Everyone, thank you for joining us here at BofA's Global Tech Conference on day one. I'm Wamsi Mohan, I cover IT hardware and tech supply chain here. Delighted that you could all join us here. Also delighted to welcome Pure, Charlie Giancarlo, CEO with a long history in tech and obviously experienced sweater in both at Pure and Cisco before that, and can talk deep into tech and deep about the market. We are looking forward to that conversation. We also have the CFO, Kevan Krysler, back there and from IR, B. I. Sanders. If you have any follow-up questions, you can definitely reach out to them as well. Before we get started, I just need to read the safe harbor statements. Statements made in these discussions, which are not statements of historical fact, are forward-looking statements based on current expectations.
Actual results could differ materially from those projected due to a number of factors, including those referenced in Pure Storage most recent SEC filings on Form 10-Q, 10-K, and 8-K. With that behind us, I guess we can kick it off. Charlie, maybe to kick it off, right? What would you say in this sort of really kind of uncertain macro backdrop? What are you seeing out there? What are your customer conversations being like?
Yeah, you know, the conversation that I'll start with is the same conversation generally that we have when I meet with executives at other companies, whether they're suppliers or potential customers of ours, which is that whether it's the global macro or the geopolitical environment changes so rapidly week by week and without, at this point in time, a clear endpoint that there's just a lot of, there's just increased uncertainty for the second half of the year. We are able to get any supplier that is able to get a signal for the next quarter, but your signals for longer in the year generally depend a lot on there being a stable macro and geopolitical environment. Of course, we're living in anything but that right now. That's when we say uncertainty is up.
What we really mean by that is it's very difficult to project when the larger macro is so uncertain. In other words, our signals that we get directly may be roughly the same as they might be in any other year, but they can change so rapidly given a change in the macro. That's why there's increased uncertainty.
Okay, great. I think one topic that I think is squarely in center for investor minds is understanding the linkage of storage with AI. Maybe you could talk a little bit about what your view on that is. Is the overall growth rates within the storage industry going to shift because of AI and how is Pure positioned to handle some of that?
Yeah, it sounds like a simple question. It's actually more complex because AI is going to affect really every part of the IT ecosystem. You have AI, of course, affecting very large scale environments for which their specialized storage is required, storage that can both read and write at very high rates. We recently had a product announcement of a product we call FlashBlade//EXA, which is a modification of our FlashBlade product that allows it to operate at tens of thousands, at the scale of tens of thousands of GPUs. A lot of the attention of the media is on these very large scale environments.
Yet we believe that the bigger change and the bigger opportunity for us, or for anyone actually, will be in the enterprise environment, which might only be tens, may not be any GPUs at all, or it may be tens to hundreds of GPUs. Actually, the real opportunity there from our perspective is that it's going to change the nature and the value of storage in the enterprise environment. To give you perhaps another change that it's creating is up until now, of course, software has really dominated the way that companies compete when they compete with IT, right? A new software program moving to mobile, giving customers the ability to transact or to interact on their mobile devices. It's all based on software. One of the things that AI has changed is the relationship between software and data.
Increasingly, data is becoming more important than software. Let me give you an example. We are in San Francisco. It has only happened in the last year that the number of robotaxis, the amount of trips taken by robotaxis, has surpassed the number of trips taken by ride-hailing or ride-sharing environments like Uber and Lyft. What does that mean? Robotaxis, the way they operate has been because of all of the AI associated with self-driving, right? The data was more important in that sense than the software, whereas the software is what created the ride-sharing apps. You have one example, maybe one of the first, of data becoming more important than software.
If this is true in the enterprise environment, that the importance of data is going to rise, the importance of being able to get access to real-time data for AI analysis is important. That real-time data is going to sit on traditional environments, not on necessarily an AI environment. What we're doing with what we called our Fusion V2 product, or capability that we recently launched, it's just as part of our normal software updates, allows all of an enterprise storage environment that it sits on Pure to operate as a cloud of storage rather than individual storage arrays. That means it's all accessible by AI. Real-time data, production data, accessible by AI. We think that's a very important part of the AI picture.
