Everpure, Inc. (P) Earnings Call Transcript & Summary

Good morning, everyone. Thanks for joining us here at the BofA Technology Conference. I'm Wamsi Mohan, and I cover IT hardware for BofA. Thank you all for joining us this morning. Today, we're delighted to welcome Pure Storage. We have CEO, Charlie Giancarlo; and CFO, Kevan Krysler with us today. And before we get started, I do need to read otherwise, Paul is going to come up and beat me up over here, the safe harbor statement. So 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's most recent SEC filings on Form 10-Q, 10-K and 8-K.

With that, let's get started. So well, thank you again for being here. Charlie, you just reported a very solid quarter, right, like you beat by about $20 million. Your guidance came up also ahead in a fairly difficult macro backdrop. So can you talk about where you saw the upside relative to expectations?

Yes. I think a couple of things are really working for us. One is the -- we sell based on superior performance with a better total cost of ownership. And that matters in an environment like we're operating in today. That total cost of ownership comes from what we -- I'll go into later. But unique competitive -- sustainable competitive advantages that we have that, as I mentioned, I'll go in. Also, we've been focusing a lot on sales force training. And as you might imagine, with the kind of euphoria that we've had over the last several years, the focus on getting back to basics and selling into a tight economy were perhaps not reinforced. And so we've reinforced that, plus really focusing our sales force now on being able to go in and work with the customers to really be able to provide an infrastructure to satisfy all of their storage needs. Up until now, Flash could only address what was called primary storage, primary storage being the storage for the most highly performing environments because of Flash's higher cost relative to hard disk. Now we have a portfolio that can address all storage within a customer account. So all of the disk-based storage that traditionally was too lower price for our flash to be able to go after. We can now compete for the entire storage of state in our customers' environment and getting our sales force fully engaged on that aspect. So between the TCO benefits, being able to sell across the entire account, getting better at being able to sell into a tight economy, I think that's starting to make a difference.

Yes. No, absolutely. There's been a lot of talk about AI pretty much in every discussion. And we wrote about your product, I think, now several years ago. I guess as you think about AI, what about Pure's architecture makes it sort of a good destination for AI workloads.

Yes. Actually, there are a couple of very different reasons why we believe that our architectures are uniquely suited to an AI environment. First of all, we first built a product called FlashBlade 5 years ago in anticipation of AI workloads. FlashBlade is a product that's highly parallel, so-called scale-out architecture, able to deliver hundreds of gigabytes of data simultaneously across hundreds of different streams to GPUs. And it's really made that product the go-to product for AI environments, machine learning environments. Part of the reason why Meta in their research supercluster satisfied or satisfy their requirement with FlashBlade. The second reason, however, is not so obvious, I think, with all the AI euphoria that exists right now. And that is that the machine learning environment requires that very high performance. But the inference environment or rather the ability once the models are created, to be able to use data for answers, requires data to be generally available in what we might call a warm environment. It doesn't require that type of highly -- hugely performing parallel environment just requires data to be available. And most data today is locked up in hard disks, in siloed environments that aren't even in many cases connected to an environment that can use that information as inference. And so we believe that AI is going to drive enterprises to put more of their data in an environment in systems that can support both their primary use case, which might be back up but also things like analytics and AI. And so it's going to shift data from being in cold environments to warm environments. And so this is perfect for flash because now we can replace the disk environments at prices competitive with disc, but with much more highly performance systems. And so we believe that AI is going to drive more data out of silos, into warm environments. So between our high-performance systems to drive machine learning and the lower-priced systems that just takes data that's today trapped and putting it in available environments, we see great opportunities.

Well, what percent of your workloads are AI based workloads? And do you have a view on...

We really don't break it out, but it's a substantial number. It's a number that's meaningful every quarter. But we -- it's one of those -- we try not to -- don't try. We don't make a practice of breaking out individual -- either products or product segments.

And do you have a view on whether AI is going to drive increased growth in on-prem storage spending?

