Great. Thank you to everyone for joining us today. My name is Mike Cikos. I'm the Lead Analyst here at Needham covering Pure Storage. We have 40 minutes set aside here for our fireside. I have some drafted questions on my side, and to keep it interactive, I know that some people are tuned in. You can submit questions on your side as well. Before we dig into it, I'll go through some prepared remarks, and then we'll just start on the questions. With that, statements made in these discussions, which are not statements of historical fact, are forward-looking statements based upon 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 Forms 10-Q, 10-K, and 8-K. Awesome. Fun stuff is done, right?
Thank you very much to CFO Kevan Krysler, Founder, Coz and then both Paul and Sanjot for joining us today from Pure Storage. Really appreciate having you guys as part of the conference tonight.
Good to be here.
Thanks for having us.
Awesome. Very simple, but just for people who are less familiar, can we just start? What does Pure Storage do, and what is the market you are addressing?
Pure Storage, we've been around fifteen years now, and we've really been focused on three main things, leading the industry transition to all-flash. A key part of our founding thesis was that flash was going to replace all disks, and now with the hyperscale design win, which I'm sure you'll ask us some questions about, it should be obvious to everybody that all-disk is going to be replaced by flash. You know, the second thing, you know, and we're addressing the enterprise storage market. The other hallmarks of our products are being really easy and simple to use. Again, something that we've completely changed the storage industry, and our Evergreen business model. Evergreen really is two things. One is when you buy a product from us, we can update all of the hardware as well as the software non-disruptively, effectively forever.
So when you buy, you never have to rebuy. Our very first customer from December of 2011 is still running the same array they bought then. The array is bigger, it is faster, and it looks like a brand new array, and they have never had to do a data migration, never had to replace it. Then, of course, we have our Evergreen Storage service where not only does the box stay fresh forever with hardware and software, but we guarantee the SLAs, and we, in effect, run the storage for you. And, you know, so mainly focused on the enterprise storage market and revolutionizing the entire industry.
Excellent. Thank you for that. And I apologize for the background noise. The radiator in our office is on a little haywire right now. But just to Kevan, again, in the interest of dusting off the story for some folks who are trying to think back to what feels like a lifetime ago with the October quarter earnings, can you just hit on the quick highlights from that quarter, just to bring people up to speed on how the numbers have been progressing?
Yeah, I can. I mean, at a high level, you know, we achieved expectations, slightly overachieved, frankly, and guided up a little bit on revenue for the year. And that guide increase was really the result of seeing a little bit more activity of Evergreen// One, which is our storage as a service offering, seeing more of the velocity opportunities converting to a traditional sale. And as a result, that created some short-term tailwind for us on revenue for the year. And that's really what drove us to increase our revenue guide for the year. We continue to get very strong traction on our FlashBlade//E offering, and now FlashArray//E. That offering, you know, is continuing to, or those offerings are continuing to get really nice traction. Then obviously, the design win was a significant milestone for us as well, which was part of our announcement.
Great. And I'm sure we'll be hitting on all those, but just one of the things that we receive from folks too, it's a question more on competition. Are you in any way seeing new entrants more frequently? Have there been any changes in win rates? Any comment on those two points?
Yeah, no, no, we're not seeing, I mean, obviously the competition continues to be very strong, but that's not any change than what we've seen. And frankly, it's the players that you'd expect, on the legacy side, Dell, HP, NetApp. You know, every once in a while we'll see some new entrants and we'll compete against them, but nothing that I would view as significant, nor changing in altering our win loss. Our win loss is relatively consistent, as we've seen, over really an extended period of time.
Great. And then the hyperscaler design win. I mean, that is the number one thing that we are fielding questions on for Pure Storage, right? So, congratulations with this most recent quarter. You guys were able to announce that design win with a top four hyperscaler. I just wanna go big picture for a second, and just stress test this, but you guys have had this effort now underway for some time. And so what has been the path that finally got you to be able to make this announcement with that design win? Are there other hyperscalers that you guys are currently courting as well, in addition to this first design win?
