Home / Transcripts / Everpure, Inc. (P) · December 3, 2024

Everpure, Inc. (P) Earnings Call Transcript

December 3, 2024

New York Stock Exchange US Information Technology Technology Hardware, Storage and Peripherals earnings 63 min

Earnings Call Speaker Segments

Paul Ziots executive
#1

Thanks, and good afternoon, everyone. Welcome to Pure's conference call for the third quarter of fiscal year 2025 earnings. Joining us today are Charlie Giancarlo, our Chief Executive Officer, Kevan Krysler, our Chief Financial Officer, and Rob Lee, our Chief Technology Officer. Once Charlie and Kevan finish their prepared comments, we'll open things up for questions. We issued our press release after the market closed, and it's available on our website, where this call is also being webcast live. You can download the slides for this webcast at investor.purestorage.com. Today we'll be making forward-looking statements, and those carry a range of risks and uncertainties. They cover things like our financial outlook and operations, our strategy, technology and what makes it advantageous, our existing and upcoming product offerings, plus competitive, industry and economic trends. Every forward-looking statement we make reflects facts and assumptions as they stand today, and we're not obligated to update any of them. What actually happens could differ materially from what we've forecasted, and results we report shouldn't be taken as a sign of how we'll perform going forward. Some of the risks and uncertainties tied to our business are discussed in our SEC filings, and we point you toward those public filings. On this call, every financial metric and the growth rates tied to them are non-GAAP measures, except for revenue, remaining performance obligations or RPO, and cash and investments. You can find reconciliations to the most directly comparable GAAP measures in our earnings press release and slides. This call is streaming live on the Pure Storage Investor Relations website and is being recorded so it can be played back later. A webcast archive will be posted on the IR website and belongs to Pure Storage. Our quiet period for the fourth quarter of fiscal 2025 starts when business closes on Friday, January 17, 2025; with that, I'll hand things over to Charlie.

Charlie Giancarlo executive
#2

Good afternoon, and thank you for joining our Q3 FY25 earnings call. Pure Storage turned in strong results for the third quarter, with revenues and operating income both coming in above the guidance we provided. As I've mentioned on earlier calls, we've been working toward securing a major hyperscaler design win before the year was out, and I'm happy to report that within the past few weeks we signed exactly that with one of the top four hyperscalers. This marks the very first design win ever to supply flash for standard hyperscaler storage, and it serves as the opening wedge for flash to handle all online storage across major hyperscale environments down the road. To give you some background, right now, between 60% and 70% of all Hard Disk Drives purchased globally go to hyperscale. Aside from providing storage that keeps costs low, this Top 4 hyperscaler picking up Pure technology ought to open up major savings in power and space across their data centers. On top of that, it should significantly lower failure rates and maintenance expenses associated with legacy disk storage, and it should double how long they can expect their storage infrastructure to last. With Pure Storage Purity Software and DirectFlash Technology, this hyperscale customer gets the pricing, speed, density and power required to deliver their cloud-based services at a level of performance and energy efficiency that nothing else matches. Our two organizations have worked closely together on engineering, and this leading hyperscaler has done a great deal of testing. Together, these efforts have proven that DirectFlash from Pure is now capable of providing affordable data storage at capacity levels built for hyper-scale, extending all the way to the inexpensive bulk data price tiers. To show how large this opening really is, we are partnering with hyperscalers to bring our technology to bear via one uniform architecture spanning the entirety of their on-line storage — from budget-friendly bulk storage, to nearline storage with greater performance, to fast storage serving their most demanding scenarios, AI included. Also unveiled today: we're expanding our work with Kioxia, a worldwide frontrunner in NAND Flash, to satisfy the growth in Flash demand we expect from the hyperscale sector. In our dealings with the hyperscale community, Kioxia has proven a dependable ally. This design win shows that DirectFlash from Pure is set to displace hard disks across the board, and NAND suppliers are paying attention and mapping out how they'll capture a market worth 700 Exabyte annually. With data volumes climbing, our paired technologies let hyperscalers handle the strain of growing data volumes while cutting power usage, labor, and the physical footprint of hyperscale data centers. The design win that was announced today has its roots in efforts that began more than twelve months back. Standard product sales to hyperscale customers had already occurred, but these hyperscalers build their own software for storage services, running it on commodity HDDs and SSDs. Our initial outreach to hyperscalers was met with skepticism regarding our ability to achieve the price points and speed required to replace low-cost hard drive storage. Despite that, this particular hyperscaler chose to explore the opportunity further alongside us, and via collaborative engineering, Pure enhanced the Purity design together with our DirectFlash technology, enabling smooth integration with their compute and storage architectures, while simultaneously adjusting the cost structure to align with their financial targets. This win marks Pure's entry into an exciting new hyperscale market. The design win on its own indicates that this top four hyperscaler's future data centers are cleared to adopt Pure's technology as their standard for data storage. We anticipate initial field trial buildouts during the next year, followed by substantial full production deployments — on the order of double-digit Exabytes — in calendar 2026, which lines up with our fiscal 2027. Meanwhile, conversations with other major hyperscalers are ongoing. Because of the considerable opportunity present with this hyperscaler and others, we expect to boost investment in our hyperscale Line of Business across the coming year, and Kevan will share more specifics on that. Let me now turn to Artificial Intelligence, which represents another big opportunity area for Pure. There are several important openings that AI creates for us. First off, we continue to provide top-tier, high-performance storage designed for GPU farms—both public and private—used in machine learning and training. Last quarter, we achieved official certification for NVIDIA's DGX SuperPOD architecture, which is designed to offer turnkey infrastructure for the largest training environments globally. We also recently shared news of a strategic investment partnership with CoreWeave, a GPU cloud provider that specializes in this space, aimed at improving how we serve our AI customers. Expanding on what we've already achieved with a super-computing scale installation that supports thousands of GPUs, we teamed up with CoreWeave so that Pure Storage could be offered as a standard choice inside CoreWeave's dedicated cloud offering. Beyond that, a lot of enterprises are now evaluating inference engines along with retrieval-augmented generation, or RAG environments, as they look to use commercial large language models to examine data they own. During this quarter, we rolled out the Pure Storage GenAI Pod, a collection of full-stack solutions that cut down on the time, cost, and expertise needed to launch generative AI projects; in that same quarter, we closed an agreement with a medical device manufacturer that had been dealing with multi-million-dollar interruptions, since their legacy storage technology was unable to support a real-time AI imaging system designed to catch defective products. Pure gives them the ability to run AI analytics, gather metadata, and teach their systems to spot and stop flaws—delivering major gains in how they operate and assure quality. Third, AI keeps pushing customers to update their setups and tear down the walls between infrastructure and data silos so information becomes simpler to reach. Different from what other vendors offer, those who choose Pure won't have to juggle separate storage operating environments for their assorted AI requirements; during this quarter, a top global provider to the defense and aerospace sector picked Pure to handle its AI data storage infrastructure. A Fortune 200 company picked our platform to handle everything from training to inference to fine-tuning, letting many teams and AI projects share data sets and storage environments without friction. They plan to build several AI technologies on top of the Pure Storage Platform, with the goal of boosting human abilities, improving aviation safety, reducing what pilots have to carry, and delivering autonomous systems designed around people. Building on top of the Pure Platform and shifting focus to the enterprise side, Pure is fueling the largest transformation in enterprise storage since Flash came along; through Pure Fusion, we're changing enterprise data by virtualizing both data management and storage, which lets enterprises build their very own data cloud environment spanning their worldwide operations. This quarter, Pure Fusion becomes available as a free, non-disruptive upgrade for every existing Pure block storage array, and it will come standard across all new Pure block products along with storage service offerings. Early next year, Fusion will reach our file and object platforms as well. The strides we're making in data storage innovation, serving enterprises and now hyperscalers too, are reshaping the industry. Thanks to what we've learned and built around making flash storage work better for large organizations, we can now go after the largest hyperscale customers, bringing our DirectFlash technology, which is built on Purity, into play at massive scale. Because of how big these hyperscalers are, the way they manage storage looks nothing like what traditional enterprises do. At traditional enterprises, storage comes as separate arrays, and each array serves just one set of workloads. If an array is set aside for a single workload, no other workload can use it. That means information stored for one workload usually stays out of reach for any other workload. So the storage systems and offerings that traditional enterprises rely on end up creating silos of data. In contrast, hyperscalers maintain just a small handful of storage environments, divided only by price-performance tiers — low, medium and high, for example.