As I said, I think it's a question that sounds easy, but actually is complex, has many different parts. We're very confident about both our current position as well as our strategy in this environment, but it spans everything from just organizing data better all the way to being able to provide the kind of systems necessary to operate at the highest level of scale.
Okay, that's super helpful. As you think about Pure's growth in the future, how much of that would you say is maybe what you could attribute towards AI versus what you could attribute towards market share gains when you think about, and where within the market would you most anticipate taking share?
Right.
If we just look at the numbers, it's pretty clear. We operate in roughly a $50 billion market in enterprise storage. We're currently just over $3 billion. That's $47 billion roughly, or $57 billion of raw opportunity. We estimate last year that storage that was directly tied to AI was probably on the order of $2 billion. I think that'll grow, but I don't think it's, AI doesn't require a lot of specialty storage. I think that'll grow double, possibly triple, but it's still going to be a very single digit percentage of the overall enterprise storage market. Of course, we're going to be competing as well for storage in the hyperscale market, which is another opportunity for which a lot of growth is possible. I think the AI opportunity changes the nature of storage.
I think we're well set up for that with our Fusion launch. In terms of, and look, I love being in an F1 car race, which is the way I look at the, about storage in the AI market, but you don't find a lot of F1 cars on the road. I think it's an exciting area to be, but I still think it's going to be a small portion of the other two markets.
Okay, that's great. Maybe if we just think about, you mentioned hyperscale as an incremental TAM. How large do you think that TAM is, that market is, and what's your opportunity there?
The overall opportunity right now, the top five hyperscalers are roughly 65%-70% of the total hard disk market. I just want that to sink in. 60%-70% of all hard disks are sold into the top five hyperscalers. It is on the order of 600 EB-700 EB a year. It is a very big opportunity. We have now, as we have talked about, gotten a design win. We are busy working on finishing the next generation data center design with Meta that they will start stamping out across their system. They are one of those large, they are a strong portion of that 600 EB or 700 EB that are sold a year. Obviously, it will take ramp time for us to be able to get into those sorts of numbers.
We plan on, or it's our understanding that we'll ship one to two exabytes in the second half of the year, which is just going into a pilot system with buildouts of their data centers starting the year after that. We are very bullish on that. We are in some POCs in other of the top 10 hyperscalers that hopefully will go positively. Those will roll out obviously, or if we're successful there, that'll be in later time frames. It is a very big opportunity for us. Now, and to be very clear, with those top five hyperscalers, we have, or at least with Meta, we've structured it where they will buy their own hardware. We'll be providing them and supporting that with our software to make that direct flash capability work. Nonetheless, it is a very big opportunity.
Yeah, maybe if we can dive a little more into that and just think through. You mentioned sort of the one to two exabytes in the near term. We would look at, like, I think you noted a double digit exabyte number slightly further out. The scale of what you're talking about relative to what you just said of like 600 EB-700 EB at hyperscalers just being HDDs. What of that 600 EB-700 EB would you say is the area that you're first replacing and what would you call like maybe the low hanging fruit in terms of the exabyte, the incremental EB that you can grow into?
Right. In each of the, again, I'll go back to the top five hyperscalers. They have horizontal storage tiers, meaning that they don't build their storage dedicated to any specific application or any specific customer. They generally have three or four tiers of storage that are based on price performance, right? A low price tier generally has lower performance. A very high price tier that has high performance, a couple in between. We are effectively competing for all of those tiers. More likely than not, the hyperscalers will start off with the highest price performance tier, largely because there's a higher price umbrella and also more performance is necessary, which Flash is able to provide. We've also proven that we can be on a TCO basis performant at even the lowest tier.
I do expect them to start from the highest tier to go to the lowest tier.