Okay, so to just go over the history of the engagement.
Sure.
We started talking seriously to these guys in, let's call it, the fall of 2023, and the key thing, you know, hyperscalers, they're very different in the way they think about data centers and build data centers. They want to build them uniform. Again, they're at massive scale. The simplicity of building uniformity is critical to their ability to be at that scale, and so they've all built their own environment, their own software layers and things, and what they really want to buy is not storage arrays like we sell to enterprise customers. They want to buy the lower-level commodity components. They buy hard drives, put them into servers that they have designed, and use software layers on top to serve out the storage, so we did not approach them as selling a traditional product. We approached them as selling a drop-in replacement for those hard drives.
And so that's our DirectFlash modules and the DirectFlash software to control it. The key thing was being able to fit into their environment. They, you know, have certain ways in which they monitor the systems, report on them, deploy them, purchase them, and all of those things are at massive scale for millions of units, and we had to fit into that. What we, you know, the leverage that we're using is the DirectFlash is so much more efficient than hard drives and so much more efficient than other off-the-shelf flash drives. And so, you know, I mentioned at the beginning, Pure has been about all-flash from, you know, from day one. Starting about 11 years ago, we started working on what we call DirectFlash.
So whenever any one of you buys an SSD that is a solid-state disk that is flash pretending to be a hard drive, so that it just drops into your laptop or drops into your whatever other system, and that's inefficient. Hard drives and flash are different technologies. They have different needs. They have different ways that you want to access them efficiently. And that layer of software in those SSDs makes them more expensive. It uses up more flash. It actually burns out the flash faster, lowers the performance, and it has lots of bugs. Okay? If you've ever had a solid-state disk fail, it's not 'cause some moving part wore out. It's 'cause there was a bug in the software they put in it. And DirectFlash , it's the same hardware. You know, we didn't invent some new hardware.
We basically, we procure flash chips, common flash controller, all standard pieces, put it together. The magic is we've taken all the complexity out of the firmware in those drives, in those solid-state drives. That's why we call them DirectFlash modules, and that's the technology, that flash management, of the DirectFlash software and the DirectFlash modules. That's what we're in essence selling to the hyperscaler and being able to fit that into their environment is what led to that. Now, we started having the conversations. They got very excited about the possibilities. Over the course of the first, you know, nine months of, last year, they spent testing, you know, what we were talking about, and we spent time working together on exactly how we would fit into the data center.
And that culminated in them recognizing that by deploying Pure DirectFlash, they can make massive power savings, thus enabling them to put more GPUs into their data centers. They can, you know, have a standard layer. It will fit in. They will monitor the systems the same way. They will manage them the same way. They will be able to go through all the same procurement. And they have sort of validated all that. And that led to the design win. You know, from this point forward, what we are now doing is we are doing testing at greater scale. So, you know, if you think about how this is going to go forward over the next year or so, or a few years, so they have now decided, great, we are gonna take this gigantic business and we are gonna shift it to DirectFlash.
And this is their AI storage, their disk storage, all of it, right? Again, a uniform layer. They're gonna do testing now over what you'd call several hundred petabytes, which may seem like a lot to most of our customers, but to them it's a small amount. Then think of that as a basic Product Validation Test or PVT, kind of thing. And then after that, they'll say, great, passed all our tests. And we already kind of know it's gonna pass the tests. We're just working together still. If we didn't know, they wouldn't have gotten to the design win point. Let's just be clear on that. And after that, they'll basically say, okay, now we're building a new data center. We're gonna take one of the zones in that data center and we're gonna deploy all the Pure stuff. So that'll be some number of exabytes.
And then after they do that successfully, they'll then go and say, oh, now we're building a new data center. We're gonna deploy the whole data center as Pure. And after they do that successfully, then it'll now be all the new data centers we build are Pure, right? So it'll sort of ramp up, in that way. And that's why we talk about, you know, the ramp in number of exabytes over the next few years.
