I think the second part of your question was to better understand the product mix as part of the sale. I will remind you, this was a large Evergreen/One deal, which is I think speaks to the dynamic nature of a lot of the work in this space. as a lot of the AI workflows expand beyond just high-speed training into the broader set of data preparation, all the elements that I just walked you through Evergreen/One creates a lot of flexibility in that model for the service provider, the cloud provider, in this case, to deploy the right service levels for the right usages. That said, [indiscernible] environment is being served with the FlashBlade product set. So hopefully, that gives you a little bit more color to think about our opportunity and really the points of validation in the space.
Operator: Our next question comes from Jason Ader from William Blair.
Jason Ader: Just wanted to ask if and when you might be willing to give us revenues from Evergreen/One and Flex.
Kevan Krysler: Jason, great question. Let us work through the transparency we’re giving you on TCV sales for the time being and the conversion of those TCV sales in terms of normalized revenue, and you’ll see that in the slides we presented, I think that is a level of bridging we’ll want to do at this point in time in terms of where we’re sitting.
Operator: Our next question comes from David Vogt from UBS.
David Vogt: And Kevan, I’m going to come back to that prior question. If I — I’m just trying to think through all of the different disclosures that you gave for ’24 and ’25, and trying to normalize for the traditional model historically, product and/or services. I mean just based on my math, it looks like your revenue over the last couple of years would have compounded at roughly 9%. Is that a reasonable framework when I make all the adjustments for the model transition and the telco customers push out from ’24 and ’25. And if that’s the case, what’s the core underlying sort of storage demand that you think underpins that over the last couple of years? And how did your share during those periods?
Kevan Krysler: Charlie, do you want to hit storage demand and market share first and then I’ll hit the modeling question.
Charles Giancarlo: Yes. I think the core, David, of your question really relates to are we picking up market share at what rate? We feel pretty confident that we’re in the still — we remain in the 15% market share pickup rate for a variety, which is not necessarily reported by the reporting agencies mainly because they do not incorporate our Evergreen/Forever subscription, which, as you know, means that we don’t resell the same storage when an array becomes obsolete because in our case, arrays don’t become obsolete. So with that, I can’t — I think the team is taking a good look at the — we know what it would have been — and Kevan stated it in his preamble that it would have been around 7 — a little over 7% this year. I don’t have the numbers for the prior year.
Do you have that, Kevan? No. We love to get back to you on that. But your basic premise that the market is under accounting, what we believe is — would be our growth if our Evergreen/One sales were actually in standard product sales is absolutely correct.
Kevan Krysler: The other thing too, David, is you can kind of do back of the envelope calculation to estimate the amount of revenue that’s coming off Evergreen/One and Evergreen/Flex given that average duration of these contracts is around 3 years, and obviously, we’ve given you the TCV sales for FY ’24, growing in excess of 100% as well as our expectations for TCV sales for Evergreen/One and Evergreen Flex for FY ’25.
Operator: Comes from Tom Blakey from KeyBanc Capital Markets.
Thomas Blakey: Across your products, Charlie, could you just maybe talk about the FER and straight-up product sales kind of like exit rates in fiscal 4Q and what you’re kind of seeing in fiscal 1Q to maybe even and get some insight into the level of possible conservatism in your guide there on the product side.
Charles Giancarlo: As we indicated, I think, at the end of our — in our Q3 call and now in Q4, we are seeing strong indications of improving demand both, I think, in the actual results as well as in pipeline as we go forward and in our conversations with customers overall. So that’s what’s giving us confidence. I think, Kevan, in terms of — we generally don’t provide direct exit rates, but…
Kevan Krysler: Yes, we don’t. But I think we’ve taught it directionally. And obviously, the strength we’re seeing, both in FY ’24 and FY ’25 is really being driven from a growth perspective. Buyer subscription services and subscription services revenue. And when we think about it directionally next year, but to Charlie’s point, he’s right, we don’t guide specifically on product revenue and subscription revenue. I think it’s a good way to think about product revenue being flattish to maybe slightly down and really that growth being driven by our subscription services overall for FY ’25.
Operator: Our next question comes from Simon Leopold from Raymond James.
Simon Leopold: Earlier in the Q&A, Charlie broke down the AI opportunities into sort of the 3 buckets, and I found that helpful. And what I really wanted to see if we could unpack a bit is the item described as data uplift. In other words, I’m trying to get a better sense of what you’re assuming for really the timing of when that becomes material and how you expect it to manifest itself in terms of — is it sales to enterprises as opposed to needing to break into hyperscalers or the operators of the AI platform. Just a little bit more impacting if that would help.
Charles Giancarlo: Absolutely, Simon. So I think your instinct on this is absolutely correct. What the concept is that, of course, regardless of the manner in which enterprises want to build, whether they want to build their own data environment for AI or do their own modeling or do their own inference or not. They do have to make their data available for analysis by the AI engine. In order to do that, their data today is largely trapped, I say, trapped in silos. It’s on storage environments that were largely purchased very much — storage is bought with economics in mind. And they are purchased at the performance level and at the capacity level necessary for their primary job, which generally means there’s not a lot of performance left over to be able to serve up data for analytics of any type, let alone AI.
And even if they had that performance, they’re not networked. Traditional storage arrays are not networked at the array level. You have to go through the application environment. which is not terribly useful and very indirect. That would mean that customers would generally have to copy that data and buy new arrays anyway. And in our case, what our platform allows them to do is replace those arrays for their primary purpose going from disk to flash and have plenty of performance left over for the AI environment. So they get to modernize their environment reduce their power spacing cooling by a factor of 10, have much higher reliability, lower labor associated with it. And at the same time now update their data environment so that it is available for AI analysis.
