All right. Hello, everyone. Can you guys hear me okay? Excellent. Thank you so much for coming, especially for those of you who hopped on a plane. Thank you so much for those of you on the webcast. We've got a pretty good schedule lined up for you today with a lot of the product announcements and some product heads, so it should be a pretty good day. Just a real quick note, this is our typical safe harbor statement. No different than the earnings one, but just to pay attention to. Logistically, if anybody needs plugs, they're up here in the front. We'll have a break halfway through. If you need anything else, I'll be in the back, so you can just come ask me for whatever you need. Okay? With that, we're just going to jump right in to Charlie.
Good afternoon. Thank you for joining us. Thank you. Been getting a little hoarse lately, so I have a bottle of water just in case. I'm really pleased to be speaking to you today. Before I start, because I'm going to reprise some of the conversation that we had this morning in our keynote session. How many of you were at the keynote, if I might ask? Okay. About a half to two-thirds. Terrific. We also have people joining us on webcast who probably were not seeing it. I'm going to reprise some of that, and perhaps give you a little bit different view of some of the things I did this morning, and then we'll talk a little bit more about where we are as a company, going forward from a financial perspective. Oops, there we are. Let's go to the next one. Right.
As we spoke about this morning, we're celebrating several things at this Pure//Accelerate. One is, it's going to be our 10th birthday in just a few days. Many of you know Coz. He'll remind me it's exactly October 1st, that's our 10th anniversary. It gives us both a chance to look back and a chance to look forward. Part of this is that, having been through the introduction of a lot of new technologies during my career, I know what a difference a decade can make. A decade can make a great difference. One of the things about decades is, as any startup will tell you, and certainly anyone who's introduced new technology will tell you, it takes 10 years to be an overnight success.
Every new technology that goes into place takes time to really get its traction, certainly with customers, to become mature enough to where it really hits, as we tend to say, the knee in the curve and starts taking off. In our first decade, we've gone from a raw startup, now obviously to not only a public company and, I'll remind you, a profitable public company, but one of the fastest-growing, at least B2B companies, in infrastructure ever. We introduced a lot of new things and a lot of firsts. In particular, I think one of the most important is this idea of non-disruptive upgrades. Where when we sell a product to one of our customers, that product effectively never gets old. It's always new.
The reason why it's always new is not only do we update it periodically with new software, that's been very common for many years, in every industry, but we upgrade it in terms of hardware as well, so that the hardware is always new. We do this without the customer ever suffering downtime. That's new. That is different in the industry. You find very few products or capabilities that can be consistently new year after year. While the first sale may be a perpetual sale, from that point on, it's always subscription. Always subscription. We were also one of the first, at least in our industry, to do cloud management or AI-driven operations, where we can predict issues before they actually are seen by our customers and make changes or work with our customers to make changes so that they don't have issues in their environment.
We now do cloud data protection, did that in our first 10 years, and have other things that our customers value a lot, such as always-on encryption, increasing their overall security as a company. As our customers look forward, they're increasingly challenged by all these words that you hear all the time. Whether it's agility or orchestration or migrating more to a DevOps-type environment, making the digital transition, and so on and so forth. As we look back over the last couple of decades, we see that a large part of their infrastructure had made major changes that really increased the agility, the productivity of IT organizations to be able to do more with less. With less people, with less time and effort, and allowing their organizations to be able to do more in terms of software and services, for their customers.
Examples of that are the migration from custom bespoke networks. When I got started in my business, every company had a dozen different networks with dozens of different protocols, all deeply tied to the application. Now those are all handled by one network, and it's called IP and the Internet, all based on, for the most part, Ethernet technology. They also had application stacks that were hardwired, with specific hardware, specific operating systems at specific rev levels with very specific security software, very specific drivers. God forbid they ever wanted to move that application to, they wanted to expand it, they'd have to have exactly the same hardware and exactly the same software in order to be able to do it. Of course, over the last decade, that's all changed with virtualization. What have we learned from those two things?
What we've learned is that taking an industry or a technology from being very vertical and deeply tied into an application, to horizontal and common across application environments is a very valuable thing and drives the industry forward. Where that has not occurred is in storage. Storage is still deeply tied physically to the applications that it serves. You can move an application around, and you can copy and move data, but with data getting as big as it is, really, you don't want to be moving it very much. We say data has gravity. What does that mean? It's just getting so big, petabytes of data, not easy to move. Besides which, why have to continue to copy it? Why have two, three, four? The average company has five to seven copies of their data. Why?
They need it because for different applications, analytics versus production, dev versus production, backup, et cetera. For different applications, right now they have a storage array that's tuned to that application, they have to copy and replicate that data to the other storage environment because that's the one that's tuned to that application environment. Data is copied constantly, and storage arrays are bought based on being properly tuned to the specific application environment. That's why you see with many of our competitors, a wide range of different arrays at different price performance, all catering to a particular set of application types. We think this is old-fashioned. We think this is no different than the way that applications or networks were treated decades ago, and that the world needs to migrate to a place where data and data storage is dynamic, not static.
We think it needs to move to something more like this, where data is a set of resources, data storage is a set of resources that's called by API from the application, from the operating system, or from the operators, from a management perspective, depending on what the needs of the applications are, and provided dynamically. Dynamically in terms of protocol, dynamically in terms of what we call storage classes, that is to say, different price performance levels, and driven by something new. Which cloud do you want it to be in? Where do you want it to be? Private data cloud, public data cloud, or somewhere else. It's a very different way at looking at the way that data and data storage is handled. Very dynamic, virtualized, made into a horizontal capability rather than vertically tied directly to the workload or to the application.
This is what we are now calling a modern data experience. We think a modern data experience has several attributes. Among them, it needs to be simple, needs to be seamless, and sustainable. I'll just describe what we mean by each of these words. Simple, to some extent, it goes without saying, as your environment gets more complicated, if we just leave the same complexity in the technology as companies scale it, the complexity is just scaling with the environment, which of course, nobody wants to go to. As you scale your environment, you don't want it to get harder. If anything, you want it to get easier, right? Simplicity for us means it needs to be API-defined. Companies need to be able to automate this based on their internal practices so that it's completely driven by their automation system, by software.
Secondly, it needs to be able to be consumed as a service. That means both how we provide the product to our customers, so that if they prefer to consume it as a service, that is on a monthly basis or weekly basis or hourly basis, that they can do that. It also means the way that the applications or, let's say, our customers' internal customers consume the storage. The application wants to be able to consume it as a service, not demand a fixed amount of it, but just get it as it needs it. Needs to have AI-driven automation because the level of complexity is going way up.
If the infrastructure is going to manage now more and more of the data, then it needs rules, and it needs to learn how the customer is changing their use of the data to be able to manage it more effectively. Nowadays, we can do that more and more with AI. Finally, it needs to be an effortless experience. One of the reasons why Everpure has the NPS, the net promoter score that we do, which I'll just remind you all because it's my favorite number, 86.6. It's the most important number in the company. The highest in any industry. Highest 1%. One of the reasons why we have that is we make what is typically a very difficult experience for our customers, that is managing their data and managing their storage, and we take it off their minds. We give them their nights and weekends back.
That goes to this effortless experience. I like to remind people that making things simple is very hard. It's very easy to make things hard. It's hard to make things simple, and we put a lot of work and effort in that. Secondly, we believe it needs to be seamless. Let me just go back. Okay. Seamless means that we need to have it be capable of producing or providing any service that the application demands. It needs to be multi-protocol, right? If an application needs a file system, we need to provide NFS. If it needs a block system, we need to be able to provide volumes. If it needs object, we need to be able to provide object, all by API, not by humans moving a different storage system into place.
Just by an API call. We need it to have multiple storage classes. What does that mean? Well, not every application needs the highest performance environment. Higher performance is going to cost more. That's just the way the world operates. Some applications require lower performance, and they might want lower cost. Again, that should be by API, not by which array you attest and buy. Right? It needs to be multi-cloud, because that's a fact of life right now. Customers should be able to put the same workload on-prem or in the cloud without having to change the design of their application. Today, it's a completely different environment. Cloud and on-prem, completely different environment. You have to what they call refactor an application, and it's a rewrite of the application to utilize the different storage environment that exists in the cloud. By the way, it's different in every cloud.
What we've done with Cloud Block Store is make it look the same on Amazon as it looks on-prem. As we develop it on Azure, it'll look the same on Azure as it does everywhere else. It just makes the practice or one's ability to move an application that much easier among these different environments. Again, this all needs to be done by API, not by humans, so that it can go into a completely automated and orchestrated environment. Oh, I'm sorry. It looks more like this. That is a modern data experience where the type of data and data storage that you need is provided automatically by the APIs you write on the left and delivered to our customers' customers, their internal customers, whether it's DevOps, production, backup, analytics, again, by the calls that their applications make to this infrastructure.
Finally, sustainable. What do we mean by sustainable? Well, sorry, go back again. A sustainable architecture has on-demand consumption. That is to say that only buy what you need, when you need it, delivered effectively on demand. Non-disruptive upgrades, which I mentioned at the beginning, is absolutely critical in making a product sustainable. A product that is obsolete in five years that you need to replace, and by replacement, that means that you have to take the system down. It means you've got hours, if not days, of when your customers aren't ordering or your developers aren't developing or your traders are not trading. That's not sustainable. When you go to your favorite SaaS application, whatever it is, you don't know from day to day whether they've upgraded the hardware or software. You just know that somehow today's experience is better than yesterday's.
That's what we deliver with our non-disruptive upgrades. The product that a customer buys today will be brand new 10 years from now. That's a really important sustainable characteristic. Finally, continuously improving. It's not only new 10 years from now, but it has our latest software and it has the latest hardware associated with it. Of course, if you buy it as a service, that all happens transparently behind the scenes. I might extend sustainability. Sustainability also means a lower footprint on the environment, lower power, smaller space, and less waste at the end of the day. Because we recycle any storage that comes back to us. It either goes into our service catalog, or we dispose of it sustainably when it comes back to us. A sustainable environment. That's what we mean by a modern data experience, simple, seamless, and sustainable.
Okay, now, how is this working for us? This is our vision for the next decade. How are we doing right now? Well, I think many of you have followed us over the quarters. Oh, actually, let me start with this first. As I said, this show is about both celebrating our first decade as well as now identifying for our customers, our partners, as well as all of you, what our strategy is for the next decade. This journey actually is something that we started some time ago by creating a very sustainable focus on really redefining what we would call tier 1 storage. That is, we were the first ones to successfully bring what was inexpensive consumer flash to an enterprise storage market at the quality levels that was expected in that market.
It also meant that we were bringing 10x better performance in a number of different ways, 10x in terms of the actual performance of the data storage, but a 10x easier experience for our customers in terms of how they experienced the quality of the product itself, meaning that it took 10x less time for them to work. Literally, when I say 10x, many customers that went from disk to Everpure required 10x fewer people to manage that same amount of data. Okay? Fewer people to manage the same amount of data. It was important. Where we are today, I would say that over the last 10 years, we redefined what it was like to own and operate a storage array for your primary workloads.
Where we're going is we're going to redefine this new journey of a modern data experience, redefines what a storage system looks like in an enterprise. In fact, today, there is no storage system. There are only storage arrays serving individual workloads or workload types. This is a vision for how storage will become a system within an enterprise that is completely interconnected and driven by APIs. We've introduced today a number of new products that are GA actually today that are part of this journey. Among them, we announced FlashArray//C, which is our capacity-optimized, so this is the multiple storage classes now, as well as storage class memory.