Got it. Got it. Okay. And I know you had spoken to it in your comments, but just to further hash it out, like, there is this thought that Gen AI is changing the demands on these data centers from a power consumption perspective, the energy requirements, right? What was how much of a push, how much were the hyperscalers pulled in that direction because of Gen AI or no, this was just on their radar way before Gen AI hit. Gen AI might've been a catalyst, but there was enough of an economic reason for them to make this decision even before that.
It's on their radar independent of the Gen AI, but Gen AI pushes it further. You know, I mean, you, you see it in the news. I just saw, you know, something around, oh, hey, Bill Gates or something got a new nuclear power plant approved for 345 MW and a few gigawatts of molten salt energy storage. Every hyperscaler, and, and indeed enterprise data centers, power is their biggest problem and power density. You go back, you know, like go back 30 years, and I remember when we were talking about having like a kilowatt per sq ft kind of thing in the data center and, and 2 kW sq ft. And now they're talking about a rack having 100 kW. And, you know, you pump chilled water into these things on the front of it and blow the air over.
So you're taking sort of direct cooling there and they can cool a 100 kW a rack. Its power is the biggest problem for just about every data center right now. They're all thinking about it, the hyperscalers most of all. You know, their disk storage is much, much larger than their AI storage and consumes amazing amounts of power. So by replacing all that, yes, we free them up to put more GPUs in the data center. That's great. But we solve their data center power problem. Or I shouldn't say we solve it completely, but we're a giant part of the solution. That is what would drive the disk replacement independently. Now the way the hyperscaler thinks, they want, as I said, a uniform, consistent environment, right? That's how they scale. Everything is simple. Everything is the same.
So one of the things that they find very attractive about the direct flash solution is they can run our biggest, densest, direct flash modules with a lot of them in a server behind one network port and replace their cold, cold storage. They can take that same design and just use smaller direct flash modules, which have the same performance, but because they have fewer terabytes, it's more performance per terabyte and put more CPU and DRAM and network in front of it and thus get their higher performance for the AI workloads. So they're looking at it and saying one consistent layer with just slightly different amounts of ratio of direct flash to memory to CPU to network can hit their cold, cold storage needs and their ultra hot AI storage needs, and that's extremely attractive to them.
Yep. Excellent. Thank you. Thank you for calling that out specifically. I think that there's also been a debate out there because you had built out this initiative with Bill Cerreta's team. Let me ask, now that you have this design win in place, is there still more investment needed for the sales effort behind Bill Cerreta's team or no, is the predominant focus now on this R&D initiative, like you were talking about that, I guess the scaled testing that's gonna be taking place now? Is that where we should think about most of the effort is for this design win?
So, there's still R&D effort on that. And it's things, you know, it's not challenges that are hard. For example, okay, the software we wrote, it reports all the performance data and other statistics in our format. And they're like, no, no, no, we want it in the format that we use with everything else. So we're just sort of, okay, we've gotta just do the software to print it out in that format instead of the other format. So there's still stuff we're doing there to fit into their data center. But then remember, this is the first hyperscaler. We wanna get the second and the third and the fourth. And, you know, there's, say, depending on how you count five to 10, some number in there that are really, really, really big and matter. And we're talking, you know, we're talking to all of them.
The first win has helped spur conversations. I mean, we were already trying to talk to them with varying degrees of success, and we just are now having more success in getting deeper engagements, because they all have evidence that someone besides us thinks this is a really good idea, so there'll be software efforts with that because each one will be a little different, not big software efforts, but, you know, there'll still be investment there. We need to accelerate our NAND roadmap. Okay? We've gotta push on our NAND, more NAND supply, and that means more NAND suppliers that we'll want to bring on board, it means pushing the density roadmap as fast as we can. We are already pushing it pretty fast, but we're gonna accelerate it even more as we can here, and, you know, it's highly, highly leveraged, that, that work.