I think this year is a year of planning for enterprise storage spending. As you know, budgets are constrained. But I do think that, certainly, as we start to emerge from this tougher economic environment, I think it is going to allow the enhanced spending in enterprise IT to continue to accelerate going forward.

I know you guys, in your annual report spoke about TCO advantages. So would you mind talking a little bit about TCO advantages relative to your competition, particularly other all flash vendors. And maybe we can break this out in different ways. But first of all, maybe where is that advantage coming from for Pure Storage versus HDD players then we [Technical Difficulty]

Not just buying the system, but then operating it for a number of years. And I mentioned earlier that we have unique sustainable competitive advantages. And that's where our superior total cost of ownership come from these advantages. So if I list them as 4 advantages, the first one is that, we have our operating system, which we call Purity. Purity is the only storage operating system out there in the world that operates on raw flash rather than SSDs. To put that in perspective, SSDs were a brilliant invention by the flash manufacturers who wanted to get flash into things like laptops and desktops and servers and storage systems. And they said, well, what's the best way for us to do it? Oh, we have a good idea. All those systems use hard disks today, why don't we design flash to look like a hard disk. In my opinion, that's like making a personal computer look like a type writer. You're taking a semiconductor and making it look like a mechanical device. But it didn't require any of those systems to change their software. So of course, SSDs took over. And everyone, literally everyone else other than Pure uses SSDs in their systems through laptops to use SSDs, the servers inside the major data centers use SSDs, all of our competitors use SSDs. And to the extent that hyperscalers use flash, they use SSDs because they didn't have to change any of their software to do it. But SSDs have compared with us, higher cost, higher complexity, lower lifetimes, lower reliability. And so it really limits if you will, the power of the raw semiconductor. Our system uses a direct to flash. We control it out of our main software. We manage a flash across the entire array and even set of arrays. And that software is called Purity. That gives us better overall cost and performance than anyone else out there. Second, we have the most consistent and consolidated product line in the business. Most of our competitors have somewhere between 3.5 dozen different system designs to handle that full range of high performance, low cost but also a block file and object, and they have lots of different systems to do this. We have one operating system purity, operating on 2 hardware architectures, scale up and scale out, both of which use the same digital flash module, all of which is -- and by the way, that same software operates on the hyperscalers as well, AWS and Amazon, all of it managed by one management system called Pure1. So it's a highly consistent product line to cover the entire range of customer needs. And then you add on top of that our Evergreen program. Evergreen, and this is, again, something that's quite unique in the world. We're able to consistently and constantly upgrade all of our systems on a continual basis now for over a decade, nondisruptive, meaning we never take the customer's environment down when they do so. So it really is a SaaS service, regardless of whether it's on the cloud or on-prem because the service is always up and operating and always looks like our latest release, not just software but hardware as well. We have customers that have been upgraded over a decade now, where the system they bought 10 years ago looks like the system that we sold last week. So that's a really unique set of capabilities. You put all that together and it results in a much lower total cost of ownership. So taking it from an user's point of view, we are 10 to 30x more reliable than -- not just hard to [Technical Difficulty] to operate than our competitors. So that also results in lower total cost of ownership. Those customers that have upgraded just pay us their subscription every year. They've never had to buy new hardware. That lowers the total cost of ownership -- compared to SSD manufacturers are flash less twice as long. So instead of 5 years, last at least 10 years. So we can guarantee it forever because by that point, it's such a small portion of the price decline of flash. It's easy to upgrade. So we protect them economically basically forever. So again, you add all those things together, and it's just a far -- not only a lower total cost of ownership. But the customer can go to bed at night and go away on the weekends or vacations and don't have to worry about their -- about their storage systems failing.

And let's touch on that power and space as well.

Power and space, 2 to 5x -- we're 10x more power and space efficient than hard disk and 2 to 5x more power and space efficient than our SSD competitors.

What drives that 2 to 5x versus SSDs?