Can you dissect that TCO math a little bit, Charlie? Because it's interesting, right? Like, I mean, I cover like, obviously, you guys also cover hard drive companies and the TCO math across both of those seem very different. The TCO math generally, like from what you see at like from an HDD perspective, is obviously focused on the media cost delta to begin with. Beyond that, the extraneous parts of the system other than the media, the size and the trajectory of where those costs are and operating cost assumptions over time seem to be very different. I mean, I'm not asking you to opine on what their TCO, but can you walk us through the assumptions maybe in your TCO?
We had to prove out our TCO to Meta in order to get this design win. Yeah, everybody, or often when you hear the hard disk manufacturers talk about it or analysts, they'll focus on the media costs. The thing about this is that the media by itself does not do anything. They have to be connected into systems. The systems require computers. They require networks to tie them together. They require power supplies to make them work. Now, there are two things that become important: density and performance. From a density standpoint, actually, I was just handed this morning one of our engineering test units of the next generation Direct Flash Module of 300 TB that should be available by the end of the year. This is to compare it with the latest HAMR of 50 TB, all right? Already 6x faster.
By the way, currently we're shipping a 150 TB DFM. That's been shipping since last year. Next year, by the end of the year, we should have a 600 TB, whereas the hard disk manufacturers are talking about 60 TB-80 TB by the end of the decade. Flash continues to operate at Moore's Law, which is about 34% price performance improvement every single year. It's a track record that's been established over 30 years. A hard disk, which has a track record of over 40 years, improves at roughly 12% a year. These sort of curves, these long-term curves of price performance, do not change overnight. They are very steady, very stable. It is just a matter of time. Secondly, and probably more importantly, the hard disks are getting denser, but they're not getting faster. That's a really important element.
The way they only have one head per disk, and even if they were to double it, they'd only be able to double that at two heads per disk. That'd probably be pretty much it. Let me give you a sense of that. They're spinning at a certain rate. The bits coming out of it are at the rate that the head is spinning or the disk is spinning under the head. The way I would describe it is if you took a 1 gal bucket of water and you put a 1 in hose on it, you can fill it up or empty it pretty quickly. If you take a city water tower and put a 1 in hose on it, it's going to take forever to drain it or to fill it.
That's the problem with hard disks, is they're getting so big, but they're not getting any faster. That problem does not exist with Flash. We can put it, Flash is almost parallel by definition. As the density gets higher, we can put more lanes of IO into it. It doesn't suffer from the performance issue that disk suffers from. This is what is going to continue to make Flash just a superior technology over time. Whether you believe it's now, next year, or five years from now, I mean, really the trends are all in the favor of Flash.
Maybe just to push back a little bit on that, Charlie, right? Like if you think about the workloads that do require that performance improvement, granted it would take longer, even with a dual actuator, you can't match the speeds of Flash. You're on the HDD side, it is definitely going to be slower. If the growth in unstructured data is largely coming because of Sora and AI-generated videos and things like that that maybe don't require potentially that fast response time, as you think about the evolution of unstructured data growth, where do you think that opportunity size fits within the incremental growth in overall storage?
Right. So now we get into the density portion of this, right? At 300 TB versus 50 TB, we're already at 1/6 or 6x the density, which means on average we need 6x fewer of everything else, 6x fewer power supplies. Actually, it's more than that. We need roughly 10x fewer power supplies, 10x fewer network connections, 6x fewer processors to make this all operate. We've proven to be at Meta less than 1/10 the space, power, and cooling of their equivalent hard disk environment, right? On average, the average hyperscaler data center turns out Meta's even higher than this. The average hyperscale data center is about 25% of their power going to storage, to hard disk storage.
At 1/10, not only are we the same total cost of ownership over time, but we give them 20% of their power back. If you think about data centers being limited more in power than they are in anything else, getting 20% of your power back for things such as AI is a really important consideration, right? Not having to build the next data center saves a lot of money. Even compared to, if we look at hard disk from a density and total cost of ownership standpoint, we're also hard disks about 1.5% failure rate a year. We're 1/8 of 1%.
