], we will start to transfer the manufacturing -- volume production will be -- the crossover, bit crossover will be 15 to 18 months hence.

So as far as you can tell, over the next 15 to 18 months kind of ramping BiCS6, you will be as cost efficient as some of the competitors, if not more cost efficient?

That's -- clearly, that's my tracker, to make sure I am more cost efficient than my competitors.

Excellent. So we have an inbound question, and I'm going to read this out here. We are seeing some of your cloud customers designing their own ASIC controllers and just buying the flash. What is it that these cloud guys require in their controllers that they cannot get from WD, let's say, the NVMe drive? And do you see the use of custom developed controller as a long-term trend? Or is it more transitory until the SSD vendors like WD integrate all the required functionality into the controllers?

Yes. We have seen this trend ebb and flow over time. In the 2016 time frame, a lot of our customers said they wanted to build their own controller. And there's still one client titan that builds its own controller. What has happened is the requirements on these NVMe drives for each of our customers is substantively different depending on their workflows. It's not one -- like it's not a client SSD, you take one product and you qualify into multiple customers. Each one of them is very different in their needs and performance and features. So what happens is we tend to go progressively, come back and say one to the other to the other. Our -- some of the cloud titans, obviously, try to go do this themselves. And in some cases, they are successful. In some cases, they are not. They end up buying more from us. And this will emerge over time as each one gets, as you said, comfortable with us, they'll trust us more then spend the same amount of money. Because in these controllers, both the front-end interface and the back-end interface are very, very complicated. The standards that they need to, whether it's PCIe Gen 4, Gen 5, Gen 6, as they come along on the back end, trying to be compatible with BiCS3, BiCS4, BiCS5, BiCS6, it's changing the current -- the requirement on both the controller and the firmware are changing a lot. So you'll see that over time, they'll probably come back to do more and more and more with vendors from us than do themselves.

That makes sense. I think I can go on these technology questions forever, but I do want to address some of the other parts, the business itself, if I may. So currently, in the hard drive market, things have changed quite a bit. In the last, call it, 6 months, you got these cryptocurrencies that kicked, kind of came and went away. You got a really strong nearline business, not just yourself but the market in general. But obviously, one of the concerns among investors is that the nearline hard drive market may be a little overheated. Curious how you think about it. And maybe how you think about cyclicality of the nearline business in general going forward.

Yes. So in the nearline market, for the last 3, 4 years, we continuously took capacity away from client and put this into capacity enterprise. And that helped this capital investment requirements to be at a minimum. But we are coming to the end of that phase of growth. There is not enough planned HDD space to convert over to capacity enterprise. The growth continues to be strong. The growth continues to be in the 30%, 35% bit growth rate in capacity enterprise. So now we have to start to invest capital just to meet that need. We are being conservative in that. We want to make sure that the business shows the right level of gross margin, the price shows the right level of stability. The demand, there is clear enough line of sight to the next several quarters. And there's this commitment from our customers before we go invest additional capital. So we are being relatively conservative in our deployment of capital in HDD and expanding capacity enterprise. So that would naturally make sure that the demand supply stays in reasonable balance going forward because we are not getting ahead of ourselves in inputting capacity. We want these margins to come back up to justify the level of investment it needs.

That makes sense. If I think about staying with the hard drive market, you talked about signing some long-term agreements with some of the important customers in the nearline market. How has that changed your visibility and your operations going forward?

It's an important factor that we've been working on for a while. As you know, the cloud titans go through ingestion and then digestion. And 4, 5 major ones are not all perfectly synchronized, are asynchronized. So they go back and forth, and so they cost local modulation in demand supply. But over several quarter period, they do very well. We are now working very closely with them just so that we know exactly what their projections are for the next several quarters. That helps us in planning not just capacity, but also in revenue and the inventory so we can manage this and smooth this out. Because each one of them, again, the firmware revisions may be different, the testing requirements may be different. We work through with them carefully, and they are starting to show us a lot better visibility. They are working with us because they know the demands are going very fast. And there's not that many 18-terabyte or 20-terabyte vendors in the world. Their demand is growing so they want to make sure they are working with us with some more visibility.

Okay. That makes sense. Now next question is on CapEx. And then you touched upon that a little bit in the earlier question. But on your last earnings call, you indicated that you expect to spend roughly $3 billion in fiscal '22, which is the same as what you do fiscal '21. Can you talk about where you are making investments between the 2 businesses and how you're thinking about investing in capacity expansion in your NAND business versus investments in technology transition?

Yes. So in general, the JV is a big part of our capital spending for flash and then our own heads, media and drive factories for HDD. In the JV, we have said openly that we will maintain flash market share. Now in the long term, that is truly. It's probably going to be very hard for anybody to predict exactly how much the market is growing and how much we should be growing back and forth. And we are also trying to do, as you said, technology transitions, et cetera. Our primary focus in flash growth has always been through technology transitions. We did bring up the new fab in Kitakami, primarily to observe the overflow from Yokkaichi. So we are investing in capital in the JV primarily for converting prior technology to the next node, but we are also allowing for space as we grow from one side to another. In the hard drive, as we just said, we are running out of this conversion capacity. We are starting to invest in drive capacity, additional drive capacity to be built. So in both cases, we are being prudent. In flash, the simple rule is we will maintain our market share. In HDD, we will be conservative in our capacity growth. We watch the business to get to the margins that it needs to sustain that investment.

So on both sides, you're still thinking about 35% bit growth in your CapEx plans?

Yes. [indiscernible]

Okay. So this will be my last question, but I have to ask about supply chain. So everyone is seeing some supply constraints. When you think about the different issues impacting supply chains include factory shutdowns, maybe higher component cost, higher logistics cost, how much of this do you think will resolve itself when the spread of the pandemic is back under control? And how much is more related to demand environment that many in the supply chain just underestimated?

So there are 3 independent maybe related things going on. One is the pandemic. So we are ourselves making sure that in our various factories in Southeast Asia, whether it is in Malaysia, Philippines, we are in Thailand, we are making sure our employees are safe. We are making sure we can run our operations. We can locally procure what we need. We are spending a lot of careful time on that. The second is the semiconductor cycle and the semiconductor shortage, the controller shortage that we have to manage through. Meaning, we have certain amount of sharpness of controllers. So we are trying to allocate our bits to go to the highest density products, highest margin products, making sure among our 4 verticals, we are able to move the bits around so that we maximize the use of those controllers that are scarce. And the third, as I said, is the intrinsic demand environment. In the flash effort, we -- underlying, we see a general balance between supply and demand. There may be some spot shortages on flash, but otherwise, supply-demand is reasonably well-balanced in a macro sense. In the hard drive case, we can see that we are being careful in how much supply we put into the marketplace. We are not focusing as much on market share, but more on margin and profitability of the business. So when you put these 3 together, the pandemic, the semiconductor shortage and the basic strength of the market, I think we are in a good place right now.

Okay. I think we just ran out of time. Thank you for spending your time with us, Siva and Peter as well. And again, thanks. Have a rest of a good day.

Thank you, Sidney. Good to see you. Thank you.

Take care. Bye-bye.