Charlie Giancarlo executive
#3

Every workload and every customer's data sits on the same storage environments. That makes getting to data much simpler. Software, rather than purpose-built hardware, delivers the different storage capabilities. Through our work alongside Hyperscalers, we've learned how to carry the finest qualities of data cloud architectures into enterprise data centers, and that's what Pure Fusion version 2.0 does — it ships this quarter. Businesses can now construct an enterprise data cloud of their own, thanks to Pure, blending on-prem and cloud seamlessly so they remain agile and competitive in the AI era. With Pure Fusion, data management runs on autopilot, operations get simpler, and the devops developer experience improves. Fusion gives enterprises the means to build their very own enterprise data clouds that federate storage spanning cloud and on-prem environments alike, which brings effortless scalability, global accessibility, automated job placement, load balancing and, crucially, AI-ready data access. Organizations can use Fusion to define and standardize customized, global, data management classes of their own, covering performance, cost, resiliency, recovery, and location. It also automates delivery of data services to users through API, in line with enterprise policy. Across the Pure Platform, on-premises and in-cloud, Fusion completely unifies, automates, and delivers the Cloud Operating Model; we've likewise strengthened our partnership with major public cloud vendors for enterprise services. Through the official preview of the Pure Fusion powered Pure Storage Cloud for Microsoft Azure VMware Service, or AVS, we make enterprise migrations from on-premises VMware environments to AVS simpler, allowing storage to scale independently from Azure compute nodes. This first-of-its-kind solution guarantees a smooth, efficient cloud transition with minimal IT disruption, lifting cost efficiency, data resilience, and storage simplicity. Now, on the market and the wider macro environment: no meaningful change has appeared in the overall landscape, which stays relatively consistent with the muted IT spending and heightened competitive environments seen throughout the year. Higher software, SaaS and Cloud costs remain a struggle for customers, along with AI spending uncertainty, which puts unanticipated pressure on operating budgets. Evergreen//One could have shown more strength this quarter, and I would have welcomed that, yet our confidence is strong that we're well positioned across every segment. The cloud isn't a location; it's an operating model that enables self-service, speed, consistent operations, and faster scaling with greater efficiency, at lower costs. Together with Fusion and our data storage platform, the vision of an enterprise data cloud is becoming a reality for enterprises. Industry analysts have recognized our consistent innovation year after year, such as our recent 11th time Leader position in the 2024 Gartner Magic Quadrant for Primary Storage Platforms, and 4th time leader position in the 2024 Magic Quadrant for File and Object Storage Platforms. Energy availability concerns the whole globe, and has turned into a critical risk to hyperscalers' operations. Certain ones are even signing contracts with nuclear power plants to lock in a dependable supply of electrical power. Low-cost power is no longer what hyperscalers chase; they want power at any cost. Bringing more generation online won’t deliver meaningful new capacity for a long time. On the other hand, swapping out power-hungry hard disk storage in favor of Pure DirectFlash gives hyperscalers one of the biggest energy resources they can tap right now. With power constraints increasingly standing in the way of data center expansion, Pure Storage stands alone in being able to let hyperscalers both upgrade their storage in a cost-effective manner and, at the same time, release enormous quantities of electricity and floor space for other uses, AI among them. All told, we feel great about how far we’ve come on the hyperscale front and about how our enterprise offerings have grown. Speaking for myself, my enthusiasm for what lies ahead for Pure has never been higher, since we are leading a new chapter in how data storage is managed. With that, I’ll hand things over to Kevan.