If you talk about replacement cycle, about 10 years for a hard disk, sorry, about 5 years for a hard disk, 10 years for one of our Direct Flash Modules, it just, one thing adds on another and it just gets to the point where if it's the same total cost of ownership, even if we're requiring slow performance, but same cost of ownership and five times the performance, why wouldn't you do that?
Maybe let's switch a little bit to think about where all that NAND supply is going to come from in terms of all the bits that are needed. Because this is a frequent argument that I think we hear: if you look at, relative to the, call it 1,000 EB roughly of hard drive shipments, total output of NAND might be, I don't know, I had like 900 or so for the market, but enterprise SSDs are what is being diverted at the moment to satisfy enterprise demand, maybe only 15% of that. There's a hunk of it that's going towards smartphones, to PCs and other things. Conceptually, how should investors think about your access to those bits without having the price go up or the returns be as favorable to the NAND industry?
Right. Of course, we would like nothing more than to get the 600 EB or 700 EB ready to go for NAND tomorrow. That is not going to happen. It is going to be a ramp over a period of time. Even we have talked about the Meta ramp, one to two this year, some double digits next year. Even if we were to get another design win, that is at least two years out. These things will be ramping over time. Again, semiconductor ability to ramp capacity is legend, right? We have been working very closely now with three major suppliers, Micron, KIOXIA, and.
Hynix.
Hynix, thank you. Little brain freeze there. And Hynix, we've already been able to make sure that we had enough supply necessary for the next two years. So we feel confident there. And of course, as we see success, these companies will continue to expand their production rate.
Maybe switching away from tech for a bit, like just to think about the financial profile of the company since we do not have that much more time left. When you think about Pure's growth in the past, you have kind of gone through these cycles. And correct me if I am maybe not articulating this correctly, but I think that you have high investment years and then you have a big uptick in feet on the street. You really start to penetrate like new accounts in years where you are harvesting margin. You kind of see the inverse of it. It is not perfect, but it is kind of like over time, it is an observation at least for me, and that I think that that generally holds true. Part of, I think, Pure's challenge is that there is this incumbency that has been out there. It is sticky. Storage is a lot stickier than, say, servers.
It's a hard market to kind of penetrate. You're investing now for, it's an investment year in some ways, right? Like your growth rate's in the low double digits. You've got investment year that is compressing maybe what you would normally have had, like margin expansion into not. How should we put all these pieces together and think about how you want to manage the company over the next few years?
Absolutely. I might just re-characterize a little bit of our past. It was less about us making changes in our investment profile for different years. We've tried to keep a very steady and consistent investment profile. What did happen in past years was that despite the fact that NAND on a long-term basis declines in price, what we've seen twice now in the eight years that I've been here has been some wild swings in NAND, either going to half or doubling pricing within a four-quarter timeframe. That actually has largely affected the top line more than it has the margin, more than the margin line, while we've kept investment steady. Now, a couple of things have changed. One is that was during a period of time when Pure was much more specialized in terms of the type of storage that we could provide.
It gave us limited flexibility when NAND prices change. The second thing was that, of course, we were very dependent on effectively a CapEx sales model. Now we are in the high 40% range in terms of having an as-a-service model that has much steadier, obviously, both revenues and cash flows. I think we're much steadier now. Five years ago, together a stated strategy where it was our intent and we've had a high fidelity to this of putting another point or two of operating margin on our bottom line every year.
Now, given the opportunity for the hyperscale, which was going to come with some heavy R&D investment without revenues associated with it for a period of time, we identified last year that for this year, we were seeking to put that one to two points of addition to our operating profit line on hold for one year, which continues to be the plan. We plan on picking that up again next year. It is part, I think we have been very transparent and thoughtful in terms of the way that we are operating the company. We hope to not create surprises for our investor base.
Yeah. No, no, that's helpful context. I appreciate that. Maybe we can't kind of not talk about tariffs, I guess, to some degree. What can you say about the known state of affairs and how you're thinking about managing the various elements, including managing to a more uncertain demand environment?