Now what would be called tier 0, tier 1, tier 2 workloads are all open for Flash. Importantly, we've opened up this, what is typically considered to be secondary tier of storage, which is today mostly hybrid and disk, now can go Flash. We are at that level of price performance. We introduced the GA of our Cloud Block Store, which is this multi-cloud capability so that customers can develop their applications once and operate it both on-prem and in the cloud. Finally, we have introduced Pure as-a-Service, where all of our products can either be purchased as a capital purchase or basically operated on an as-a-service basis based on the capacity that the customer actually uses, and not what's reserved.
Delivered over the next several years, but of course, as I say, every overnight success takes a decade, all applications, all protocols, all clouds to be able to be managed as a pool of resources based on policy, API, and AI management in the background. This is what we've set out for ourselves to deliver to our customers as a modern data experience. One of the most magical items, and it is almost magic, for our customers, is that everything they've ever bought from us, everything they will buy from us today, and everything that they buy from us in the future, will all continue to contribute to this modern data experience without replacement. Just based on our subscription model, it'll constantly be updated to get to this modern data experience.
They have no fear that anything we've ever sold them or will sell them today becomes obsolete on this path. That's really quite remarkable, I think you'd all have to agree. Again, now modern data experience. Now let's just talk a little bit about where we are as a company and a little bit about what we see. This last quarter has been truly, I would say, astounding and surprising to many of us. It was, as you all know, for us from an economic standpoint, a good quarter, right on target at 28% year-over-year growth. Of course, we were rather surprised by the competitive announcements that largely came out afterwards.
In terms of the financial announcements with the five competitors that we tracked that announced publicly their results, we had been saying we were growing 10x faster, but when the others are negative, you can't really have a mathematical comparison. We've been asked quite often why we think this is the case, and we really assign it to two different things. One is we are one of the very few large storage competitors that are investing in this technology as if it is a technology market, not a market that is being reaped for profits, not a market that's going to white box and open source software, but something that demands focus and attention. The second reason is honestly the quality and the nature of our sales and marketing operation.
By definition, being a new company, 10 years old, every dollar of revenue we've ever gotten was a competitive deal that we had to take out of someone, some install base. So we have a team of, you know in the business as hunters, whereas the macro will affect much more farmers than it will hunters, almost by definition. So between those two things, I think it really explains our continued growth. With the roadmap that we've put out for ourselves, and I have to say, this roadmap gives us a very rich set of organic investment opportunities. With this roadmap, I have to say that I'm very optimistic about the future.
If we look to that, and if we keep close to anything like our current rate of growth, we'll continue marching up this ladder in terms of where we are in terms of overall storage. This is the market share for overall storage. A recent analyst report put us at number 2 for all-flash arrays. We're already pretty far along, but as we've told you many times, we're not measuring ourselves against all-flash arrays. We're measuring ourselves against all storage. There's a good reason for that. Let's just not forget that hard disk is a mechanical device. How many of you have recently seen a mechanical calculator on anybody's desk? Anyone? Okay. Solid state will win over mechanical. I don't think that's a controversial statement. It's just a matter of price.
As flash continues to decline in price, as it will, it will take over for all mechanical. Mechanical is still 85% of the market by volume. It's less by dollar, but more by volume. It's still 85% of the market. The opportunity for us as flash prices decrease is really quite tremendous. All right. A lot of you have told me that Pure, you've built a beautiful house, but you're in a lousy neighborhood. Come on, right? You've told me this, right? Well, the truth is, this is the house I grew up in. Okay? Many of you probably know, I've talked about this before, that I grew up just outside New York City, right? I'm a Jersey boy, grew up just outside of New York City. I think a lot of people in New Jersey say that.
When I say I grew up just outside New York City, this was the view from my front door. That's The Riverside Church. That's Manhattan. This is, of course, a modern picture of what it was. If you look south, and we had a picture window looking south, you saw this. Of course, picture is never as good as eyesight, but that's lower Manhattan. My mother used to love seeing the cruise ships come in and out of Manhattan when they did that. It was really a great place to grow up, and we loved it. The coastline, that is, the riverfront, did not look like this in the 1970s. Do any of you remember the riverfront in the 1970s? The riverfront looked like this. Okay. It wasn't dilapidated. It was a disaster area. It was abandoned, okay?
It had been abandoned by the piers, by the railroads, by the industries that were polluting the river at the time. The last thing you'd ever want is to fall in the Hudson. You'd come out green if you fell in the Hudson. You'd be sick for months. It was dilapidated. It was abandoned. It was an eyesore. It was smelly. You can tell sort of where this is, behind lower Manhattan. If you looked up the entire riverfront, such as this, you can really see what it was like. It was not a pretty spot to be. Now, you may consider this a bad neighborhood, right? It was burned down. It was deserted. It was controlled by the mob, okay? You know what I think of it? I think it's prime real estate. Okay. This is what our industry is like.
If you look at the profit pools in a data center, what do you see? Let's talk about data center infrastructure. You've got servers. Servers are probably about 40% of data center infrastructure spend. How much profit exists in selling a server? Not a lot. Networking. Networking is about 10% to 15% of a data center spend. Good profit pool, 65%, roughly, gross margins by the networking guys in that space, but only 10% to 15% of the revenue. What about storage? About 40% of the overall revenue in a data center environment, not including software. Even not good competitors in that space are getting high 40s, low 50s margin. Our most recent quarter was 70%. It's a great profit pool, it's a large market. This is something we're very excited by, rejuvenating, gentrifying the data storage industry. With that, we think we're very well-positioned for success.
It's a big market. We think it's a great profit pool as well. We consistently take market share. Good markets and bad, I believe we'll continue to take market share. We're winning and innovating in the cloud. The dark cloud that's been hanging over our head, I believe we can make as much money in the cloud, as we can on-prem. One of the things we've done very well as a company is taken lousy infrastructure, consumer flash is lousy infrastructure, and turned it into high-performance, high-value, enterprise-class infrastructure. A lot of cloud infrastructure is lousy infrastructure. We're taking that and making it enterprise-ready, and because of that, AWS and Azure, et cetera, are very interested in working with us because it makes it easier for them to convince customers to move enterprise workloads into those environments. We're expanding our subscription-centric business model.
That includes arrays that we've already sold that now we just keep updating with new software, new hardware, where customers never need to make a new decision as to replacing that environment. Then finally, we think that with our gross margins, and as we grow and are able to get more leverage in our business, we have a very compelling financial profile. We've got a packed agenda for you today. We'll be starting out with a number of our product leaders who'll be taking you through more detail about this strategy. As you'll see, this is not a pie in the sky strategy. Everything we have announced that is GA at this show, it all contributes to the strategy. We'll have Prakash on FlashArray, Matt Burr on FlashBlade.
Dan DeCasper, very important, because it's Pure1 that ties this all together. Pure1 is the single pane of glass that creates the system environment associated with this new modern data experience. We'll have Dave come on to talk to you about our sales and marketing operations. We will have a customer panel that'll be driven by our Robson Grieve, who's our relatively new, I think 10 months, CMO. You'll get a chance to meet him as well. Finally, last but certainly not least is Tim Riitters will come up and give you an overview of the company from a financial standpoint. With that, I want to thank you very much, and we're going to move over to Prakash. Prakash?
All right. Welcome. I think this is probably my first conversation with this group of folks. I've been with Pure a little over 14, 15 months now, leading the FlashArray business. A little bit of an update around our business today. We've got over 16,000 arrays in the field today all phoning home, which gives us a lot of interesting data around what's going on from a telemetry standpoint with data usage overall in the industry. It's across 6,000 customers, and we, using our telemetrics in Pure1, as Charlie mentioned, we show consistent uptime with six nines today. Prior to Pure, I was at SAP for about 13 years, built mission-critical applications and systems there.
What's interesting about Pure, as I take a look at the journey and the momentum, mission criticality is built in by design, and we're able to showcase real uptime numbers on our website overall. The net promoter score, interestingly enough, last year, we had about 10,000 arrays. This year, we have about 16,000 arrays. Our net promoter score was 83 last year. It's 86.6 this year. It's gone up as we have more arrays in field. If you think about what that means, we're getting more signal directly from all of the usage of our arrays, and we use that signal to continually optimize our product such that we can deliver higher quality. This is kind of the foundation to a self-correcting, better quality system, and our net promoter score reflects that.
The announcements around modern data experience, I'm going to talk a little bit about how we make it real. In most of the customer conversations that I've had at Pure, everyone's like, "I love you for what you do really well. Performance block, you guys nailed it. It's great. The simplicity, the economics, the Evergreen, all the things that Pure has done really well, we love it." They're like, "I just wish you would do something else for me." We've had a ton of customers approach us and say, "If only you did this, I could go ahead and kick out this competitor." We've internalized that.
From an investment portfolio standpoint, I took a look at our strategy and said, "Do we just double down on performance block, or do we decide to go wider?" We're in this kind of enviable position of customers like us for what we do. They want us to do more in their landscape. You've got people willing to give you their money, and we're like, "Well, why don't we do the more?" That's the strategy that we've embarked upon, and at this conference, we've announced a lot of what that next set of more is, and we're just getting started. The first step was saying, you know what? How can we bring the performance, reliability, and economics that we've seen into the hyperscale world with Cloud Block Store?
It allows us to define a little bit of a leadership position around what good is in a public cloud environment. Things that the enterprise has grown up kind of used to. You have 100 terabyte database, you make a copy, still 100 terabytes because you use data reduction technologies. Those things don't exist in an over-provisioned cloud environment. We're bringing the benefit of, hey, you know what? Instead of when you copy your production database, it takes actually twice the capacity. On-premise, it takes no incremental capacity and requires no incremental purchase. We're bringing that into the cloud environment. This is the reason why we've got a great partnership with Amazon and have been working with them around the Cloud Block Store. The other approach is we've started with this product that's great, and we decided to make it 10x faster and 10x slower.
Now, you'd argue, why do you want to make something 10x slower? It turns out that a lot of people we see that there's a market very clearly for people who want to buy the premium thing for the most important thing, and there's a market for people who want to buy cheap. Cheap and deep. We even evaluated whether there's a little bit of this market for the middle, right? The retail analogy I use is, there's a Nordstrom, there's a Walmart. Is there a market for Target? Yeah, in retail, there is. In technology, is there a market for Target? We call these hybrid arrays. Our philosophy is no. Who wants to roll the dice and say, "You know what?
If I get a cache hit, it's 10x faster, and if I don't, it's 10x slower. As we take a look at introducing new products, we said, "Okay, let's go ahead and introduce a capacity-optimized product." Then also for the mission-critical workloads like SAP HANA that are in-memory based, or workloads that have high sustained peaks, we're going to introduce ultra-low latency with storage class memory. What's interesting around this market, if you just think about this capability, it's all block at this stage. Now instead of just staying in the performance block, we can go into the ultra-low latency and the capacity-optimized block market. In the block market, we're optimizing the addressable use cases and scenarios we can go into, by over 50% in the next 12-24 months. We've already got industry-leading win rates.