Kevan Krysler executive
#4

Thanks, Charlie. We're happy with how Q3 turned out financially, coming in above the guidance we'd given for both revenue and operating profit. Revenue came to $831 million, up 9% from the same period a year ago, and we also posted robust operating profits of $167 million. Demand remains solid for our //E family and FlashArray//C solutions, which lets customers shift workloads that are sensitive to cost onto all-flash. On top of that, renewals of our Evergreen subscriptions throughout the installed base are still healthy, showing solid growth compared with a year earlier. When it comes to our storage-as-a-service products, total contract value, or TCV sales, hit $96 million in Q3, bringing the figure to $253 million across the nine months of FY 2025. Larger Evergreen//One deals, meaning those worth more than $5 million, are converting at a pace in line with the expectations we lowered last quarter. The faster-moving portion of the Evergreen//One pipeline — which we classify as under $5 million — is performing well, but during Q3 we saw a notable rise in those opportunities turning into a traditional sale instead. That drove product revenues in Q3 above what we'd anticipated, while TCV sales for Evergreen//One came in below expectations as a result. Annual recurring revenue for subscription services, or ARR, climbed 22% to $1.57 billion. Total RPO at the close of Q3, covering both subscription services and product orders, rose 16% versus a year ago, reaching $2.4 billion. When product orders are excluded, leaving only RPO tied to our subscription service offerings, the increase was 17%. Lower than expected TCV sales from our Evergreen//One offering are weighing on RPO growth. US revenue in Q3 reached $562 million, while International revenue came to $269 million. We added 340 new customers during Q3, and we still serve 62% of the Fortune 500; total gross margin in Q3 held up well at 71.9%, which reflects a strong Subscription services gross margin of 77.4% and a solid product gross margin of 67.4%. As we'd expected, product gross margin felt the effect of our deliberate push to move customers' cost-sensitive workloads onto our //E family and FlashArray//C solutions. While that strategy brought a slight dip in product gross margin, it highlights how committed we are to providing cost-effective, high-value solutions to our customers throughout our data storage platform. Turning to Q4, we anticipate continued solid growth for our //E family and FlashArray //C solutions, which should mean a modest sequential decrease in product gross margin, much like what we saw between Q2 and Q3. Operating profit of $167 million and an operating margin of 20.1% in Q3 got a boost from beating on revenue, strong gross margin results, and ongoing discipline on operating expenses. Our headcount rose sequentially by close to 140, landing at roughly 5,900 employees. Pure's balance sheet and liquidity remain solid, with $1.6 billion in cash and investments as of the end of Q3. Cash flow from operations for the quarter totaled $97 million, reflecting a sizable upfront payment for new software technology we licensed. Capital expenditures came to $62 million, including substantial investments to scale operations for our hyperscale opportunities. In Q3, we bought back 3.6 million shares, returning about $182 million to our shareholders. Additionally, we made $55 million in withholding tax payments tied to employee equity awards, which likewise countered dilution by roughly 1.1 million shares. Roughly $213 million is still available under our current repurchase authorizations. Now for guidance: we're lifting our FY 2025 revenue outlook to $3.15 billion, which works out to about 11 and one-half percent growth year-over-year. This upward revision to our FY25 revenue forecast stems from observing more Evergreen//One deals, those under $5 million, shifting into traditional sales. Although that trend lifts product revenue projections, it simultaneously lowers our FY 2025 TCV sales growth outlook for as-a-service offerings. Given weaker TCV sales growth in our as-a-service portfolio, we don't anticipate that growth in consumption and subscription offerings will meaningfully affect our FY 2025 revenue growth. Because of that, we aren't revising our FY 2025 TCV sales guidance for as-a-service offerings. For Q4, we expect revenue of $867 million, which reflects 9.7% growth year-over-year. Shifting now to operating profit projections. In line with our higher FY 2025 revenue guidance, we're also boosting our FY 2025 operating profit outlook to roughly $540 million, which implies an operating margin of 17%. For Q4, we're guiding operating profit of $135 million alongside an operating margin of 15.6%. Before I wrap up, I want to offer some preliminary thoughts about our first transformative design win with a top-four hyperscaler, to help shape your models. The commercial structure of this win centers on licensing our technology plus providing support services. No hardware will be part of what we sell to the hyperscaler. We expect meaningful revenue contribution and operating margin expansion from this win starting in FY 2027, which matches our projections for full-scale production deployments hitting double-digit exabyte capacities by then. To seize this milestone, we'll boost operating investments in FY 2026, targeting operational scaling and speeding up our chance to deliver Pure's differentiated technology for hyperscale storage. Taking these investments into account, we anticipate FY 2026 operating margin will be around 17%, in line with our FY 2025 operating margin guidance. To wrap up, looking forward, our strategic investments and innovation establish Pure as a leader in reshaping the data storage landscape; we stay committed to execution while steering through a persistently subdued IT spending climate. The growth prospects ahead are driven by our breakthroughs across our data storage platform, enabling organizations to achieve greater efficiency, scalability, and resilience in their operations. We're thrilled about what's coming and confident in our capacity to generate sustained growth and innovation. With that, I'll hand it back to Paul for Q&A.