Right. The known state of affairs is that if you just look at last week, tariffs were off again and then on again, all within one week. The exact nature of those tariffs, and that's one of the things that's a little bit created the additional uncertainty, is that it's even worse when you look at the fine print. That is, the details behind the large statements around tariffs is really quite missing. I think for anyone, any manufacturer, there's a lot of uncertainty associated with tariffs. There'll be announcement, for example, of a certain tariff on a certain country, but what does it actually include? Does it include the value add of that country or the entire bill of materials that comes out of the country? A lot of times that type of detail is just not forthcoming.
What we can say about tariffs is that there's still a lot of uncertainty as to exactly what that will look like when they are finally put into effect with the detail behind that. Now, the other thing we can say about tariffs is we've done a lot of planning around it. I think we're prepared for a lot of different scenarios. We do have a very flexible, very distributed supply chain. We are in quite a few countries around the world. By the way, when we talk about tariffs, you also have to talk about your demand chain. It's not just about where we source product, it's where we ship product. We have a lot of flexibility in our demand chain as well. I think we're well prepared, but until the final rulings come out with the detail behind it, it's very hard to predict.
I think that's what's creating a lot of uncertainty in the market, not just with us, but with our fellow travelers in this environment.
You brought up NAND pricing, Charlie. Just kind of curious, what's your outlook on that in sort of the next six months to a year? I know it's always pretty volatile, but how are you thinking about that? Are you doing anything around inventory and managing that?
I mean, as you know, we've announced just in the past six months two different relationships that we've extended. I think we're well prepared for any fundamental changes in what would be spot NAND pricing. I think we're pretty confident about our costs as we go into the second half of the year. Of course, what we can't be necessarily confident about is tariffs. That could make a change in the costs or in the macro, which could change the demand side.
Maybe just one more thing on margins around product gross margins. In particular, like E Series has been significantly lower product gross margins. Can you just help us think through strategically, is that where they need to be to drive penetration? What kind of penetration does it need to drive? Why is it priced the way it is priced?
Right.
There are two reasons why it's priced the way it's priced, which we signaled many, actually several years ago. That is, we believe that we had a, that we were the prime mover, the first mover in driving flash into lower tiers of storage. We wanted to be aggressive there. We are being very aggressive there. The second is this is a new product area. Customers are just not used to buying storage that's typically hard drive-based storage in their lower price tiers. To get them focused on doing this, like any new product, you generally have to come in at what are typically lower margins for an organization. Now, let me just provide some context for this. I've been part of organizations that have introduced a lot of new products into a lot of new markets before.
I have never seen a case of a successful product being introduced that represents the same gross margins or higher of an existing organization. Typically, in a new market, you have to really incent your early adopters, your early users with a very strong pricing model. That is a second reason why we're doing this here. I have always lived by the statement that with market share comes margin. As we gain market share in this segment, the margins will come up. They will. I have high confidence that we're already starting to see some of that, as we saw in the last earnings call.
My expectation is we gain more market share as customers say, "Yeah, I can actually replace or start to think about using Flash in place of disk and start to realize some of the benefits of doing that," which go way beyond just the cost. I think we'll see margins come up.
Okay, great. I think we're just about out of time, but Charlie, maybe any parting thoughts for the investor group here just in terms of why you're so confident about Pure as an investment case?
You bet. For those that have been following us for some time and even for those that have not, we started out selling a specific storage product in a specific storage area within the enterprise. Today, we can really satisfy any storage need of our customers, which is a huge change. We have done it with a lot of discipline, meaning we have one operating system that supports any need they may have: block, file, and object, very high performance, very low price. We have now introduced something we have not talked about, but our Pure Fusion that really allows us to enable the company that is the customer to operate all of their storage as a cloud of storage rather than as individual arrays. This, for the first time, is going to create a network effect in storage in the enterprise, which has never really enjoyed a network effect before.
We believe that this will really allow us to start to really penetrate even deeper into the enterprise environment.
Amazing. Thank you so much, Charlie. Really appreciate the time. Thank you so much.