We'll get great, great returns on it. But we do still need to do that. You know, beyond those top hyperscalers, others would get a version of this that is more standard. You know, sort of, it's like, here's our standard thing. Oh, you want the reporting to be different or the logging to be different. You do that yourself. Here's the API. You know what I mean? Sort of the difference between us doing it and them. But you'll still see a set of work on NAND quality. There'll be some CapEx that goes with that too as we bring in more types of NAND to test with. You know, and that's kind of where the major investments will be.
Awesome. Awesome.
And, when I say major investments, the major portion of that, compared to our total R&D, it's not some gigantic thing that's gonna bloat it up.
Right. But it's from a sizing perspective. I think it's just important to understand we're and we've talked about this in our prior earnings where our operating margin, we're expecting that to be flat next year with our expectations that we set for this year. Generally, we're expecting moderate expansion year- over- year specific to our operating margin. And so when you think about that potential expansion, that's really getting absorbed with the scale and driving the hyperscale opportunity, as well as other opportunities in the hyperscale environment.
Excellent. And great to hear as well as far as. I mean, that is the other thesis. Once you get that first design win, there should be some sort of increased activity or conversation taking place with the other hyperscalers that you've been courting. So good to hear on that engagement.
Following up on some of the NAND comments and some of the testing that you guys are gonna be doing with suppliers, great to see the expanded collaboration with Micron. Between the announcement with Micron and the Kioxia agreement, is it fair to think that you guys are now ramping up capacity and vendors to support the hyperscaler design win, or is there something else that we maybe need to consider?
No, it's largely to for the hyperscaler design win. I'll also say, you know, our E family drives more NAND as well, you know, where we sell to the enterprise, the E being the really large- scale systems designed to go after sort of hybrid and hard disk in enterprises. But you know, you do the math, we get a number of hyperscalers. We're gonna need, you know, Kioxia has been a great partner for more than a decade. Micron's been a partner for like nine years or something like that. And you know, we've, you know, say expanded those, but we're gonna go after other NAND supply too. You know, if we get to where we want to be, which is replacing all of the hard drives, we're gonna need a lot of NAND.
Yep. I'm gonna shift gears for a second 'cause I know that the platform has some key offerings that I think you guys are just architected differently, which allows you to deliver them to the market in the first place. One of the things that we've really been highlighting for folks is Fusion, right? And so Fusion became available as a free upgrade for all existing block storage arrays, and a standard in all new block products and services. I know it's still very early, but what has the feedback been thus far from the field?
It's been, you know, really positive from the customers. The, you know, earliest customers to deploy what they're seeing, you know, is exactly what we hoped to deliver with it. It's an ability to run their entire fleet as a simple, consistent layer. So very much all the standardization that I talked about that a hyperscaler does to be efficient at scale, you know, we're now able to let customers with two, three, five, 10, you know, 20, 50, 100 , whatever, you know, of our products deliver that same kind of consistency, and, you know, Fusion, it's, you know, the start of, of where we are.
We're going to be improving this for years, I mean, three, five years, and we're still gonna have a great roadmap on it because, you know, what we're, you know, able to do is take an organization and, you know, let them run so much more efficiently, so much more reliably. You know, you know, like if you have a library at home, you organize your books in a certain way and you can go in there and you can find books 'cause you know how it's organized. If you just have a jumble of books, every time you're looking for a book, you're like, oh my God, where did I see it, and you're searching through everything.
Today all these customers, they've been buying us, I buy a storage array and I configure it, and then I buy another one and I configure it, and they're all different, right? Just the way they configure it, things change. There's a new feature, there's whatever. And Fusion allows them to configure all of these products the same way. So they're organized, they're orderly, and people understand them. And that lets them run so much more reliably. And, you know, they can monitor their risk, their compliance, their data protection, their cybersecurity. It just, it orders of magnitude improvement in how good a job they can do. And one of the things, you know, we actually just did a review of this, internally, on Tuesday morning, with our Fusion team.