We've already got great growth. We're going to be making it easier by actually expanding our portfolio that way. If you think about the technology trends, flash started with single-level chip and then multi-level chip and then tri-level chip. It's one of those things that, with every generation, things get better, faster, cheaper, right? To some degree, reliability was the challenge. As you pack electrons together, reliability becomes more of the challenge overall. If you think about, I've got a 16 gig SD card at home for my digital camera. I got a 512 gig one as well. The 16 gig SD card I've had from many years ago, still using. The 512 gig one, I've burned through a few. Why is that? Because this is just a little bit of physics.
You pack some electrons together, you write to them, you've got higher wear rates. It turns out that we've written software that optimizes for the inadequacies of the technology and medium. 90% of our Purity engineers are software engineers. We point them at a medium, we say, "Hey, look at the problems that are associated with that medium, engineer around them in software." That's largely what's allowed us to go ahead and build a QLC-optimized system. In the same way, we were able to take advantages of Optane technology to expand into the ultra-low latency market as well. We're taking advantage of the new media types that are there and introducing systems there.
If you think about our portfolio, FlashArray//X, largely, you're talking about 55 to about three petabytes, 70% of which is virtualized, mission-critical transactional databases, Oracle, SQL, SAP HANA, Epic in healthcare. With DirectFlash Fabric, we've been able to consolidate a lot of DAS-based systems in that workload. We're now expanding to the ultra-low latency workloads with our DirectMemory Module, and we've introduced a new product, FlashArray//C, that actually starts at one petabyte and actually goes up from there. Going into next year, we expect to even build bigger systems there overall. The sustainability Charlie talked about in terms of mass consolidation and bringing it all together, we're able to go ahead and consolidate large data center footprints, more workloads into FlashArray//C. This is our entry point in terms of the journey, in terms of where we're starting capacity footprints.
Next year, we expect those numbers to go up. If we take a look at the workloads, the analogy I use is a little bit around kids and candy. We've introduced a product with FlashArray//X, where you've given the kids the candy, right? This product has a high-performance profile. They love the candy. That's great. Now, from a market standpoint, we've said, you know what? We're going to introduce something that's a little bit cheaper. It's the healthy protein bar. We'll call that the FlashArray//C. If anyone's come from the all-flash world and they take a look at, hey, you know what? I'm going to go to something 10x slower, right? No business is going to accept that.
If you think about the disk-based world where people were living in the past, they can actually get an upgrade where they get slightly better latency and better economics by moving to the all-flash data center. We find that this is actually creating a very interesting market opportunity, overall, for us to expand, to cut some of these cheap and deep stores and take share in the disk world, where previously our portfolio was, you know what? We were trying to convince everyone that they needed the candy. Now we have this opportunity to go ahead and play, saying, "You know what? If you need the performance, we can give you the performance. If you need even better performance, we can do that." You know what?
If really all you need is a technology upgrade for simplicity and Evergreen, you're buying cheap and deep, we have the ability to do that. This is the broadest portfolio expansion we're doing in the block market. Use cases that we see today, tier 1 virtual machines. 70% of all of our arrays today are virtualized. We see a lot of container workloads as well, which is interesting. With our Pure Service Orchestrator, we have transparent container pools for Kubernetes, software developers using APIs can actually go ahead and get storage pools from any performance tier, overall, across FlashArray and FlashBlade as well, which is interesting. We have the ability also, however, for VMs to set policies that you could actually tier. These are my important VMs, these are the VMs that I want to layer on a lower performance thing.
Maybe it's test dev, et cetera. Data protection. Of the 16,000 arrays we have out there today, one of the capabilities we've introduced, not today, but we've had it for a while, is synchronous replication, and we have over 1,000 arrays in production leveraging synchronous replication today. What are the other 15,000 doing for data protection? Maybe they have a cheaper thing on a secondary site, and this is now a market opportunity that's immediate, right? This is our installed base. These are people using Pure today. We can actually bring an offer to them in their landscape today overall for data protection. Workload consolidation is another strong use case as well.
The other piece that we started seeing a little bit of, with compliance and regulatory reasons, we see people looking at more of all data encrypted coming from their applications or having some subsets of data encrypted. We're introducing a new capability called EncryptReduce to ensure that we can protect the value proposition of data reduction, the benefits you get that make Flash economical, in an encrypted data world by integrating with the leading KMIP key store providers on the planet today. Thales is one of the ones we've worked with initially on initial product introduction, this allows you to ensure that whether your data's encrypted or not, you have the ability to go ahead and get the performance and economic benefits of Flash. Now, our cloud journey is interesting. Why did we do it, what did we do it for?
Well, we've seen a lot of customers say, "You know what? I want to go ahead and start, with a system, mission-critical transactional database, and maybe I want to lift and shift it into the public cloud." Right? It's kind of a migrate use case where I need to move it up there. Now, no one just moves storage independent of their application and compute. Some people are doing that with bare metal instances, some people are doing that with virtual machines, but the storage worlds are very different across the landscape. Charlie mentioned that every hyperscaler has their own independent storage architectures. On-premises, you have different storage architectures. You don't really have management monitoring and consistent APIs across any landscape.
What we've done by enabling Cloud Block Store on Amazon, available in GA now, we've been in beta with over 60-plus beta customers over the last year. We've enabled the ability for customers to have that consistency across that environment. DR, we've introduced capabilities for DR, where you can actually do cloud-based snapshots across on-premises and public cloud as well. Dev test, you've got a production system, you want instant agility in the public cloud to get started, or even HA. This is one of the unique things. In our beta program, we actually had someone set up a physical FlashArray in Equinix direct attached to our software in Amazon and are running an Oracle database with synchronous replication across the two. Meaning it's the same software. You could actually run active-active configurations between a physical array and software on the other side.
We didn't roll in hardware into the data center. We actually wrote this as all software. One of the things I mentioned earlier, we have software engineers that are optimized for looking at a medium and writing in software engineering the ability to go ahead and optimize for that landscape. Instead of thinking about it as, hey, it's general purpose software, we optimized and worked with Amazon directly to optimize for that landscape. The architecture gives you the reliability of S3 with the performance of NVRAM all in VFlash. If you take a look at the cloud architecture overall, the ability to get performance, reliability, and cost savings in the public cloud with the ability to go ahead and deploy hybrid is something that would've been challenging and customers would've had a lot of hard work to try to go ahead and build a consistent API.
Today, we had Amazon at our user conference. They said the quote. I'll quote them. They said, "We've done the hard work for you." That's the reason why Amazon's been a good partner in this joint engineering exercise. As we talk about unifying on-premises and the public cloud, we didn't stop with the technology. We actually thought through the entire business model on how people get this. Customers can go to the AWS Marketplace as of this morning, and in the Marketplace, they can actually deploy Cloud Block Store to their accounts directly from the AWS Marketplace. We also have extended our subscription, Pure as-a-Service, where a customer can buy a single subscription and across their on-premises environment or the public cloud environment, they just buy capacity.
Whether they today decide to start with 70 terabytes over here and 30 terabytes over there, and that mix changes to 50/50, or that mix eventually changes to 100 to zero, they're actually not recontracting, relicensing anything. It's one unified subscription across on-premises and public cloud. What does the Cloud Block Store do? At the end of the day, in standard storage terms, better, faster, cheaper. Right? We provide better by giving you better reliability. We give you better performance. We've optimized it for high-performance compute, and we give you consistent APIs across your entire landscape, so it makes it easy, and data mobility becomes possible. If you think about infrastructure decisions today. You're a developer, you're an individual, or you're a corporation, and you want to put your Oracle database on the public cloud.
You have over 300 different storage and compute instance types to choose from, and Amazon's adding new ones every year. We just started on this journey, but if you choose our Cloud Block Store, we abstract a lot of that decision away from you. If you want to basically non-disruptively upgrade in the public cloud between instance types, you can actually upgrade from one of our medium instance types to our large instance type without actually doing a data migration. We've brought the benefit of Evergreen and non-disruptive upgrade into the native hyperscaler environment. Going forward, we're actually working on some research now where we can identify all of the new compute and storage instance types that Amazon introduces and actually start leveraging that for you so your investment is future-proofed. We're bringing that storage virtualization capability into the public cloud by enabling non-disruptive upgrades.
I talked a lot about Amazon. We introduced CloudSnap for Amazon for taking your snapshot offloads for data protection to the public cloud. We've added Azure now. This is GA as well. Between both AWS, we support AWS S3. We support long-lived, infrequent access on AWS as well. We've added CloudSnap for Azure, where you have the ability to offload your snapshots for data protection. Companies now can leverage the cost economics of data protection in the cloud. This is one of the things that I'm going to discuss. It's around an acquisition that we did earlier in the year. We announced acquisition of a company called Compuverde in February, that is a file protocol stack. Interestingly enough, we're embedding that file protocol capability into FlashArray because Charlie talked about multi-protocol.
This is the IP and the technology we're using to bring multi-protocol capability into FlashArray, where previously it's been block, but this will allow us to expand the use cases into general purpose file shares and low latency file. If you think about the market today, we've had a product in the file market for quite some time. FlashBlade is a scale-out, bandwidth-optimized file and object platform. Low latency, you have mission-critical databases. Those mission-critical databases might have log volumes that are low latency, or you would have general purpose user shares that are low latency sensitive. We've optimized FlashArray for latency-optimized use cases, and we've optimized FlashBlade for bandwidth-optimized use cases. If you think mission-critical databases, you have transactional databases and you have analytic databases. Analytic databases are more bandwidth optimized because you need to churn all the data. Transactional databases are more latency optimized.
We now have products that actually cover more of the market end-to-end overall. The good news is, unlike the competitive landscape, we actually, on this project, started with the FlashBlade management and monitoring in the APIs, creating a file, same across the both products. Creating a directory, same across the both products. How you look at them in Pure1, same across the both products. Unlike the industry that is, you take a look at different vendors, they've consolidated different products from different vendors, and they all have different APIs, different management, different monitoring and data migrations between the two. We've actually done a lot of investment to ensure that we have consistency so people can start anywhere and go anywhere. You'll see that that's going to be a consistent investment theme as we take a look at how we're building out the portfolio.
Today, imagine a world if you could start anywhere and go anywhere. If I take a look at the Dell portfolio of PowerMax for the high-end and VNX and Unity and XtremIO, or you take a look at HPE's portfolio of whether it's Primera or Nimble. Everyone's got products that they've done market segmentation for. It's like this is a product for this market, mid-market. This is a product for the large enterprise market. This is a product for this market. I almost think it's an afterthought beyond acquisition because the company hit this, I can't grow, so I need to acquire. I take a look at that market, and that market complexity doesn't make sense to me. Why do you have this complexity in the market? Why do you need multiple products for multiple segments of the market?
The analogy I'll give you is, Pure, as a company, we've got back-office financial systems. We used NetSuite. I think we still do. Tim will correct me if I'm wrong. At some point, as a company, maybe you outgrow a mid-market ERP system. Then you have to take a look at an enterprise ERP system. Then you have to go ahead and say, "Oh, crap. I got to do a big IT project." Right? That's what happens when companies start with a segment product, a mid-market product, and grow into become an enterprise. What happens if a company starts small and needs to grow? In the early days of Pure, 10 years ago, we had companies, a lot of our early companies were within 25 miles of Mountain View, our headquarters.