The density of flash that we can put in our systems that are managed by those controllers versus SSDs. SSDs are far more complex. They can't become as dense. They require more processing power to manage overall. And so we're able to get more petabytes per rack unit and use far less common equipment to manage that than our competitors.

And Charlie, you also mentioned the fact that your flash would last longer. What is the reason for that? Is this the wear leveling algorithms that you use relative to what SSDs can do?

So it's a variety of things, one of which is exactly that. Remember, we're doing wear leveling -- I'm glad you figured that out pretty easily. We can do -- we do wear leveling across the entire array whereas our -- the competitors don't do it. They allow the SSD to do it, and it's only doing it within the SSD itself. The SSD also is much more complicated. It has a lot of DRAM in there. It's the DRAM that tends to fail more often than the flash itself. They have super caps in there. They tend to fail -- they have bigger super caps, they tend to fail much more often if -- so just a variety of different things.

Okay. That's helpful. You announced a large cloud block [Technical Difficulty]...

We're operating natively on Amazon or on Azure. We support those 2 today. The way to think about it is just like the Purity software that operates on-prem, but instead of controlling our digital flash modules, it's controlling internal Azure or internal AWS storage infrastructure. But it's providing all of the enterprise features that customers expect for traditional workloads that we provide on-prem. And all of the APIs, all of the interfaces are exactly the same. The reason why customers use it is, one, to get those additional features like snapshotting replication, things they expect from enterprise storage that they don't get in the cloud or they'd have to implement themselves. But 2 is we have advanced data reduction capabilities, everything from thin provisioning to deduplication, et cetera, so that the combination of what we charge and what they pay the hyperscaler for the storage underneath our system is far less than they would pay for the native cloud storage. So the one customer that we announced this quarter was -- we announced it because it was close to an 8-figure customer for this -- for a Cloud Block Store, they estimate that it's going to save them about half of their total storage spend. We estimate -- I believe it's actually going to be more than that.

Okay. That's really impressive. Can you talk a little bit about Meta? And what sort of use cases as Meta use Pure for? And do you see that expanding to other use cases over time?

Well, our audience really took note of the one use case, which is the research supercluster. To be clear, we announced that for a variety of reasons. They announced it, it did affect our reporting. And so therefore, it was called out. And Meta, of course, announced their entire research supercluster, which they've advertised, and we agree, is the largest AI supercomputer in the world. But it's not our only relationship with Meta. We've been supplying them in different areas of their business for quite some time. That being said, we don't believe that the architecture that we've used in the research supercluster is significantly different than what they would use, a, for any AI environment; but b, even for their mainstream environment. And let me just give a little bit of background on that. So there are 2 major parts to our infrastructure for that RSC. One is the machine. The part that feeds the machine learning infrastructure. This is our very high-performance scale-out, highly parallel FlashBlade system. This was the premier system that I mentioned. We developed starting 5 years ago to focus on AI environments. It's machine learning environments require highly powerful streams and very, very fast data to keep the GPUs very busy. That's about 10% to 15% of the data storage that we shipped into the RSC. Over 80% of the data storage we shipped into the RSC is what would be called bulk data. It is what intern feeds those FlashBlades. It is much more price sensitive than the higher-performance systems. And that is just bolt data that's stored -- it's the same kind of data that would be stored in their AI environment that might be stored in their other bulk data environment, just Facebook. And we believe that could be true in any hyperscale environment. So that, as you see, when you're dealing with AI, you have what I'll call hot data that has to be fed at these very high rates, but then warm data has to be prepared, ready to go in those dates. That warm data environment, which we ship now with our flash array see, but also our newest product, FlashBlade, exists everywhere. It exists in hyperscalers and is now price competitive with disk. Just one last thing. One of the big reasons, there were many reasons we won that RSC footprint. Performance was certainly one of them. But another one, especially on the flash, they could have back-ended our FlashBlade//E with their software and their disk-based environments. We were the only vendor, the only system that could fit within their data center footprint, both power and space. Without our solution, they would have had built another data center at tremendous expense. They would also have had to source more power, and they didn't believe they could source more power. So the power and space arguments, once you start running up against those walls, it starts becoming a very compelling.