Robert Lee: And just to build on that a little bit. Charlie mentioned a lot of these silos and fragmented pools of data storage really aren’t network today. Let me take it from a customer and a procurement lens. If you look at a lot of these environments, data storage has historically been purchased and configured application by application, department by department completely independently. And in a world where that data is really only being used for a single purpose, that worked okay. But the whole power of AI technology is being able to connect all these data sets and glean from them in unison greater insights in order to do that, you’ve got to actually connect all of these things. And so when we step back and we think about our position with our platform strategy, being able to pull all these pools of data together, that’s what we really see as the larger opportunity set here.
Charles Giancarlo: And then when you think about timing, what I’m seeing is it’s the most — it’s the companies that are the most advanced in their thinking and analysis of what they might do with that are just beginning to realize, oh my gosh, regardless of how many GPUs we buy or lease or rent, at our data governance is very poor. Our ability to get access to that data is very poor. So I think it’s going to take time for the general market to come to full grips with this. So I see it more late this year, early next year. In the meantime, we have — there’s plenty of economic reasons why customers would want to use our e family to replace disc when that’s coming up anyway. So as I see it, the e-family kills two birds with one stone. It’s a better replacement and it prepares them for AI.
Operator: Our next question comes from Nehal Chokshi from Northland Cap Markets.
Nehal Chokshi: Yes. Thank you and congrats on your results, and thank you for that detail on what the Evergreen run bookings does to your overall revenue. That’s great. I want to double click on the major one with the — on the major GPU cloud provider. Specifically, I’d like to understand how penetrated are you within that opportunity as well as if you’re not — if that deal does not put you up 100% penetration than what is being utilized alternatively at that particular GPU provider?
Robert Lee: Yes, Nehal, I’ll take that one. This is Rob. We’re — it’s not 100%, so I’ll tell you that. Look, we’re very pleased with the win. I think, as I mentioned before, what’s particularly exciting is the mix of use cases being deployed on our footprint there. And look, I think as we mentioned, it’s one of the largest GPU cloud providers out there. Our focus at this point really is in ramping that Evergreen One consumption driving those initial project deployments to success, and we’ll see where that takes us.
Nehal Chokshi: I was just wondering can Rob describe what is alternately being utilized in before [indiscernible]?
Robert Lee: I really can’t speak to that. I don’t know that we have the visibility into that.
Operator: Our next question comes from Eric Martinuzzi from Lake Street.
Eric Martinuzzi: Yes. So you had a small workforce realignment charge in Q4. I was curious to know what the goal of the realignment was. And are we done with that?
Charles Giancarlo: Yes. We have restructured to some extent, inside the organization to focus much more on customer segments, enterprise commercial and hyperscaler and putting specialized teams focused on driving business process to fit those 3 models in a very coordinated way across across the company. And based on that, we’ve moved people and moved roles to different areas. And that also meant that there were some roles that were being eliminated, and that’s what really caused the restructuring. And no, we don’t — I mean, that was limited really to the reorganization that we put in place.
Paul Ziots: Thank you, Eric. I think this next question is a person who got back in line. So I think this will be the last question.
Operator: Next question is Aaron Rakers from Wells Fargo.
Aaron Rakers: All right. Just because it has not been asked, I’m curious, given the AI discussion and the narrative around the pure, just where we’re at with regard to Meta and that deployment, AIRE, I know there’s been a lot of discussion about the expansion of Meta’s GPU footprint. I’m just curious of where you currently see yourself at and whether or not there’s any assumptions of opportunity at Meta baked into your guide this year.
Robert Lee: Thanks, Aaron. This is Rob. Welcome back for round 2. Our relationship with Meta is stronger than ever. We’re working with them on almost a continuous basis. And as we said before, they continue to realize incredible value from our solutions in place. I think specific to your question, just asking about did we see additional sales to the RSC environment. We — as we’ve said before, we have sales to Meta in almost every quarter, including in other AI environments. We didn’t have sales into RSC. But as we’ve mentioned, we typically wouldn’t be updating or commenting on sales into other environments, including other AI deployments there.
Paul Ziots: Thank you, Aaron. Charlie, before you give concluding remarks, maybe you could comment on Meta’s second really excellent question about whether we’re at a different crossover point now to the all-flash data center.
Charles Giancarlo: Yes, Meta, I think that it’s a great question. I still believe we’re early in that process. I started saying last year that I expected the last disk storage array to be sold in about 5 years’ time. We’re now 4 years in front of that. I’m going to stick with that timetable. I think customers — the growth of our e-family has been really tremendous this year, but it’s early — still early days. We’re expecting even greater growth, obviously, this year ahead. But considering how much disk is out there, it will take a bit of time. I am very bullish on this, and I’ll stick with it, that I think the next 3 to 4 years, we’re going to see the decline of Disc Systems.
Paul Ziots: And then you had some concluding statement.
Charles Giancarlo: Well, I do want to thank everyone for joining us today, as always, on today’s earnings call. The platform strategy that we’ve built is now leading the data storage industry’s transformation. We are seeing just incredible response to really unifying the way that data operates within the enterprise environment. With this unified and modern storage platform, enterprises now can really uniquely tackle their fragmented data environments. And that’s what’s going to allow them to unleash the full potential of their artificial intelligence. I do want to thank our customers, our employees, our partners investors and suppliers, your dedication, your collaboration, your trust are really the driving forces behind our progress. Thank you all.
Operator: That concludes the Pure Storage Fiscal Fourth Quarter and Full Year 2021 Earnings Call. Thank you for your participation. You may now disconnect your lines.