They happen to be large SaaS companies now, like ServiceNow and Workday and Adobe and LinkedIn, and we have all these very large SaaS companies. Back then, they were SaaS startups, and they grew with us along the way. That's why our product grew up as infrastructure as code, API-driven, et cetera. We support every API-driven framework on the planet. We grew up as a company, and people actually didn't need to make a switch. Think about that, right? They're a small startup, and they grow into being these semi-large or very large enterprise companies. They didn't hit an inflection point. Evergreen supported it, and our business model supported it, and we gave them the ability to say, "You know what? There's generational upgrades, controller swaps, NAND swaps, new technology, new engineering, et cetera." Instead of going ahead and saying, "Guess what?
We grew up with you, and we made this non-disruptive. What if we can bring that same concept to more of your landscape? You know what? Across tier 1 and tier 2 or even tier 0, across block file, object, between FlashArray and FlashBlade, we cover all of the use cases. Multi-cloud, Matt's going to talk to you about our perspective on object replication in the cloud. I've talked to you about some of our block capabilities. We're on this journey to say, "You know what? What if there's no penalty to make a decision?" The number 1 problem in the storage industry today is before I actually get started to do anything, I have to be very careful before I hit buy or procure. If I choose wrong and I implement wrong, the penalties of changing are very, very hard, because the APIs are different.
I have to do forklift migrations, all of these different things. That's the problem we want to solve. This is an industry that's great at going ahead, and you'll see the news in the media and the press, and here's a new spec sheet. Here's the new 10, here's the faster box, here's the new feature, right? It's missing the one macro point that changing is hard. I know I'll use this iPhone analogy, which is probably very well overused, but I do remember going on vacation with multiple devices. Drag the digital camera around, drag the Walkman around. I didn't drag an eight-track around. Maybe Matt did. He'll talk to you about that later. I had a lot of devices. I had my phone as well. I had a beeper at one point. Anyway, that's a little bit of history.
When you think about what Apple did to say, let's consolidate a lot of these environments into one thing, we forget that we live in a world that's consolidated today and greatly simplified into one device. Now we can all talk about the new camera features and resolutions, and we can talk about all of the new shiny features and iOS updates and all of those great things, right? We've all gotten used to the new reality of a simplified, consolidated world.
This industry that's super fragmented, where vendors have fragmented products, and they're like, "Okay, well, on this product line, I'm going to introduce a Storage Class Memory shelf, and it works on this version of PowerMax that if you bought it from this year to this year, then you can add Storage Class Memory to the shelf." That's kind of what the industry looks like today. Our goal is to drastically reduce it such that you have a product for latency-optimized, you have a product for bandwidth-optimized, largely because the architectures of today don't allow you to consolidate the two things. I would love to get to a point where the networking technology moves, where you could link multiple motherboards together and actually blend scale up and scale out and disrupt the entire notion of scale up and scale out. Hopefully it happens in our lifetime, right?
This is a long-term technology inflection point that may happen in networking. Today, there's a reason for scale up and scale out technologies, and that's why our FlashArray and FlashBlade portfolios look the way they do. We've built a single management monitoring and API solution in Pure1 with a single product for latency-optimized and a single product for bandwidth-optimized with a single API, anyone can start anywhere and go anywhere. You know what? The good news is the application simplicity is actually starting to consolidate. People have largely learned that, you know what, for application virtualization, you're going to use VMs. For container management, the world has largely standardized on Kubernetes today. Right? Sometimes it's LXC containers, sometimes it's Docker containers, container management is largely simplified.
You want to know what horses to bet on, we know which horses to bet on in terms of application portability, whether it's modern apps or legacy apps. Having a consistent API-driven approach allows big, large SaaS providers to get started today. With infrastructure as code, so developers can start with agility. It's the way people are building in the cloud. It's the way Everpure helped enable early SaaS companies to build their business models. We've been doing that since the inception of our company. The biggest opportunity going forward is this ability to have AI-driven recommendations. Dan's going to talk to you a lot about that later today. We're on this journey where we have the ability to take a look at anything that's out there in the field and say, "You know what? Now that we have ultra-low latency storage class memory," right?
We don't have to go to the entire market or spend expensive sales dollars figuring out where it's good. We can actually go to the customers and say, "You know what? You have sustained reads for an hour and a half every Friday afternoon, and we can shrink that to seven minutes." We can do that before we even show up, right? We can have an SDR do that over the phone. If that's valuable to you, yeah, Intel Optane, a price premium in the market over Flash. Guess what? If the value's there, we can tell them exactly what it's going to be before they even get started. We can start getting smarter and recommending different workload profiles. Imagine when we extend this further with multiple clouds, and we can start doing workload recommendation across hyperscalers.
Imagine when we extend this across on-premises and public cloud, and we can say, "Hey, you know what? These workloads have high peaks, so they might be better in a cloud environment that's kind of bursty," versus, "You know what? This could go in a private cloud to save you money because it's not a bursty workload." We can start building all of that based on the telemetrics we have today. This is the reason why I'm really excited about the future of FlashArray. I'll turn it over to Matt Burr, and he's going to walk you through FlashBlade. Thank you.
Thanks, Prakash. All right. The first thing I'll say is, Accelerate's always fun. It's more fun when you're winning. It's nice to see all the customers in one place, and just, I think the more time you spend with them, the more time that you realize there's sort of a core principle that, I'm sorry, it can't be analyzed, and it's tough to measure. If you walk around and you talk to customers, they'll tell you that simplicity still reigns supreme. We're going to spend a lot of time talking about markets and products and NAND and NVMe and all the sort of expansionary things we're doing with APIs, et cetera. Like it or not, the recipe still lies in the simplicity of the systems. Case in point, we have a major oil and gas company here.
In fact, I didn't even want to participate in the transaction. I asked our sales team not to even try. They said, "Well, why are you going to block this?" I said, "Because it's impossible to win. You can't win. You have a one in a million chance." They said, "No, this guy's different." I hear that all the time. I talked to him, and he wasn't any different. I went back to them, and I said, "You're basically trying to be Sisyphus here. You're pushing this rock back up the hill, and it's going to roll right back down on your face, and it's going to crush you again.
You're going to feel bad, and you're going to complain to me that you don't have some feature. I told this to the customer, who on one hand thought I was rude and on the other hand thought I was pretty incredible for actually telling him that. The point of this is I'm setting this up in kind of a way for you to think about markets. The FlashBlade business in file and object is repeating the exact same thing that happened in Block on these S-curves of adoption that are tied to economics, and they're tied to workload types. As the workload types modernize and the economics change, the momentum builds into the business. We'll talk a little bit about the momentum in the FlashBlade business. The counterpoint to this, which sort of, that's what we can analyze, right?
You guys can go figure that out. Right? You can go talk to the NAND suppliers, and you can go figure out what the inflection points are in those S-curves. You're probably better at it than we are. I think our hardware people are pretty good at it, but you're certainly probably better at it than me. What this guy told me was, he said. And we were asking him to do something really hard. He had a huge 20 petabyte workload, and that 20 petabyte workload was predicated on about a petabyte and a half of performance and 18-ish or so petabytes of capacity. Where's our bean counter? Is our bean counter here? Is Tim here? There he is. There's our bean counter. If you went to our bean counter with this proposal, our bean counter doesn't have any budget, so you wouldn't go to him.
You'd actually go to me. If you went to our bean counter and you said, "Hey, bean counter, I have a 20 petabyte project that I need to do. On one hand, it's going to cost me X, and on the other hand, it's going to cost me X plus five." Bean counter goes X all day long, and you don't even get a chance because these are $20 million, $40 million projects that have 7 to 8 to 10-year ROIs on them. Long story, so stick with me.
My point was, if you go back and say, "Look, this is a seven-year project." If you can bifurcate your workload in a way that you can get economics, you can get capacity, you can get performance economics and capacity economics, you've bought yourself the optionality throughout the total project to basically get to a place that says, if I need performance, I can dial it up. If I need capacity, I can dial it up. What you can't have when you choose one or the other is the optionality to go choose the one that you really need in real-time. That dynamic, right, that's the one in a million part. This was the customer that was willing to go take the one in a million. When I asked him, I said, "Hey, we had this really long conversation about optionality.
Is that what drove it?" He said, "Nope." He said, "I was willing to bifurcate the workload because I knew that was important. Your system installed in a day, it took up a half a rack. There weren't a whole ton of cables. Your guys took my guys to lunch in the middle of the day because they had time to do that, your competitor was at least a rack, if not a full row, and it took each of them a minimum of a week to get set up." When this project gets going and I'm managing 80 petabytes, I don't have time to be babysitting storage systems anymore. The complexity of these 20-year legacy NAND systems, I can't state this enough, we way over boil the ocean. We do it worse than you guys do.
We way over boil the ocean when it comes down to what customers really want, and they want something that solves really simple problems that they don't have to manage. This $50 billion plus market that we talk about, these S-curves, I think everybody is pretty familiar with what they represent. I've been with Pure for nine years, so I lived the green S-curve, and I'm currently living the orange S-curve running the FlashBlade business. I myself have felt a little bit like Sisyphus in a way that I didn't feel it when I was on the green curve. One of the inherent differences there that we'll talk about is this file and object market happens later than the block market because, a, block is a lot easier to do, and you have transactional-oriented infrastructures that require low latency and bulletproof protocols. File and object market is different.
It tends to be a lower price point market. I don't feel like Sisyphus anymore. The rock has probably rolled over me four or five times. At least it feels that way. The rock stopped rolling over me about eight months ago because I don't have to argue with customers about why they should or shouldn't buy or look anymore. We're talking about, okay, here's the next thing I need, here's the next thing I need. The thing that's different in there is customers resisted the concept of all-flash in file and object. They believed that HDD prices would drop at a rate that NAND would never be able to outrun them for the bulk of the workloads. Fast object has changed that. S3 has changed that. Even some SMB workloads have changed that. We'll talk about that in a bit.
Our business is generally a really healthy business, and I know we haven't shared breakout numbers with you guys previously. We're not intending to break that out today, although we will talk about some aspects of the business that I think are very representative of our momentum. I can't tell how easy this is to see or read, but the shading on the left side, you can see we've built a really good enterprise business. Between enterprise, public sector, and healthcare, 3 businesses that you really want to be in. They're large public sector markets. You see commercial not being represented there very high. That's very much a low-end SMB, general purpose NAS-type file share environment. Our world is not there. Our world is in enterprise class, really fast file and object. If we look on the right here, it's represented across 3 primary workload types.
It's modern analytics, rapid recovery, and restore. This rapid recovery and restore concept, we'll talk about it a little bit, but think of that as there's nothing that can restore a database fast enough to be able to have business continuance meet their targets inside of IT organizations. Typically, if we were talking 15 years ago, you'd have a Data Domain box, and that Data Domain box was faster than Iron Mountain. Everybody remember Iron Mountain? Remember the guys, like, you'd take a tape. Literally, we'd back something up to tape, a guy would take it away in a truck, and you restored it by calling Iron Mountain, saying, "We need our tape back." That's crazy. I can't even imagine that that ever happened, but I did it. I sold it, built systems for it. Now it's not even disk-based restore anymore.
FlashBlade restore becomes a primary way to get into solving those database problems. What you see in terms of our business model is taking that door that opened for us around Rapid Restore and leveraging the data hub strategy to get inside. Basically, what that means is using backup budget to get in and then leveraging the FlashBlade and the data hub for a much broader set of use cases. Two out of three of our Rapid Restore wins, actually it's a little greater than that number, reference the data hub as the reason that they sold the deal. Think of it this way, as you're looking at the flowchart in your mind, it's have a database restore problem, also have a modern analytics problem, so I'm moving to something like Elastic or Spark or something like that.