You know, we've got one customer, for example. They're a state tax agency in here in the U.S., and they've deployed it now on their first five arrays and they're sort of going to expand it. They've been asking for a few feature requests. One of the things that made me really happy is they're asking for what I think are really good feature requests. A lot of times you come out with a product and the customers, they see the differences between the product and the old stuff they had, and they ask for all these features to make it just as bad as the old stuff. These guys are asking for features that are right in the direction I wanna take the product. They're asking for features that will help them. Oh, gee, I'm deploying a new array.
I wanna be able to copy this configuration from over here, do a little template editing and just get it done real quick. Oh, I wanna be able to, you know, the things they're asking to be able to do are right the direction we wanna take the product, which shows that they're getting a good understanding of why the product does what it does. And that means they'll, they'll get better value, but it also means that the way we're selling it and the way we're talking about it is really resonating with them. And so I was very excited to see the kinds of feature requests we're getting from a set of these early customers.
Excellent. And I think Fusion in its current state is available for FlashArray. Is there an expectation to extend that to FlashBlade as well?
Yes, it's gonna, you're gonna see that this year, FlashArray and FlashBlade as part of the same fleet. And it's the same fleet. It's not like, here's a fleet of FlashArrays. It's like one fleet. You'll see Portworx in there, Cloud Block Store, all of it. 'Cause that's the whole benefit.
With that search capability as well, like, am I gonna be dictated or told I can only search on array versus blade or no, you, it's a mixed fleet you'll be able to run your analytics over?
No, it's one fleet, everything. The whole point of it is to let you run everything like it's one simple service.
Is there maybe to kind of make sure that the customers are seeing this, right? But does the partner channel, do they need to be quote unquote like activated to help customers understand the value of Fusion? Like how, where are we in communicating the message of what Fusion will bring to them, to the customers?
Yeah, I do think educating the partners is, you know, so that they can help sell it better is part of what we need to do. You know, the way we sell, some customers, they're very much dealing directly with us, but there's a lot that are indirectly, they're talking to their traditional partner who they work with for everything. And so yes, we have to get those folks educated on it. Again, I come back to the early feedback from customers. The customers are getting it. And so if the customers are getting it, I think the partners will be able to get it, as well. You know, when you come out with something new, getting people to understand that message, getting them to not yearn for the past, but to look forward to the future is one of the hardest things.
Like I said, I'm really happy to see that resonating. As you can imagine, when we came out with FlashArray and then FlashBlade, the first several years of both of those products is just an endless stream of, no, we're not going to do the same feature you hated in the past. I cannot tell you how many times I've refused to add some of those features to our products.
So we did have a couple of questions coming in. I wanna be true to my word to make sure that we're honoring those. The first one here is asking how would you characterize your on-prem non-subscription business and how do these margins compare with the SaaS business?
Yeah, I think over time, the as a service business will generate improved gross margins for us as well as improved operating margins for us. I mean, obviously the selling effort associated with the renewal of an as a service agreement is gonna be far different than selling via traditional sale. That's really what's helping drive the operating leverage. Then generally, you know, we can operate the service generally more efficiently across the platform than our customers can, which really drives more efficiencies for us on the gross margin level as well. That's kind of how I would view the gross margin profile and the operating margin profile when comparing and contrasting a traditional sale versus an as a service model that we sell.
Great. And another question here asking about some of the new products and initiatives, beyond just fusion, but looking for a status update on the Gen AI pod, AWS Outpost relationship and Azure VMware.
Wanna hit those?
So, okay. So, the Azure, you know, Cloud Block Store with Azure. So that's moving along. You know, you go through the different phases.
Maybe you wanna explain a little bit just so folks understand.
Oh yeah, actually.
Yeah.
That's a good point.
Yeah.