Paul Ziots executive
#5

Thank you, Kevin. Prior to opening up the Q&A, I'd like to request that each of you stick to a single question with just 1 part, which lets us reach as many folks as we can. Should you have further questions, we'd appreciate it if you would kindly get back in the queue, and we will gladly address those extra questions, time permitting. Operator, please begin.

Amit Daryanani analyst
#6

Many thanks. Good afternoon to all of you, and congratulations on winning the Hyperscaler deal. I realize your team has been pursuing this for quite some time. Could you maybe explain what led this hyperscaler to pick Pure Storage rather than constructing it in-house, since the majority of these hyperscalers do possess a substantial amount of resources; and after that, regarding the financial effect, I heard you mention < > sizing it up somewhat, but at scale, how large do you believe this revenue opportunity or TAM might become for Pure? Thank you.

Charlie Giancarlo executive
#7

Thanks, Amit, and it's good to hear your voice. As you're aware, we've been pursuing this for a long while. A number of separate factors were in play here. In the end, what mattered was whether we could present a collection of best-in-class qualities that persuaded this hyperscaler our solution stood out above every alternative available to them. And frankly, this came down to how much we know this area, plus the fact that we already had something ready to hand over right away, so they wouldn't have to develop it themselves. Let me list them, with no particular ordering. The first is reliability at scale that genuinely has no equal. For one thing, we're able to reach those exabyte volumes; for another, on the reliability side today, we're driving toward an annual failure rate of 0.15%. That figure is remarkably small. Compare it with the failure rates for SSDs and hard disks, which run 5 times to 10 times greater. This brings down what we spend on labor overall, as well as what outages cost us. Next, the huge impact when it comes to saving power. When you look at it, we achieve somewhere between a five and 10 to 1 gain — meaning we use less electricity than setups built on hard disks. At that kind of ratio, you'd find our data centers maybe cutting 20% of the electricity they consume in total. If you stop and consider it, that's an enormous power source. Rather than thinking of it as saving power, think of it as a source of power. That electricity then becomes something they can use to scale out into other applications, like AI. Faster time to value — we're able to get this into their hands right away. They don't have to do any additional R&D work to get this up and running. It aligns really nicely with the architectures they already have, too. Keep in mind that our software is built on a processor-based architecture, not one based on SSDs, so it slots right into their virtualized approach where all their software runs on a processor. That makes it a far more software defined model. And last, you get unmatched price and performance — reaching the same TCO capabilities as hard disks once you account for everything, while delivering roughly five times the performance those hard disks give you. It's simply an equation that's really hard to beat. Kevin, I'll hand things over to you for the economic portion.

Kevan Krysler executive
#8

Absolutely. I appreciate that, Charlie, and thank you, Amit, for raising that question. There are a few points I'd like to bring up at a broad level regarding the financial picture, including how this chance might eventually appear once it reaches scale. As Charlie noted, roughly 60% to 70% of hard disks being purchased at present are essentially propelled by the hyperscaler segment. That, I believe, provides a solid reference point regarding the potential available to us over the long run. Clearly, for us that will translate into considerable revenue contributions as time goes on, along with growth in operating margin across the long term too. Over the nearer term, a few factors deserve mention, which I also put into my prepared remarks. Regarding operating margin for FY 2026, our expectation is that it stays level with the 17% guidance for this year that we provided at the start of the year. The reason is really to make funding available for expanding operations and speeding up the hyperscale chance sitting before us. We do not anticipate a meaningful revenue contribution from this chance before FY 2027, and that would likewise encompass operating margin growth; it's also worth noting that we will not be including hardware in what we sell to the hyperscaler; that's a key factor for us too. And finally, when we consider revenue contribution from this chance, and we view it relative to data storage capacity purchases, it would be considerably smaller than, for instance, a sale of our FlashBlade//E solution. So I believe those are a few key points we ought to be keeping in mind.

Paul Ziots executive
#9

Thanks, Amit. Let's take the next question.

Aaron Rakers analyst
#10

Yeah, apologies for that, everyone. I hope my audio is coming through now. I appreciate you taking the question, and congratulations on the announcement too. I suppose I just want to dig into a bit of what was just described. So, Kevan, here's what I'm asking: regarding the financial characteristics of this, it sounds like a licensing arrangement in which — are you granting a license for Purity based on how much capacity gets deployed, and also licensing the DFM module, meaning you aren't building those yourselves? Could you maybe break down how that works? And then, when we look at this, do you expect that this hyperscale customer, as they shift toward next-generational data center footprints, will choose to go 100% flash, or will it be some combination of hard drives and flashes? Any perspective on that would be useful. Thank you.

Charlie Giancarlo executive
#11

Yes, thank you; allow me to begin there. So your understanding is correct. What we're essentially doing is licensing both — the two components of the technology, meaning the software plus the hardware design. The customer will purchase directly from their integrator; they will be acquiring the hardware, so to speak, from their integrator. So you have that right; from our standpoint, it largely arrives as licensing fees, software fees and support fees. As for how — the revenue will roll out, as we've said, mostly in our FY 2027, with next year being this customer's continuing development of their next-generation data center design, and we're deeply involved in that. At present, what we needed to demonstrate we could do — proving ourselves in at the price performance level of the bulk of their storage, which has held true for all the hyperscalers, namely hard disk, but what we expect is that this will be a consistent architecture they'll use across the vast majority of their storage, including their higher performance storage, which today is generally provided by SSDs.