I have funding for this, I don't have funding for that. I can bring the system in, I can bring the platform in, fund it under the backup budget, and then I can modernize my infrastructure for file and object through the data hub through that budget. It's a really good way to get into the business. We've got FlashBlade all over the world today. Tremendously successful business. I'm from upstate New York. It's generally not a place known as a very high technology place, and when I say upstate, I mean really far up. There's even a FlashBlade up there, so that means it must be incredibly well adopted. The customers that you see shaded in, not sure what color that looks like to you guys, but it's supposed to be orange.
Those are our customers that are here, that should you so choose, we could hopefully find a way to connect you with them if you're interested. I was going to flip this slide, but did you get the picture? Okay, cool. All right. That success has largely led us to something that I'm incredibly proud of. I think half a billion sounds better than $500 million, it's half a billion. We're on track to hit all-time bookings of half a billion dollars by the end of this year. It's a really good, strong business, and when we look at what that means, really good margins inside of the FlashBlade portfolio, but also inside of the overall file and object market. What it really means, though, and we'll touch on this again in a minute, so I want to just put this out there.
The startups are coming into the space now. There are startups that are beginning to emerge in this market. When FlashArray started, there's a lot of parallels between these two businesses, but one of the contrasts is when the FlashArray business started, there was a whole bunch of startups that were already there. Most of them are gone. Nimble is at HPE. Kaminario's kind of still around. XtremIO was there, Texas Memory Systems, but everybody generally started at the same time. The startups in this space are beginning to come now and first releases of their products, et cetera. We view this as having a three to three and a half year head-start and a $500 million kind of cumulative gain in dollars. It's not even the money, right? This is another hard concept to explain. It isn't the money, it's the experience.
File an object is so much harder than block that for those startups to catch up, sure. The comment that I always get is, well, they're a startup. They can build things faster. Well, sure. If they were simple things to build, yeah, that's great. The depth of what's required and what's required to sort of cross that moat and file an object to do, whether it's SMB, object or NFS in the core data path, that's really hard. Right? That's really hard. The infrastructures or the formats that you're seeing them adopt are complex software-only configurations, or you're seeing scale-out systems that are reliant on really intricate cabling and networking schemes. That's the antithesis of what that oil and gas customer wanted, right? It's such a taboo thing to say, oh, well, there's an appliance. Well, sure.
If your software is so good because you're a really good software company, that you have to wrap that software in tin that you make because that's what makes the whole package better, that's the simplicity that the customer wants to see, so they're not managing 400, 500, 600 different cables. That's a big deal because cables cause outages, right? When you looked at scale-out architectures in the block side of the world, those cables, when one was pulled or when a power cord was pulled, if the systems weren't cabled correctly, the metadata dumped out of the system and you lost data. We're going to repeat that pattern again in this space.
I don't want to get complex and technical here, but just understand that there's something to this three-year head start with a whole lot of customers that have been willing to pay for your product because you learn from what's in their environments. That is much more important in file and object than it was in block. What we've announced, or what we're announcing here at the show, are 150 blades. The importance of this is we go from a system that scales to about four petabytes that doubles in size to eight petabytes of capacity. This really speaks to modern workloads, whether it's log analytics, some legacy workloads, so less workloads, workflows like EDA, et cetera, where you have really large datasets. The modern datasets outstrip eight petabytes.
Yes, we're announcing support for 150 blades, but we have multiple projects underway both through software and hardware to yet again increase the size of these types of environments. What we're really trying to do there is we're trying to hit the inflection point of cost and capacity. We don't want to go build 100 petabyte system just to be provocative if the cost per gig or cost per terabyte on the NAND side in file and object is such that that's prohibitive. Right. I had someone ask me a really good question, which was, "Why double, not triple?" I don't think the economics support 24 petabyte systems yet, or 16 or 24 petabyte systems yet. We're probably still in that sweet spot of 8 for about another 12 to 16, 18 months.
Then we'll look to double in capacity again, at least with a new form factor. We think we're right at the sweet spot of this market. By the way, the sweet spot of this market is still the lower, it's still the bottom portion of this curve, right? It's there. It's barely onto the orange for the all-flash side of the file and object market, right? You can sort of look at, and I wouldn't say these things are correlated, so don't try to mathematically correlate it, but you look at Splunk or you look at Elastic and you look at the adoption rates of those technologies that rely on Fast Object or start in the cloud like Vertica. We benefit from the data gravity of those Fast Object systems.
I hate to say this because I think HDFS was a great protocol, but it's a 20-plus-year-old protocol. Cloudera and Hortonworks are good partners of ours for sure, but the adoption rate of net new opportunities for HDFS pale in comparison to those for S3 and Fast Object, and that's really the market that we aim for. One of the comments that I consistently get, or one of the questions I consistently get is, "Why did you not support HDFS?" It could have been a good business model. It could have been a good choice to support HDFS and give customers that off-ramp from, or give Cloudera and Hortonworks customers that off-ramp from HDFS to S3. That's a very expensive proposition, and we would've mortgaged our future.
The choice was really to move into modern file and object, and that was largely predicated on where the economics of the market are. Second, object replication. Prakash set this up really nicely. Object replication for us, think of this as modern S3. This is native S3 inside the system with object replication that allows you to do geo-targeting, geo-replication, et cetera. You can use any S3 target as a landing spot for FlashBlade data.
This would give you the ability, for example, to process something in the cloud or process certain workflows in the cloud, or store certain workflows in the cloud, or store certain amounts of data for workflows in the cloud when you don't really want to be doing very much with them to get the value of that economics, and then bring them back down to a localized processing entity, which would be FlashBlade via S3. The cloud is just another hardware target. I know that there's this mystical thing, and we're all trying to figure it out, and we've been doing it for 10 years, and we'll do it for another 10. The cloud is a hardware target. Right? What we do there will change, and it'll evolve. For purposes of workload and IT organization understanding, it's just another hardware target.
How you manage things there, which is why we're doing Cloud Block Store, is where the value comes into it. File replication. File replication is asynchronous, kind of what we're all familiar with in terms of table stakes for mission-critical workloads that need replication for disaster recovery type scenarios. Okay. This is worth spending a little bit of time on. When we sort of think about where we are in this journey, and I think Tim or Matt, it would be good for us to develop a slide somewhere along the way that sort of compares FlashBlade from the product perspective to FlashArray on the product perspective, to just kind of give everybody a sense because I think there's certain people that have followed our journey from the beginning and certain people that follow in the middle.
Last night I had a real interesting conversation where someone said, "Well, these products started at the same time, didn't they?" It's hard to explain to someone that one started four years after the other, and what does that mean, right? You can think of it in terms of ages. If FlashArray is kind of a teenager getting its driver's license or moving into adulthood or whatever it is. FlashBlade is kind of coming into junior high school or middle school, kind of in parlance of where I'm from. This really represents the next step of sort of getting S3 fully complete, moving to SMB 2.1 and 3.0. That will allow us to move into legacy workflows in oil and gas, media, and entertainment.
Really what the combination of these things adds up to is use case expansion from really five core use cases to about eight. Two of the three additional use cases represent a really material portion of the TAM in the file and object market. As you're thinking about what Prakash just discussed relative to FlashArray//C and FlashBlade, don't expect the economics of FlashBlade to dip into the sub $50K or sub $25K general purpose NAS market. You're always going to have some level of overlap in your portfolio. I'm going to get in trouble for this. I actually think that our opportunity with FlashArray//C is bigger than we've actually talked about because, we're not thinking about consolidation of $25K systems when we think of how big the FlashArray//C opportunity is.
I'm the only person inside of Pure Storage that believes this, so I don't think this could be an on the record statement, right, Tim? At the end of the day, the way we cover the overlap portion of that market is FlashBlade won't go that low. There's just no point in us. No one's going to put file shares on a box that has 40 gigs of throughput capacity. That's just not what people do. Right? We will continue to stay at the high end of the performance market. The way to think of it is high object count, high throughput, low object count, low throughput. Those are really the two different products. Over time, you really have a very nice platform that covers both of them, all tied together with Pure1, with the ability to move things up and down from the cloud.
I won't spend as much time on that as Prakash did. You can envision a world sometime down the road, when we reach a point with the core product on the FlashBlade side, that we're also doing something with file in the cloud. That's after we sort of complete the product from where it is today. All right. This sort of concept of consolidation as a financial driver. When you build systems for, or when you build architectures for block-type systems, you're generally servicing or what you're looking for from a consolidation perspective is you're getting databases, you're getting virtual server workloads, and you're getting VDI workloads consolidated onto one system. That gives you high performance, high IO and low latency. This is basically what you get, right? File and object workflows are inherently different.
There may be five different infrastructure stacks for a specific workflow. Pretty hyperbolic. It's probably a very provocative statement again, but it's not all that uncommon. I talk to customers that have five different stacks inside a workflow. Let's just say that the average person or the average IT shop has kind of at least three per workflow, which means you're managing generally three vendors, right? This space has broken out across a bunch of different or a couple different vendors. You're generally managing three things. You're managing like a purpose-built backup appliance. You're managing something for object. You're managing something for whatever your file-serving choice might be. The opportunity is yet again analogous to FlashArray to take, it's not the same, but databases, virtual server, and VDI, it's the equivalent in the file and object world, which are less end user facing. Okay?
Who uses databases, right? We do. We access that data. Who here flies United Airlines? Almost everybody, right? Anybody does not fly United Airlines? How annoying is it when you go to United Airlines to use the reservation system, and the thing just turns? Like that happens more with United Airlines than any other airline. That's because they're having a database problem. Right. We're the ones that feel that. In the file and object world, it's developers. The developer is the user. You know who gets more frustrated than us? Developers. Because when they're developing something, they have to actually run a unit test, or they have to run a test that the sooner that test completes, the sooner they can go on to writing more code. That's what happens in file and object. The user's different, the buyer is different, the people that run the systems are different.
When they can get to a point where they're only troubleshooting one thing to actually solve for where the problem might be, as opposed to looking into four or five different things, they will reduce the number of full-time engineers that it takes to run this environment. Each one of those five, on average, this is Gartner and IDC, not me, has five people to run each of those infrastructure stacks. Consolidating them down to one platform will be about a half of a full-time engineer. One of our largest customers does self-driving autonomous vehicles. They have a very large data set, a very advanced AI machine learning cluster, and they have a lot of really smart data scientists. They have one half of an FTE managing the storage portion of their infrastructure. That is down from five.
The number just happens to line up and support my argument. I didn't make it up. It is down from five. I hope the theme that's developing here is, yes, these markets are different, but there are some similarities. We've had a lot of success in it, but there's a lot of just history repeating itself, and we expect it to repeat again. Okay. Anybody here from the Northeast? Has anybody here ever driven through the Holland Tunnel? Has anybody ever driven through the Lincoln Tunnel? He's got both hands up back there. Holland Tunnel, Lincoln Tunnel, whether it's the GWB, whatever it is. That picture on the left is the traffic backup. I'm also from the Northeast, so I've spent a lot of time in this, where the traffic's backed up or it's closed, et cetera.
We all know what's going on with the infrastructure budget in the Northeast Corridor and the impact that it's having on economics. I think the thing that's interesting, and I wish I had built this so that it was circled. I think the numbers on the left are really interesting. By the way, anybody from Boston? I think this is true. I don't know if it's true. The completion of the Big Dig resulted in $7 billion in private investment over a four-year period in that part of Boston and created 43,000 jobs. Not to work on the Big Dig, but the output of the result of the Big Dig changed that. It was $167 million a year in traffic pattern changes. The point here is these numbers on the left.