Okay, so Cloud Block Store runs currently on Amazon and Azure, and it is effectively the Pure Storage software running on the infrastructure services supplied by the hyperscaler. So it's not putting our hardware into a hyperscale environment. It's running our software on top of whatever storage and compute services that they supply, and you know, the basic premise behind it is to enable people to move applications from on-premise to the cloud or from the cloud back to on-premise, without having to change the application. You know, when you build an application native in the cloud, it's sort of trapped there and can't go back and forth.
And there's a whole lot of people that, you know, obviously have had stuff on-premise that they've wanted to move to the cloud, and then they have to change how they run it because you can't run a VM the same in the cloud as you can on-premise. Then if you're using various services, you switch to the cloud-based services. So we started with that on Amazon, running on AWS. We achieved some, you know, nice successes there. One of the issues that we've had there is, you can't boot from one VM, you know, boot one VM from another in the cloud. And so that still necessitated some changes. We now have the Pure Cloud Block Store running on Azure, and it's part of the Azure Marketplace. So you can buy it through that, as well.
The Azure Solution, the nice thing is you can run that and boot VMs from other VMs there and such. It, it's sort of a much better VM solution there. And so you can seamlessly move things back and forth from on-premises to the cloud using that. You can replicate, use the cloud as your disaster site, use the cloud as a bursting site, you know, exactly the kinds of things you want to be able to use the cloud for. You know, we've achieved some good success with a few customers there. We're continuing to work out, you know, where you end up with issues. Customers aren't used to the unreliability of the Cloud Services, right? So Cloud Services, like in the data center, you know, you see us and our competitors always talking about running five- nines, six- nines of availability.
The cloud runs like three- nines, four- nines of availability, and so people have to they can't just blindly go there. They have to think through the architecture a little better, make sure, for example, oh, I'm doing disaster recovery between two zones. So I'm independent of failures, not set up everything. So it's going through my one Cloud Account Manager, so that I can't do it. So we're working through getting people to do a better job of deploying that, but you know, the product's out and people are starting to use it more and more and achieving some good success there with Portworx. Portworx is container-based storage services. We purchased a company back around the start of COVID.
Now, containers have slowed down in their adoption because as everybody went through COVID and then the time after COVID, the easiest thing to cover or, or to cut back on, I should say, was your new initiatives, which was your new applications being containerized. So again, with Portworx, what we're seeing is some really good successes with the customers that are using it. We've got a number of, you know, very large customers and very large deployments, but, it's been slower to ramp than, you know, we would've hoped because containers have been. We are still 100% committed to the fact that containers are going to be the future. You get into the next decade, every new application, every application being done, will be containerized, and Portworx, and that's where Portworx syncing with Fusion as well is important.
You know, Portworx is really the only solution for stateful containers out there, you know, for managing their storage needs. And so, we still think it's got a fantastic future. It's just taking a little longer to get there than we wanted.
Excellent. And then the last one that we'll probably have time to hit before we have to wrap it is the Gen AI pod, and some of the partnerships that you guys have with NVIDIA.
Yeah, so you know, AI is driving, you know, there's basically training environments and inference environments. The inference environments, you know, our existing products have just been great for, because it's really using your same data, but using it more as it were. The training is a lot more like HPC. We and the other legacy players haven't done quite as well there. You've seen some of the niche players like a DDN or a VAST that have, or WEKA that have, done better there, but we now have SuperPOD certification with NVIDIA and a much stronger partnership with NVIDIA, and so what you see is in things like some of the cloud GPU offerings, you know, not the hyperscale cloud, that's one thing, but you know, the GPU cloud providers, you know, we've achieved some pretty good success in those.
And, you know, we expect that to continue and to grow. You know, we've done a partnership with CoreWeave, and you know, we're doing other things there that are just designed to capture that market much better now that we have the SuperPOD certification with NVIDIA.
Excellent. Excellent. And I think we'll leave it there, but thank you very much for the time, guys. Really appreciate it. And good luck with the rest of the meetings, everyone. Thank you.
Thank you. Appreciate it.
All right. Talk soon.