Paul Ziots executive
#12

Thanks, Aaron. Let's move on to the next question.

Howard Ma analyst
#13

Excellent. Appreciate you fielding this, and congrats everyone on that hyperscaler design win. In my view, it's a game changer for Pure and possibly a pivotal shift for the industry; so Charlie and Kevan, I'd like to dig a bit deeper into how big this deal is and how it's put together. So I suppose the first thing is, can you tell us how long the deal runs? And regarding the revenue potential, if I run a quick rough calculation, double-digit exabytes in calendar 2026, even if you simply assume — if you assume $0.10 per gigabyte and perhaps half of that gets recognized because it's software only, then you arrive at roughly $500 million in calendar 2026, which implies that 2025 is somewhere — it might be zero or near $500 million. So any detail you can offer, assuming those ranges hold for 2025, would be useful; and if I could squeeze in a follow-up about margins as well. Could you just say whether this deal will be gross margin accretive? And what precisely are the investments you're making on the OpEx lines? Thank you.

Kevan Krysler executive
#14

Yes. Thanks, Howard. There are many questions packed in there. Let me begin with the margin, gross margin impact. Look, I think the way we're — and once more, we're still at a fairly high level regarding how this looks, and we'll share more specifics down the road. Your thinking is correct in terms of the fact that most of the value from the deal will still be hardware. And I think that's crucial to factor into your models. I think when it comes to gross margin, what we've been tracking as company gross margin is a solid approach to have in mind right now. And once again, we'll provide updates as we proceed; and then duration, it's a multiyear duration, even though the volumes are TBD.

Howard Ma analyst
#15

Great.

Paul Ziots executive
#16

Okay. Thank you Howard. Next question please.

Pinjalim Bora analyst
#17

Thanks a lot for fielding this, and congratulations on the results this quarter. Charlie, here's what I'd like to understand from you: as you keep working through your strategy of increasing density with the DFM modules, in what ways does this opportunity shift or develop? Right now, is it built around the 75 terabytes or the 150? And once you get to 300 terabytes, what sort of volume growth would you anticipate in this specific situation?

Charles Giancarlo executive
#18

Yeah. The — it isn't tied specifically to any single one of those items, even though they are familiar with our road map, and that was a significant factor when they decided to go with us. They recognize that we typically hold the lead when it comes to density. For most hyperscalers—and definitely for this particular one—the majority of their storage tends to sit on the cheaper end when you look at cost per gigabyte. So for them, it matters a great deal to have a product that remains ahead of everyone else when it comes to cost. But beyond that, what really impresses them is that this same technology can serve workloads that demand more performance, and that's where the lower density DFMs come in. So the truly remarkable part is this: you're dealing with a single technology that spans virtually the entire price-performance spectrum. For a hyperscaler, having a single technology that meets every distinct requirement for running the operation is a major advantage.

Robert Lee executive
#19

Right, and let me just tack on one more point there, Pinjalim. Like Charlie said before during the prepared remarks, we have spent quite a while collaborating with hyperscalers to push this technology toward replacing disk drives. What I'd point to is that the launch of the 75 terabyte modules was the thing that truly made this specific customer sit up and take notice. Then going forward — as we consider the 150-terabyte generation that's shipping today, plus the road map we have in front of us — every stage on that journey simply renders the solution all the more persuasive, really on the strength of every superlative Charlie laid out before, costs included.

Paul Ziots executive
#20

Thanks, Pinjalim. Let's take the next question.

Analyst analyst
#21

Hey, everyone. Matt Litchi (ph) here, filling in for Mike Cikos at Needham. Appreciate you taking our question. I was hoping to get some clarity — when you talk about Evergreen subscription TCV and revenue, is that restating what you'd previously guided to? And can you give any sense of direction regarding how subscription revenue should move as a share of the total mix?

Kevan Krysler executive
#22

Yeah, good question, Matt. And yes, you've got that right. What we really observed was a change during this quarter, specifically involving our Evergreen//One higher velocity opportunities, where those deals ended up converting into a traditional sale — in other words, a CapEx sale. So the implication was, clearly, product revenue came in higher, and that is essentially what pushed up our guidance for the year.

Charles Giancarlo executive
#23

Absolutely, Matt — let me lay this out a bit further. What we're observing across the market, I'd say, is that customer OpEx budgets have come under mounting strain. That's because software costs went up — an unforeseen jump that hit software and SaaS spending during this year — and on top of that, there's uncertainty surrounding AI. As a result, a larger number of customers than we used to see, ones who previously had been looking at a Storage-as-a-Service option, which clearly is funded through an OpEx budget, are instead shifting toward buying the standard product, which gets funded through the CapEx budget, and that budget hasn't faced the same kind of strain this year. This market continues to develop. We're still in the early stages of it; our view is that this is a short-term occurrence; given the better economics and the better capabilities that a Storage-as-a-Service offering delivers, we'd anticipate seeing growth resume, though my sense is that what we've had here is a short-term occurrence stemming from the surprisingly heavy OpEx spending organizations took on over this past year.

Kevan Krysler executive
#24

Yeah, and I believe that point matters a great deal too, correct, that whenever we consider offering data storage technology in a service model, it remains at an early stage of maturity. And alongside that, people tend to fall back on whatever is simpler, which is, honestly, buying technology by way of a CapEx sale or a conventional sale.