Instead of repairing the infrastructure, it costs you three times to repair it than it does to replace it. The systems that we are replacing are from companies that are trying to repair, and they have a three-year lag if they started today and said, we need to build from the ground up. The hard thing here, because it's a little glacial, it's a little like I'm going to go find oil. It's like, wow, these things go so slowly on the file and object side because disks have been around for so long. Do I really see movement there or don't I? There is a train at the end of that tunnel and it's coming for them. It's coming right at them. That's what those systems are. It's bottlenecks getting into the Holland Tunnel or Lincoln Tunnel or GWB or whatever it is, or it's closed.
If you show this to a customer and you walk them through, I promise I'm not going to do this to you, and you walk them through the life of an IO inside a legacy file and object system, and then take them through our life of an IO, they will say, "I do not see those bottlenecks and I do not see those closures." Those closures are things like locked files. What's your name?
Arkay.
Arkay. You're a developer, and I'm a developer. You're accessing a file and you need to do something with that file, and I need to go get that file, and we don't have-- There isn't even a way to prioritize you over me. I would never prioritize me over you, but there's no way to do that in a system. If you're using it, I can't use it. If I'm using it, you can't use it. In our system, that doesn't happen. You can't go back and retrofit these systems in order to get there. You might say, how can you get to $500 million in all-time bookings in a market that you're saying you're still on the bottom part of the curve? Well, to me, that's just opportunity.
If we can make that happen in a market that is just starting to tip, that conversation, that customer that's thinking about how do we bifurcate these performance workloads, capacity workloads, it's a really, really great place to be in. All right. We've touched on this a couple of times. Up to 2016, these legacy NAS systems where you end up with the bottlenecking I was talking about before. That's good, but it's not like people just said, "Hey, it doesn't work, so we're not going to do anything." It doesn't work that way. I wish it did, because then everybody would just buy our stuff, but it doesn't. What they did do was they did what everybody else would do.
They figured out a way that they could make it work good enough for what they needed to get done. They built out DaaS architectures. If you look at Cloudera and Hortonworks and Hadoop and HDFS, these are very much DaaS architectures that have lots of drives in them, lots of servers in them, very complex infrastructure sprawl. One major problem. You have gold.1, gold.2, gold.3, gold.4, and you have disparities in these baseline data sets because you have different versions of these things based on sort of the expensive infrastructure sprawl. We solve two problems here. One is that consolidation actually leads to having one gold master across multiple sets when you're using generally S3 for object because you have a centralized storage repository. Two, we eliminate the complexity of that infrastructure sprawl in DaaS. DaaS is expensive. Right.
There are customers that have actually limited their data sets. I don't know if anybody in here has a car that at some point in time will have self-driving technology in it. I personally think we're so far away from that, it's unbelievable. I don't trust it. I wouldn't trust it. I know that the data sets aren't big enough yet to actually be analyzing the data to understand that we can recognize a stop sign. That's easy. Can you recognize a ball that's coming at a certain speed, or the density of a piece of fog that's going in front of the car, and which one's different? Yeah, sure. If you're going to hit the ball, is it that big of a deal? No. If you're going to go through the fog, is it that big of a deal? No.
What if the car thinks that piece of fog is a bowling ball because it can't analyze fast enough? It slams on the brakes and somebody behind you hits you. Right? The data sets are still really too small. I don't care what Elon Musk says. The data sets are still too small to get to the type of analysis that we need to get to, or to have the safety in those types of areas, which is why you see AI in things like Instacart, right? Amy Fowler talked about this morning, AI is an Instacart. It tells her that she's saved, whatever it was, 130 hours in grocery shopping. That's cool. All right. It's a good stat. That's largely where this stuff lives today. Some cool applications of it are Paige AI.
We've talked about it for a year or so, going back and looking at medical imaging to determine anomalies in cancer cells from cells that were pre-digitization is a pretty cool thing, right? Being able to go in and digitize those cells and then analyze them and look for pathological similarities across cell types that haven't previously been able to look at. Tens of millions of records from the '40s through kind of the '80s or '90s when they first became digitized. Those data sets are 10 or 15 or 20 petabytes. The reason that those studies are successful is not because they wrote better software, it's because they have more data. They have more data to look at, machines can look at it faster. What we see is that sort of 2016 kind of world and beyond that it's more efficient than DAS.
It scales compute and storage independently. We've all heard this one before, but this is the other one that bewilders me. If I have to build my infrastructure, every time I need more compute, do I want to pay for storage, too? Not if I don't have to. If every time I need more storage, do I want to pay for compute? Not if I don't have to. I want to be able to do those things independently. That's a lot of flexibility that current systems don't offer. Lastly, it's that sort of cloud-native architecture with the S3 protocol or the native S3 protocol, I should say, because it is different, and the ability to do things both in the cloud and on-prem.
What we see here is the business is accelerated by data gravity in general, but also where the economics sort of bisect that opportunity to really build much bigger systems. I think the key takeaway here is I'm super excited and probably a little cavalier right now because our business is doing really well. I'm one of those people that tends to get pretty excited when things are going well. The second takeaway is that this shift is here, right? The opportunity for us to really kind of excel and I think to be true to my own values, we're all going to benefit, I think, a lot more from advances in file and object than we ever did in file and block, or excuse me, in block. Block benefited business. I like to think of file and object as benefiting people.
It's just kind of a different paradigm. Thank you. I guess we're Okay.
Got shot over a little bit. He did great for the rest of us that are here. For time constraints and feel free to stand at the door in the back.
There's a lot of here. Okay, ready to roll? Great. Let's get going. Welcome, everyone. My name is Sandy Casper. I'm the VP and general manager for Pure1. This is the first time for me I'm in front of this audience. Really excited and hope you'll find this useful. What I wanted to do was start with a bit of a problem statement. Again, maybe for those of you who aren't as familiar with Pure1. In a nutshell, what's happening in enterprises is that enterprise admins have an increasingly hard time dealing with the complexity of the equipment they have to manage. These days, it's not only in their data center, of course. If you want to add insult to injury, they also have to deal with the public cloud, right? We're kind of in that boat now, of course, as well, right?
We're giving them CDS, now your typical admin will have to worry about what's on-prem as well as in the cloud. Another trend we're seeing is that there's more and more desire to have self-service IT. Different roles in the organizations, people like say a storage admin or a DevOps person or a developer, they have very different needs from their management system. These systems need to accommodate that. If you look at what's out there, it's really not awesome. A lot of these tools are very difficult to operate. Some people still sell shrink wrap software in this day and age, believe it or not. You have to install it and configure it and upgrade it and learn it. What's also sometimes true is that you get a product from a vendor to manage your storage.
You would find that that one product you bought is not actually supported, right? Sometimes they don't cover the entire portfolio. We think there's two things that are really key to do this successfully. One is automation. Again, these environments are too complex that you can do manual administration. The other, we really mean this, the best management is management that you don't have to do. The products are so simple. You can just set and forget, plop them down. They always work. You only interact with them if you want to, and not because you have to. You can also use AI and ML to really help here. It provides great tools. I'll get into that a little bit in just a minute.
Our strategy, and this has been the case for the last few years, is really on these three pillars. Self-driving, 6 nines reliability, the ultimate in storage management simplicity, set and forget. Multi-cloud, where we assist the customer's journey into the public cloud. We see that as part of our mission. Full-stack, this is where we extend our reach from the storage layer beyond, basically up a little bit into the infrastructure. VM Analytics, which I'll get to, is a great example of a sort of a first foray into that. Underneath all of this is Meta. Meta is our AI ML engine that powers basically everything we're doing. If you want to look at it more from a block diagram, it would look something like this. This is storage layer at the bottom.
This sort of jives nicely with the modern data experience that Charlie described this morning. We have the cloud side of things, the Cloud Block Store and Cloud Snaps. We're getting more and more opinionated about the hypervisor and the container layer, and so I'll get to that a little bit. There's Meta on top of all of this, and there's basically three services that we expose to the end user: global management, predictive support, and predictive analytics. In this presentation, for the rest of it, I'll take you through these three, and I'll tell you a little bit about what we're doing in all of these areas. Important to note, you see at the top the different user profiles. More and more, we're starting to customize again for the different constituents who might be using the system.
Related, but a little different, we also want to give customers complete freedom in terms of the device they want to use to interact with this. Web, mobile, or just APIs. If they want to run their own systems and just pull data from us, that's fine, too. Before I go through the three pillars, this was actually up at the keynote, so you've probably seen it already if you look carefully. We're leveraging a lot of data in Pure1. Since we started with this, and this is now about eight or nine years ago, we've constantly observed our customers arrays and accumulated a lot of data. You can see we're taking in about 50 terabytes every day. We sit on top of about 15 petabytes, and we're running about 1,600 containers and about 1,400 virtual machines to power the service.
All right, let me go through the three pillars. Let's start with global management. One thing that's really important to us at Everpure is that when we ship a product, it will be supported in Pure1. I would almost go as far as say we wouldn't ship it if the support in Pure1 isn't there. As we ship Cloud Block Store, it's now available, brand new. Guess what? You log into Pure1, support is right there. Same with Object Engine and anything else we'll be releasing going forward. We're getting better at dealing with very large enterprises. When we shipped Pure1 for the first time in 2015, we had customers with dozens of arrays. Now we're in this fortunate situation that we have customers with hundreds of arrays.
If you have fleets that large, then showing everything in a single screen is not the best user experience you can create. For those people, we created a feature, which was recently shipped, that we called Use. It's very simple. You define a filter. You can filter anything like location or the type of hardware, the firmware that's running it, what have you. You can assign that filter back to an end user. As that end user logs in, they see only the filtered arrays. For example, if you have a San Francisco site and a New York site, you can very easily just partition. The New York people see the New York arrays, and the San Francisco people see the San Francisco arrays. VM Analytics. I mentioned this briefly earlier.
This shipped about a year ago and has done phenomenally well for us. I think it was also mentioned in the keynote, actually, a couple times. We have close to 900 customers on it already, and we're closing in on 900,000 virtual machines that are being monitored. If you're not familiar with it, in a nutshell, it's basically a performance troubleshooting capability. If you look at the screenshot there, we're showing on these so-called Sankey diagrams. At the very left, you have the VM. At the very right is the storage array. We show how data flows from the VM to the array and everything in between. We annotate this figure with the performance metrics at every stage. If your VM is slow, you can take one look at this, and you'll immediately know where the bottleneck sits.
It makes performance troubleshooting super simple. We obviously have a lot of customers that use the VMware environment, and so that's one reason it's been such a hit. Now what we're doing here is, we'll continue to build it out, but we're also going to create a paid version, a separate subscription that we call VM Analytics Pro. The idea there is that we'll create, over time, a family of Pure1 Pro subscriptions for different types of functionality, and this is the first example of one of those. You see some of the things we're doing there. I'm not going to go through all of them, but essentially, you sort of start with data retention. Today, with the version that's included in the Evergreen subscription, you're going to get seven days of data.