Yes. That gets very real at scale. Maybe one for Kevan. Can you describe what sort of -- what's the impact of revenue and margins in periods of like NAND price extreme declines or increases? How does that flow through the P&L for you?

Yes. I mean, it's a great question. And frankly, we benefit from both scenarios, whether NAND is -- actually all 3, whether NAND stabilized, whether it's coming down, or frankly, when it's increasing as we've seen historically. And the reason being is, if we think about the current NAND environment, where we've seen a pricing decline, that probably accelerated our FlashBlade//E offering by about a year in terms of coming out with FlashBlade//E. So obviously, our penetration of disc and that opportunity gets accelerated as our NAND pricing declines. We take advantage, obviously, with the fact that we can negotiate directly with last vendors. And that's been a huge benefit for us, especially as they've been challenged from a demand standpoint. And then obviously, that parlays itself from a margin profile. And you've seen a little bit of that in our product gross margin ticking above 70%. Ideally, we want to keep that in the high 60s because there's so much opportunity for market share capture for us. But I do think there's a strong upward bias for us, especially in this environment. Now we do see NAND stabilizing potentially in the back half. So we'll see how that plays out. But even in a stabilization environment, given where we are with FlashBlade//E and other opportunities we're looking at, I think the benefits to product gross margin continue in terms of strength, in terms of what we've seen, but also subscription gross margins. And obviously, we can leverage the benefits from NAND on our Evergreen//One offerings in addition of the fact that with our overall Evergreen model, we're leveraging these assets, we're bringing back to modernize and refurbish and put them back out there in terms of our infrastructure to service level agreements for Evergreen//One, which obviously helps us and enhances us on our subscription gross margins.

Okay. Yes, that's super helpful. Charlie, you made a comment on last call, no new HDDs would be sold in 5 years or [ more ], right? And so as you think about the NAND pricing today, so 2 questions here. One is HDDs are also evolving, there's HAMR technology coming. How do you think about sort of price per bit storage with respect to that? And second is, this has kind of been a very tough NAND pricing environment. Why aren't we seeking higher replacement of HDDs given that this is probably as favorable NAND pricing declines we would see in a while?

Yes. Let me start with the second one, and then I'll get to the first. First of all, if you look at SSDs, they're still quite a bit more expensive than hard disks, and they've not yet reached or the system level -- system pricing using SSDs hasn't reached parity or even closed the parity with hard disks. When I speak about our ability to match hard disks [Technical Difficulty] -- okay. So that will take -- it doesn't -- at that level, it doesn't really matter, right. The reason why we're so certain about hard disk is for a variety of reasons. One is, if you look at -- first of all, whether it's TAM or anything else, if you look at that the price curve of hard disks, it's been flattening out for multiple years. So they're not able to maintain the same rate of decline in terms of price per bit that they once had. Secondly, flash has been catching up on our price per bit over many years. And those 2 lines are going to intersect in the next 3 or 4 years regardless. At a system level, we've already intersected, right? So we believe we're going to be there. Furthermore, if you take -- whether it's TAM or anything else, just take the road map at face value that the hard disk manufacturers have put out there. It's just not going to keep up with ours. We're going to be at -- we're at 48 terabytes on a card today. We'll be at 75 later this year. We'll be at 150 next year. We'll be at 300 a year after that. If you look at the hard disk road map, they're talking about 24, 26, 28 terabytes. We're at 300. They're at -- at 1/10th the space, power and cooling. They just won't be able to keep up at a total cost of ownership level, and they're not going to be able to keep up at a density level. So -- and finally, you have to look at their ability to invest. 10 years ago, hard disk were in everything. 15 years ago, they were in iPods, okay? Remember that the old iPod had a hard disk in it, okay? It was -- they were all the laptops, they were in all the desktops, they were in all the servers, and they were in the hyperscalers, and they were in all of the enterprise storage systems. Today, they're only in 2 environments. They're only in the hyperscalers, and they're only in enterprise -- low-cost enterprise storage, right? There's no -- they're no longer primary. Primary storage is all flash. Once those 2 start to decline, there's no market left for hard disk. There'd be no more investment capacity left for hard disk. So they're going to be -- it's going to be a challenged environment for sure.