Paul Ziots executive
#25

Thanks, Matt. Let's take the next question.

Simon Leopold analyst
#26

Thanks. I'd like to dig into something here. That CoreWeave announcement you put out a month ago — clearly, from what it sounds like, that isn't the massive hyperscale opportunity. So perhaps you can walk us through what kind of win that actually was, plus some detail on the use case and what that opportunity looks like. Thanks.

Robert Lee executive
#27

Sure, Simon. It's Rob here, and I'll handle that one. Without a doubt, we're thrilled to be broadening our relationship with CoreWeave — through the strategic investment we've made and through the commercial partnership we announced, which is going to deliver Pure Storage technology and make it accessible to every one of their customers within the CoreWeave-dedicated cloud. As we said in that release, for more than 12 months we've been collaborating with CoreWeave, the top GPU cloud provider, to stand up one of the biggest scale GPU storage deployments for a major joint customer. And now, on top of how well that environment has performed, this newest partnership expands things by taking those lessons and making that proven design available to a wider set of our joint customers. So we're genuinely excited to have it announced, and we're already seeing early signals of interest and demand coming from the customer base.

Charles Giancarlo executive
#28

Simon, to clear up any possible misunderstanding: roughly two quarters back, we put out word of a GPU cloud deal in the double digits — actually, I think that was Q4, which makes it three quarters ago. The customer was CoreWeave. Back then, we couldn't say who they were, but that's the agreement in question, and that's the infrastructure sitting there today.

Robert Lee executive
#29

Also, Simon, just to add one final point. Right, CoreWeave isn't one of those four biggest hyperscalers we've been talking about during this call.

Paul Ziots executive
#30

Thanks, Simon. Let's move on to the next question.

Sreekrishnan Sankarnarayanan analyst
#31

Hi there, and thanks for fielding this one. Charlie and Kevin, congratulations on what you've achieved with this big hyperscaler win. Charlie, you mentioned double-digit exabytes for calendar 2026. Does that figure rest on the agreement we already have? Or would you call it your own projection of how far this opportunity might grow? And regarding the license, does it get paid just once, or does it come back every year, and is it linked to exabytes in any way? Thank you.

Charles Giancarlo executive
#32

That figure reflects what we anticipate for calendar 2026, shaped by working alongside the customer, though nearly every sale is still to be determined. On the design win — we hold the plan of record, but to be clear, they are building around us; we are the plan of record for storage in their next-generation design. As for licensing — go ahead, Kevan.

Kevan Krysler executive
#33

Right, when it comes to direct flash technology and purity software, the licensing is genuinely going to be linked to how much capacity we end up shipping.

Paul Ziots executive
#34

Thanks, Krish. Let's take the next question.

Mehdi Hosseini analyst
#35

Sure. Charlie, so I can get a handle on what you meant about FY 2027 and that hyperscaler deal that's going to get you to double-digit exabyte shipments — could you help us figure out what the projected exabyte shipment number looks like for this fiscal year, fiscal year 2025?

Charles Giancarlo executive
#36

Nothing meaningful going out beyond test units in this year.

Paul Ziots executive
#37

Thanks, Mahdi. Let's take the next question.

Paramveer Singh analyst
#38

Yes, thanks for taking my question. I'd really like to dig into the Evergreen//One opportunity. You noted that certain faster-moving deals revert to on-prem direct sales. But given that more hyperscale chances are now appearing, on top of the one announced today, do you believe customers could be more inclined to go straight to a hyperscaler using your Purity platform and DCF rather than choosing an Evergreen//One style deal? Thank you.

Charles Giancarlo executive
#39

Yes, I'll handle that. It's a fascinating question, though in my view the answer is largely no. Here's why. What we'll be providing to this hyperscaler, and to other similar hyperscalers if we manage to win them, is the fundamental storage technology, not the storage services that the customer provides to its own customers. Those storage services rest on their own software, and thus are among the same software services they already offer today. What's the implication? For the most part, enterprises go to the cloud for reasons tied to deploying their applications, far less for reasons tied to storage choices. And so — in our view those choices are mostly independent of where they decide to place their data. So if they choose to keep applications on-prem, that's typically where we sell our standard product now. And when they choose to keep it on-prem, we offer them the option of storage as a service that runs both on-prem and in the cloud; so I truly view those as two quite distinct things, and I don't believe our provision of storage to hyperscalers built on our direct flash technology will significantly sway enterprise choices about where to put their storage. Conversely, I do believe we can deliver some of cloud storage's finest qualities to the enterprise through the other announcement today, namely that we've made Pure Fusion available for download this quarter. What that enables companies to do is construct their own enterprise data cloud made up of both their on-prem storage and storage they place on our software in the cloud. And this lets them function much like hyperscalers do, which is storage as a service, storage as a cloud of data instead of as individual arrays. So I view them as reinforcing each other. But I don't necessarily view our sales to the hyperscaler for its foundational storage as greatly influencing enterprise choices about where they put their applications.

Paul Ziots executive
#40

Thanks, Param. Let's take the next question.

Asiya Merchant analyst
#41

Thanks. Appreciate you fielding this, and congratulations on the hyperscaler news as well. Regarding evergreen TCV, it seems like a greater number of customers are shifting toward a product sale. You noted this was likely a short-term effect caused by constraints on OpEx budgets. When you consider the opportunity pipeline in front of you and look ahead to calendar 2025, could you help us grasp what you anticipate for this Evergreen offering and Evergreen services subscription services in terms of ramping back up to double-digit growth? Thank you.