If you want to do a, let's say, you want to do a period-over-period comparison, maybe you want to compare what happened last Thanksgiving, Black Friday, versus the year before. If you're sitting here in September with seven days, that's very hard to do. We're going to give customers 36 months of full resolution so they can do these types of things. Once we have all this data, we can bring in our good old friend, Meta, and I'll get to a little more of Meta in a second, and build a bunch of really cool AI and ML-driven type automation to just make having and running a VM or stack that much simpler. Subscription management. That's another sort of interesting So other than that we're getting into paid subscription, this is another interesting direction for Pure1 as a whole.
More and more, we think of it as a sort of a central delivery vehicle for the company's as-a-service subscription offerings. The idea is that a lot of the customer journey from discovering a product, purchasing it, installing, management, monitoring, all of that will take place in Pure1. It's that central portal. The first step in that direction, specific to subscriptions, is the subscription management screen. You can see it's already available now. You can see all your VM Analytics, all your CBS, your object engine, your ES2 subscription, all neatly in one screen. You can track your utilization relative to what you purchased, et cetera. This just makes, again, super simple to sort of stay on top of that. The mobile app. Again, that's another one of those things you kind of have to see to believe.
This was a great challenge for us that we set ourselves three years ago. We were like, "Can we build a useful fleet management capability condensed to the screen of a little mobile phone?" Right? I think if you asked our customers, they would resoundingly tell you we succeeded in doing that. We have people tell me it's the first thing, not that I would recommend it, but it's the first thing they look at in the morning when they wake up and the last thing when they go to bed. It's essentially a simplified version of the Pure1 web experience tailor-made for the mobile phone. We continue to build it out. We just added, it says up there, Remote Assist.
It's a feature we just added that lets support access your arrays, and you can enable that now from your phone for your entire fleet. Just a lot more convenient than it was before. All right, next pillar is support. Before I get into what we're doing there, I want to quickly offer a taxonomy for how we're thinking about support. There's different ways you can support a product, right? As a vendor, I can ship you a product. The product breaks, you call me, then together, we work on fixing it. That's sort of the most basic. That's what most consumer products do, right? The next level up is I can ship you the product. It breaks, now as a vendor, I actually find out it broke. Now I work with you to help fix it, right?
Of course, by far the best way to go about this is, as a vendor, I ship you the product, but I constantly observe the product and what you're doing with it. If it ever gets on a trajectory where it might break, I work with you to prevent that from ever happening. Right? That's exactly what we're doing with Pure1, and that's one of the chief reasons why we have a 86.6 net promoter score, by far the highest in the industry. Basically, the way this works, and this is a topic I could talk to you for hours, but very briefly, the way we do this is, let's say we have an outage, right? We hate outages. They're very rare these days, but they still occur.
Our goal is that every outage of a given type can happen only once, right? If we have an outage, we work backwards, and we look for leading indicators to that outage. With hindsight, right, we can go and find them. Sometimes that's a simple little string in a log file. Sometimes it's a lot more complicated, where you need to correlate different data sets. In either case, we mine through the 15 petabytes of data that we have and look for these. Once we find them, we add them to a library of checks. Subsequently, 24/7, 365 days a year, we monitor these telemeters as it comes back. Whenever we find these indicators, we work with the customer. Sometimes that means they have to upgrade an array, go to a different software version.
Sometimes they just need to make a config change or change of workload. In all cases, you can avert these outages. A few other things we're doing, like customizable alerts, notification, again, just to deal with larger fleets a little better so people can have more targeted ways to send alerts. Anomaly detection, that's another really cool sort of meta feature where if you look at workloads, they all have a rhythm to them, right? They change hourly, daily, weekly, monthly, quarterly, sometimes annually, depending on the application behind them. We already have that notion of that rhythm for all the workloads our customers have. We can now detect deviations.
If suddenly it behaves in a different way than it historically did, we can work again with the customer and see maybe it's benign, they just added something or made a change, or maybe again, it's an indication of a problem. All right, closely related, predictive analytics. Here I want to draw a very quick analogy with self-driving cars because it nicely illustrates what we're doing. If you have a self-driving car, it's on the road, it has all these sensors, right? It collects data around itself, and it then makes decisions based on that data. Go left, go right, accelerate, or brake. Right? It also takes all this data and sends it to the manufacturer of the car, where they go and analyze it, and they tweak their software and make it more and more reliable. Right?
You can say that each individual array, or each individual car in this case, benefits from the collective experience of the entire fleet. We take the exact same approach to storage. Right. Storage is different, but it's equally complex. Right. No two arrays, no two SAN configurations, SAN switches, host configurations, applications are ever quite the same. Right. To sort of navigate this complex environment, again, you can leverage that data and then build on top of that. One of the most important things you have to do to get a very seamless, always-on storage experience is proper workload sizing. This simply means putting the right workloads onto the right arrays. That is extremely difficult for a human to do. Because you can sort of look at a workload, and every storage admin would confirm that.
You can look at a workload and you go like, "Yeah, maybe I can move it over here." You move it, you have very unexpected results because these workloads interact in very complex ways. We see one of two things happening, neither of which is good. Either customers over-provision their array, so they buy a lot more than they actually need, right? Which does not go through the customer. Or they under-provision and put too much on it, and now they have slow application, sometimes downtime. To deal with that, we came up with this concept of Workload DNA. That's a fancy name for basically just saying we characterize each workload with over 1,000 different variables. We have hundreds of thousands of these workloads stored, right? Then we use machine learning, and this is a perfect example.
I'm sure you're hearing a lot about machine learning, and nobody ever gives a concrete example. This is a perfect example of how you can use machine learning. Basically what we want to do is given a workload and its DNA, what kind of load factor does this put on an array given a set of other workloads that are already there? You can build a virtual machine, ML model, and you can run your workload through that and let it spit out that load factor. Then you can go and compare that, whatever the model spit out, with what you actually observe. Because of course, we have that information as well. Then you go and tweak the model, right?
If you do this a few hundred thousand or a few million times with all the workloads, you get a pretty darn accurate predictor of what will happen as these workloads move around. Again, that's exactly what we've done. Armed with that, you can now answer three what seem like very basic questions, but are, again, every storage admin would tell you they're not that easy to answer. Number one is, given an array, how much longer can it run before it's going to hit the wall? Right? These things are sometimes highly non-linear. The second is, if I were to upgrade hardware, what would happen? This seems such a fundamental question, right? I have a, in our world, it's an X50, I want to go to an X70. How will my load look like?
You can stare at the data sheet of an array, ours or our competitors, until you're green in the face. There's no way you can, with any kind of accuracy, predict what will happen to your workloads when you do that upgrade. Meta will be able to tell you exactly. We have tools, and if I had more time I would demo it. If you want to see it, a lot of us can do that for you. You can simulate movement of workloads from one array to another. It'll tell you exactly what will happen on both sides. Similarly with upgrades, if you go from one array to the other, we will tell you exactly what the impact is. Lastly, we also let you scale workloads. Say you have a database, it's a very common case too, right? You have a database.
You know you're going to double your users. First order load is proportional to the number of users, right? We give you an opportunity to just say, "Let me double or triple this workload and tell me what will happen to this array after I've done that." Again, Workload Planner will tell you exactly what the impact is. Those are some of the things we're doing. There's a lot more coming here. If we do this again next year, I'll have a lot more to talk about. For today, that's all I have. There's several sessions that we're doing on this topic. The ones in gray, unfortunately, already have passed. If you're interested and want to learn more about any of these topics, please find one of us. We'll be more than happy to walk you through it.
Then there's a few more that we're going to do tomorrow on basically a bunch of stuff that I talked about here. That is all I have. Thank you very much.
Thanks, Dan. Good afternoon, everyone. My name is Dave Hatfield. I look after our go-to-market business. Nice to see many of you again. Hi, Colin. I'll give you a quick update. I will tell you, though, to start, yesterday was our global partner forum, we had kind of order 1,400 unique partner principals, presidents, CEOs of their businesses, of MSPs, GSIs, VARs that were there. I told them it was my favorite day of the year to be able to represent them because we fulfill all of our business through them, and we get a lot of value-added solution work and services work from them. This morning, as I was meeting with many of the one-on-ones that I had, I was meeting with one of the largest healthcare organizations in the U.S. and had a recent merger.
I was thinking the whole time that if I can just get those people up here on stage, which I think we are doing a customer panel after this, to express to you how different the experience is that we're providing our customers from their mouths unfiltered, I think it would have answered a lot of the questions that we keep needing to answer and educate all the different kind of constituents that we have out there, including all of you. There were two guys. One was from the acquired company, which was an HPE shop. The previous owners had become Pure Storage fans about three or four years ago off of largely an EMC and a Dell environment. They had come to the Pure//Accelerate three years ago, and they said, "We came into Pure//Accelerate as skeptics.
We ended up doing a proof of concept with you, brought you in, ended up replacing all of the Dell and the EMC. The two folks that were from the acquired company, that were HPE leaders and strong supporters, said the same thing. Last year, they came to Accelerate and were skeptics, but they kind of had to come along because these guys were already bought in. Now we've been deploying it for the last year, and it's completely different. It's just fundamentally different. It just works. They actually printed up T-shirts with their families with the Pure logo on it saying, "This stuff just works, and I get to see my dad." Like it's crazy stuff like this actually happens here.
Trying to articulate that to an investment community that is more than just flash, it's actually software, and it's a data experience that I think Charlie walked through in his introductory remarks, I think, does capture the positioning and the value of a lot of what we're doing better than we've done in the past. Quick update from my point of view, we're taking share. You all saw the IDC numbers last quarter, the growth at 28% when everybody else was flat to down. At one level, in an investment thesis, that's not a good place to be in, but it does absolutely describe what we're doing is different, and we're expanding the moats.
Our goal is not to be number 4, but we do believe we will be number 4 going into 2020 after just a few short years in operations in a pretty competitive market. We do believe and are very focused on being number 1 or number 2 in the years ahead. I get more excited about the next 10 years, and I've been pretty enthusiastic about the last 10 that we've had. Nice trajectory in this market. In cumulative bookings lifetime, we crossed over $5 billion. A lot of what we've had to deal with as we're moving into the Global 2000 or moving into more conservative buying communities is how safe are you? Nobody gets fired for buying the old guys. Our argument the last few years is nobody gets promoted either.
Now I'll close today with what I closed with the Global Partner Forum yesterday, which is we're now the innovator. Everybody expects that from us to be the disruptor and the innovative choice, but also the safe choice for large enterprises in terms of financial viability, in terms of product portfolio, in terms of ability to execute, et cetera. The $5 billion, I think, does provide some credence to that. We're growing at scale. Not a lot of growth names out there for all of us to consider that have been able to maintain mid to high 20s growth on a billion-dollar-plus run rate. The fundamentals underneath this, as you know, are about repeat purchase rates, are about win rates, are about ramp to productivity for our sellers, are about alternate routes to market and portfolio selling.
A lot of things that we got to do to continue to maintain this that I feel confident in. A 31% CAGR blended over the last four years. The number of customers at over 6,600 now and averaging around 300-400 per quarter is actually a metric that we like. We're not looking to be in the mid-market. I think most of that business goes to the cloud. We think most of that business goes to the cloud over time. It's less about volume as it is about quality and investing in the markets and the segments that matter. We feel nice about the trajectory of the customer adds and the payout on the investments or the areas where all the storage is, where all the data is, continues to grow nicely as well. A couple of new metrics here for you.