The rate of declines that you've seen though on the flash side, when you think about the CapEx that is needed to continue to go down that pace of price declines, do you think that the flash providers are motivated to induce them on a CapEx that's necessary to keep that going?

Well, to give you the sense, flash today is over twice the business that hard disks are, right? Total -- and I'm just talking at the semiconductor level, let alone at the SSD and at the system level. So the CapEx is already there. One of the great things about flash is it's about 5 node generations behind processors, okay? So there's a lot of old technology that they can continue to use to improve their overall performance. So they don't rely on any dramatic improvements in physics. It's just following the generational curve that is set out by memory and by processors. So no, I don't -- there's no worries there.

30 minutes is and finding a very little amount of time to get through so many things that I want to talk with you, but maybe 2 things to wrap up, right? Like so first, just on margins. Maybe Kevan, they can talk about what do you think is the right trajectory for margin improvement [ had on ] this year? Last year had some 1x that's causing margins to dip this year, but as you think over the next few years, what is the right trajectory for operating margins for Pure?

Yes. Look, I think we've got a fair amount of opportunity for leverage throughout the business and the way Charlie and I are looking at it is about 1 to 2 points of expansion per year of operating margins, how we're thinking about it beyond this year.

And that includes what kind of investment and sort of OpEx or is that anticipating an OpEx investment commensurate with sales growth, under sales growth?

It's great. Yes, look, we're going to still prioritize growth, profitable growth, and we believe that's very feasible. And the areas where we think we'll get that leverage very easily will be the continued gross margin expansion of our subscription gross margins where we think there's a lot of upside long term in that area. R&D spend as we continue to expand our global talent of software engineers, we've been very successful both in Prague and Bengaluru. And then look, as we continue to penetrate with our portfolio, the enterprise and the hyperscalers, that's obviously going to give us more scale and leverage on sales and marketing. So lots of difference of opportunities for us as we see about the leverage for operating margin.

Okay. That's great. Well, we just have a minute or so. So Charlie, maybe you want to take the opportunity to just [ pass ] the audience about the way Pure is a good investment at this point in time and just maybe any wrap-up thoughts you have.

Well, I think it really comes down to the fact that we do have these really unique sustainable competitive advantages that our competitors. And even in a sense, the hyperscalers are not going to be able to replicate on their own or at least not very easily. And that is the direct to flash system that we've developed over a decade and continues -- every time there's a new flash technology, new flash chip. It's something that we have to continue to stay on top of to be able to leverage the best out of any new flash environment. Everybody else uses SSDs, and that gives us a distinct advantage. It's not easy to replicate it, and you have to start from scratch to do so. The fact that we have a single consistent product line, that's very, very flexible. Our competitors leverage multiple technologies, multiple software, multiple management systems. At the end of the day, if you're an enterprise, standardizing on fewer components is generally a very good thing. And you might say, well, yes, but customers don't like to have necessarily a single -- they want dual sources of supply. Well, yes, but they don't necessarily want to use 6 different types of products for the same area, which is what they have to use when they satisfy it -- when they rely on a different competitor of ours. And then finally, our Evergreen system, where we provide a SaaS-like service even for products that customers buy with CapEx because after they buy the initial product, it stays relevant and updated forever after it. It's why we call it evergreen forever. The products never get old, which is wouldn't we like that in the rest of our lives. To have something that never gets old is always updated and never goes out of service. So I really believe that these are things that are going to allow us to continue to thrive and to lead in this industry.

Excellent. Well, with that, unfortunately, we are out of time and have to wrap. But Charlie, Kevan, thank you so much for being here and look forward to seeing you at Pure//Accelerate next week.

Same. Thank you.