Kevan Krysler executive
#42

Yeah. Thanks for that question. A few thoughts on this: look, when we examine the pattern around our Evergreen//One conversions above $5 million, nothing has really changed relative to what we described last quarter. That was in fact pretty much in line with what we expected, and again, we took some time to walk our investors through that last quarter. The place where things have moved somewhat for us is in the faster-moving Evergreen//One deals, meaning those at $5 million or below; that's where we observed a notable uptick, so to speak, in conversions toward a traditional sale. And again, we've discussed that — it doesn't appear to be lasting, particularly in light of our industry-leading Evergreen//One service offering. But when it comes to thinking about and guiding or discussing next year, I'd prefer to wait and see how Q4 unfolds, then provide all of you an update next quarter.

Paul Ziots executive
#43

Thanks, Asiya. Let's take the next question.

James Fish analyst
#44

Hey, everyone. I'll add my congratulations regarding the cloud win too. I wanted to inquire about the pipeline. First, concerning the hyperscaler cloud opportunity. Based on your remarks, you both keep hinting at the possibility of landing additional hyperscalers. So regarding similar licensing potential with other clouds, what are you observing in terms of the discussions you're having with, say, the leading 10 cloud and AI providers? And second, perhaps on the enterprise front, is calendar 2025 shaping up to be a bigger refresh year for storage? Or does that evergreen model somewhat dampen the impact? Thanks, everyone.

Charles Giancarlo executive
#45

Yes. In fact, I'm going to hand it over to Rob, our CTO, for the hyperscaler discussion. Rob?

Robert Lee executive
#46

Yes, definitely. So like we mentioned, talks are still moving forward with several hyperscalers. And the reason they're moving forward, I'd say, comes down to the things Charlie pointed out earlier: the reliability we provide that nobody else matches, the substantial energy savings, and of course the performance range we're able to deliver — all of which drive TCO reductions. Those, I believe, are the primary factors, more than the details around financial and commercial terms. But those conversations are ongoing with several other prospective customers. I do believe today's announcement — covering both our lead customer in this area and our strategic collaboration with Kioxia aimed at enabling wider adoption of our flash technology among hyperscalers — is going to help speed that up. So we're quite enthusiastic about what lies ahead. That said, like we noted, this design win represents a very, very significant milestone, one that places us in the plan of record with this hyperscaler, though obviously additional work remains. Correct? So over the coming year, we'll be collaborating very, very closely with this hyperscaler as they keep advancing through the usual stages toward production, as they ramp up to what we ultimately anticipate will be double-digit exabyte capacities in fiscal 2027.

Charles Giancarlo executive
#47

When it comes to what we expect on the enterprise side, I think what you're really asking is whether we think demand has been held back owing to a muted IT market. And I do lean toward the view that a certain amount of pent-up demand has been accumulating. Still, we'll need to find out how that plays out once we move into next year.

Paul Ziots executive
#48

Thanks, Fish. Let's take the next question.

Wamsi Mohan analyst
#49

Sure, thank you. Turning back to the hyperscaler chance, Kevan, what you said about how much revenue comes in compared with what's spent on data storage capacity — that would be a lot lower. Is the reason simply that hardware is missing from the mix? Or ought we to assume that what the software costs on a per exabyte basis is likewise a bit lower? And to be clear, when you say double-digit exabytes in calendar 2026, does that come from this single hyperscaler? Or does it include other deals won later on? Thank you.

Kevan Krysler executive
#50

Yes. Thanks, Wamsi. The double-digit exabytes in FY 2027 relates to this hyperscaler design win itself and our expectations associated with that win. And what was your second — your first part of that question?

Wamsi Mohan analyst
#51

Is the reason for that lower revenue contribution relative to a capacity — storage capacity sale simply that no hardware is included in the solution? Or does it come down to software pricing being somewhat different from how you'd traditionally sell it?

Kevan Krysler executive
#52

Yes. I appreciate you asking for that clarification. The main thing driving this is going to be the exclusion of hardware from what we sell, and that's really the primary factor; and then, if you consider scale in terms of order of magnitude, pricing does come into play, though it matters less than the hardware exclusion.

Paul Ziots executive
#53

Thanks, Wamsi. Let's take the next question.

Meta Marshall analyst
#54

Great, thanks. I wanted to ask — you said the volume is TBD, but you seem to have some idea of how the architecture will be shaped. Is the TBD because you're unsure how quickly they'll deploy this new architecture, while you do know roughly what share of that data center will be yours? Or what pieces are still moving as you get a clearer picture of the actual volumes for fiscal 2027? Thanks.

Charles Giancarlo executive
#55

Yes, thank you, Meta. What's mostly uncertain is the precise timeline for the data center build-outs. I'd also say the exact percentage of the data center that's ours is somewhat unclear too, even though we're aware that we've qualified for a very, very big chunk of it, and we're aiming for even more. That said, I'd say it's mostly a matter of schedule, which is the thing we have the least control over. They still have plenty of design work, so to speak, on their side for other components, ours included, but also for many other parts of their next-generation design.

Paul Ziots executive
#56

Thanks, Meta. Let's take the next question.

Eric Martinuzzi analyst
#57

Yeah, I'd like to ask a bit more about what's shifted in the Kioxia arrangement, since you mentioned the hardware is now left out. My assumption would be that the hyperscaler is going to deal with Kioxia straight away in that case. What has actually changed regarding that arrangement? The reason I ask is that when you mention additional production capability and more advanced technology, it sounds as though your company is getting squeezed out of the middle.

Charles Giancarlo executive
#58

Yeah. I appreciate you raising that. No, we are not getting squeezed out of the middle. Our connection remains with every one of the NAND suppliers. All of the qualification work falls to us. The design work falls to us. On top of that, we'll handle designing the direct //flash modules as well; therefore we hold a great deal of the control, and when a hyperscaler brings in an integrator, that integrator is the one buying the hardware modules and handing them over to the hyperscaler. The cleanest way to picture it is something like a meat in the channel model.

Robert Lee executive
#59

What I'd add is that this represents a broadening of a partnership we've had for a long time—one that isn't limited to commerce and capacity alone; our work alongside Kioxia on technology road maps and technology design has been excellent, and that's a crucial element of this collaboration too, namely delivering, if you like, jointly designed—as Charlie put it, codesigned—direct flash technology featuring their most advanced NAND components to this hyperscaler customer, and doing so together.

Paul Ziots executive
#60

Thanks, Eric. Let's take the next question.

Tim Long analyst
#61

Appreciate it. I'd like to focus on the gross margin side of things. My understanding is that the guidance was for a sequential drop in Q4 comparable to what we saw in Q3. Could you confirm that? Then, as we look ahead on this line, it appears the mix may be gradually tilting toward //E and FlashArray//C Series. Should we anticipate that trend persisting on the product and hardware side? Also, if you could share some thoughts on competition, given that a larger portion of the competitive set now offers QLC-based products in their lineup — is that segment of the market becoming more contested? Thanks.

Charles Giancarlo executive
#62

Tim, out of respect for those who have already returned to the queue with their second question, we may attempt to address one or two of yours, rather than all three.

Kevan Krysler executive
#63

Tim, perhaps I could take on the product gross margin and that quarter-over-quarter drop comparable to the one from Q2 into Q3, plus what we anticipate as we move into Q4, and that's genuinely coming from our approach of helping our customers shift their cost-sensitive workloads over to our //E family and FlashArray//C offerings. And it truly comes down to the robust expansion we're witnessing with those offerings; at this point, there's diminished price elasticity for them because of how price sensitive those workloads are. So when you combine the greater growth of those offerings with the lower price elasticity, that's creating strain on the product gross margins that you're observing.

Paul Ziots executive
#64

Thanks, Tim. Let's take the next question.

David Vogt analyst
#65

Great. Appreciate you fitting me in, everyone. So perhaps one for you, Kevin, regarding margins in the coming year. You mentioned putting money into the business to back the rollout during calendar year 2026. You also said gross margins would be roughly similar in fiscal 2026 compared with fiscal 2025. Could you help us get a sense of where those investment dollars are headed, since you've been investing heavily over the past couple of years? Should growth in OpEx be at that same level? Is that the way to think about it? Looking at 2024 and into 2025, should 2026 appear comparable? Thank you.

Kevan Krysler executive
#66

Yes. Excellent question. I'll take more time next quarter to walk through more specifics on what we expect. But what I did share is that we believe our operating margin of 17%, which matches what we told you for this year, will also hold for next year. And regarding the primary drivers behind these investments, that includes concentrating on speeding up our density road map with our flat DirectFlash technology and growing our supply chain capabilities, qualifying additional NAND suppliers and manufacturing sites and really integrating our technology with the hyperscaler hardware specifications. That's where we expect the meaningful incremental investments to originate. And then looking at it for this year, clearly, we're also investing substantially from a CapEx standpoint. That's the reason for the full year of FY 2025, we expect our free cash flow margin to be slightly below, 1 to 2 points below, our operating margin.

Paul Ziots executive
#67

Appreciate that, David. For our final question, we'll go to Medi, who rejoined the queue to ask something a second time. Thanks so much, Mehdi, for lining up again. That wraps up the Q&A.

Mehdi Hosseini analyst
#68

Yeah. Charlie, how many exabytes do you figure will ship in FY 2025?

Charles Giancarlo executive
#69

Predicting that isn't so easy. The number will be quite small. It's tough to pin down, since, like I mentioned, the places we're entering are largely test setups ahead of preproduction setups. So perhaps 1, perhaps 2 — roughly in that neighborhood?

Paul Ziots executive
#70

Thank you, Mehdi. Before we conclude, Charlie, I think you had some final comments.

Charles Giancarlo executive
#71

Absolutely. I'd like to express gratitude to each of you for being with us on this earnings call today. That design win highlights, in my view, the essential part Pure will play when it comes to tackling the massive data expansion and electricity needs of hyperscalers; they are up against extremely intense rivalry for power and are even resorting to quite novel new energy sources; however, bringing those energy sources online is going to require a considerable amount of time. Since power limitations remain an enormous problem for them, this stands out as one of the finest options available to them for unlocking tremendous quantities of power — roughly 20% of the entire amount of power they consume at present — and making it usable for all the expansion they have ahead of them, all while enhancing the general performance and abilities of the storage they already have. Once more, I'd like to express gratitude to our customers, our employees, our partners, our investors and our suppliers. We sincerely value all of their ongoing backing and dedication. We're eager to talk with you again next year. Happy holidays, everyone. End of Q&A

Operator operator
#72

That wraps up Pure Storage's Third Quarter Fiscal 2025 Financial Results Conference Call. We appreciate everyone for joining in. At this point, you're free to hang up your line.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete Everpure, Inc. transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.

Get an API key View API docs →

For developers and AI pipelines

Programmatic access to Everpure, Inc. earnings transcripts and 132,000+ others is available through the EarningsAPI REST API and the hosted MCP server. Quarterly plans from $162 - full transcripts, speaker segments, full-text search, and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.