One, the G2K customers, and this is a customer count number on a year-on-year basis is solid. We're lucky enough to have Frank Slootman on our board and Anil on our board for a bunch of years as they were growing ServiceNow and Workday. I'm a pretty good student of sort of probably not as well educated as you guys, but of number of new G2K customers that get added on a quarterly basis. If you're doing 20, 30, you're kind of best in class. These sales cycles are longer. It's harder to break in, et cetera. We've got a really nice run rate of G2K, and we're continuing to focus, but there is a little bit longer sales cycle there. Cloud native businesses are trending really nicely ahead of the growth of the business. We're a perfect fit for any cloud native business.
It could be software as a service, infrastructure as a service, platform as a service, MSPs, partly because our technology aligns very, very well on taking advantage of all of the innovation that's happening on the hardware and the software level, but also the commercial model that maps to this which maps and aligns very well with their P&L. Nice growth on our cloud customers as well. The reason is obvious why we're focusing on these markets. It's where the data is, it's where the revenue is, and it's where our product portfolio really shines. On deal sizes compared to company average, it's two and a half times the size on average for the first deal in, and then the repeat purchase metrics we've shared with you over the years.
I think our top 25 in our last earnings was $9 additional for the $1 in inside of the first 18 months, and then we like the trends from there. It comes in with a higher ASP, but also then it multiplies better from there. Our win rates in the G2K, a lot of other vendors in our space talking about win rates in the enterprise or challenges in the enterprise, and we're taking share there. We feel confident that we're doing well in that regard. On the cloud deal size, 1.8x so nice size, and terrific win rates in terms of compared to the rest of the company average. We look at this stuff with a lot of precision, and we feel good about those numbers.
While we've got a nice share of the G2K, we have a whole bunch more to go get. If you're 28% penetrated, you got 72% to go. That gives us a lot of TAM and a lot of addressable market to go focus on, which is partly why we added the sales capacity that we have over the last three or four quarters, largely to go attack the G2K. Obviously, that investment doesn't fall to the bottom line because they're ramping. We know with those win rates and with that ASP and those repeat purchase metrics, we need to put the investment in, and we're confident that it'll pay off both in terms of ramp to productivity, which I'll show you on the next slide, but also the operating margin leverage that we get from that once they're ramped.
It just takes a little bit of time to get them there. Lots of TAM to go attack in the G2K. This is something that I want to make sure that we actually were going to change the slide, and we just didn't get to it, and I apologize for that. Over 30% or so of our business comes from cloud native customers, as I've defined it. Another third is commercial, rough and tough, and the other third is enterprise. As you look at the mix of our business. We have a nice portfolio of customer segments that I think maybe gets overlooked a bit. We've got nice precision and different models that support each of those. We feel good about that. I guess it's in the slide after this.
Yeah, one more slide on sort of the upmarket focus and the bias in the context of number of million-dollar deals. On my right, your left, growing nicely, outpacing the growth of the overall business. This is the number of million-dollar deals. Of course, we've given you one metric, but we're not giving you the good one, which is how many numbers we actually provide there, but you get some context for the growth. On the right side, 45% growth overall in the combination of cumulative customers that have invested strategically with Pure Storage, defined as million-dollar cumulative spend or $5 million cumulative spend. The orange are the million dollars, and you see the growth in $5 million customer installs growing very nicely at 56% CAGR.
We have customers that are really betting meaningfully on Pure's technologies, and why I'm so excited about the next few years, but in particular, the announcements that we made today, is they want to buy more stuff from us. They want that same Evergreen experience. They want that simplicity experience. They want the Evergreen model from an ownership perspective. We didn't have a product portfolio to be able to go help them in tier 2, really. Rapid Restore was great for being able to restore at the high end of those application workloads. Now we have a product and a price point that can go after tier 2 much more aggressively. They want to buy business from us in a multi-cloud context. People love our support. People love the experience, the attitude, and the automation and the predictive analytics that we provide with AI.
To be able to layer that on top of what they get from public cloud vendors is pretty compelling. As we think about delivering the Pure experience that they've come to love and be accustomed to in their data centers, to now be able to do that in AWS and over time in GCP and in Azure, it's a very compelling thing from a customer point of view. They want that support. They want that experience. They want those data services. It also kicks in another route to market that we haven't largely participated in, taking advantage of marketplace, but taking advantage of credits. Many large enterprises are spending tens of millions, hundreds of millions of dollars on a multi-year contract with cloud, and they have these credits that can be applied toward solutions that participate in that environment.
Well, they can apply those credits now to us, the revenue flows through to us. It's a nice route to market. We've factored in the forecast and the revenue of that for the balance of the year. It is for Cloud Block Store specifically, it is amortized, et cetera. It'll take some time for us to build it, if you think about the route to market that it opens up and you think about the relevance of what it could provide over time, and we think will, it's pretty exciting. I think these CAGRs of people betting on us are indicative of continued growth for those types of metrics in the future as well.
Productivity, as you know, we've made quite an investment over the last several quarters in additional selling capacity because we see all the forward-looking metrics that we've shared with you supporting those and saying it makes sense for us to continue to invest. If our win rates were different, if our ramp to productivity were different, we wouldn't have had the confidence to make the same level of investment. The good news is, not only were they the largest selling classes that we've had in our history, we're on our 27th cohort class. The gray are the new hires, the larger cohort classes at 18 and 19, I believe, and 20, are outpacing the earliest cohorts, which are most productive, and those that are at the target model. We watch this like a hawk.
Anything we can do to drive higher productivity here drops directly to EBITDA or to bottom line. The more products we put into these reps' hands, the deeper we get into the G2K where there's more serviceable TAM for them to go after, the faster this ramp will occur. I think we're hiring well. I think we've got the right traits. I think we have the right onboarding. I think the real inflection here happens as they get more products in their bag to be able to sell. They enter into their territories with more pipeline because we got better awareness, and we're participating in the advance. You've heard me say for years that our biggest reason for a loss is not being in the fight to begin with. To the extent that we can get in more fights, I love our chances.
The partnership strategy that we have, more capacity that we're putting on the street, the portfolio build-out that we're doing will allow us to compete for more and more of the serviceable TAM and the customers, and they want to buy it from us. The average net promoter score as one external third-party validated metric, I think for our competitors in the low 20s, and ours is an 86.6. It's night and day. To convince that person from the acquired healthcare company to look at something besides their existing vendor, when they're comfortable with it and they're familiar with it and they've had it for years, that's the challenge. How do we get that person to come take a look at us? Our competitors are helping out with this.
When they launch products that require disruptive upgrades for their whole fleet, that gives us an opportunity to go attack. We look forward to doing that over the next 12 to 18 months. This ramp to sales productivity is a really important metric. It's early days, and the ramp is encouraging. It's also encouraging that we've got 54% of our sellers with less than 15 months of tenure. What's encouraging about this, it means you have a lot more selling capacity coming online and maturing without having to spend more selling dollars. From a margin and operational leverage perspective, this is where we get a lot of that to hit the bottom line. This was very deliberate.
We knew the products that we were working on the last year and a half, two years, were going to hit the market, were going to open up more buying and selling opportunities. As this comes on, you get natural leverage, provided they maintain the productivity ramp that they're on. Okay. Quick updates on some of the selling metrics that we look at. From a strategy and priority perspective, I'm not a big fan of big changes, especially in selling organizations. We like incremental tuning as you move forward, and that maintains productivity. These three priorities have been pretty consistent over the last few years, and it's just about specific bets and investments underneath to make them successful.
Moving toward a platform motion and really driving portfolio and getting higher attach rates and multiple products per seller or per customer or per account, per partner, are the types of things that we talk about as we move into the portfolio selling motion. We have really nice pull-through from both our core FlashBlade and FlashArray products. There are lots of customer environments that are much more distributed scale out file and object. We lead with FlashBlade there. Once we're in there and they get a taste of the simplicity and the Evergreen experience, we look for Oracle workloads or SAP workloads or other block workloads that can pull FlashArray or our FlashStack converged infrastructure in. Same goes true, obviously, it's even more so on the FlashArray side. We have a larger base there.
People fall in love with it, and they want to have us help them with other use cases, et cetera. That platform motion continues to be a priority. I think the packaging and positioning around the modern data experience, with a little bit of a services-led assessment level consultative sale on the front end as we've gotten to a larger multi-product portfolio becomes important. I believe we have the best selling organization in technology, and I'm proud of that. I think it's a competitive weapon and an advantage, and they're used to selling. Most of the people that we have in our sales organization have sold portfolio of products in the past. Giving them the tools and the actual products is something that we're excited about.
Channel independence, this is a combination of alternate routes to market to map to the segments that are important to us. You've heard us talk about MSPs with Rackspace and GSI traction and the $100 million deal plus that we signed, et cetera, as real, meaningful commitments from this route to market to help us get access to the G2K. Also it's about making them more empowered and independent so they can quote, configure, price, close, upgrade, the whole shooting match, with limited touch from our sellers. As you go into the commercial motion, we want more and more of that business and that selling activity to happen independent of any Puritan touching it. We've seen some nice progress on that. A lot of systems, a lot of automation, a lot of business process.
Part of our program update for our partner program last year around standardizing pricing and discounting levels and putting incentive structures in place was about enabling them to drive independence. It's a multi-year journey. We're into it a couple of years, and we're seeing some nice progress there. Finally, resourcing the segments that matter. By that, it's not just about span of control and territory of a seller, but it's ensuring that we've got the domain knowledge around the partner alliance ecosystem. It's having subject matter experts to be able to represent us well. In healthcare, as an example, in our marketing organization, we have former CIOs, some of our favorite customers. It's one of the challenges we have with our go-to-market motion and our products is that our customers love us so much they want to join us.
It's like, "Well, you're maybe worth more to us on that side of the house than on this side." There's always a balancing act. For the select few, where we can have them be part of the team and bring that competency and that domain knowledge in helps us articulate an overall go-to-market strategy for a specific industry vertical with domain knowledge. It's a horizontal technology platform. What we do, it's not a bespoke thing that you have to go do a bunch of tuning for it. There are different partnerships and different routes to market. As an example, in healthcare, you want to make sure that you're riding in and partnering with and compatible with Epic and the other EMRs that are out there. It starts with, okay, let's be compatible.
Let's figure out the right words and packaging and dedicated sellers to go bring it to market. Become the enabler underneath their own hosted infrastructure as well. Those are the types of things that are specific to a market segment, but it's the same core IP and technology that underpins all of it, so there's a lot of leverage in it. As we focus on the industry-focused segment approach, that drives better engagement. It gets us higher in the organization, selling to CIOs, selling to chief data officers, selling to the line of business executives, understanding the application infrastructure, and becoming more consultative. You have much richer and deeper engagement on the things that matter, which translates into the revenue multipliers that we have there. On the partner side, it's all about empowerment, and I think I hit that one well enough.
Subscription. When you're moving toward a platform or portfolio motion, and you have a strategic advantage with our unified subscription model and a customer base that is increasingly wanting to buy as a service, formed largely by the success that they have had in buying services from AWS and Azure and GCP. The expectation starts to fold into the rest of their environment, whether it's making AWS ready for tier 1 in a Pure-like experience with our Cloud Block Store, or it's keeping it on-prem, and being able to move it back and forth. We want to meet the customers in how they want to buy, in the most efficient way they want to buy. That should drive velocity and differentiation for us. We don't want to force it though, and make it unnatural.
It's still reasonably early days, but it's pretty clear that it's going to be a meaningful chunk of the infrastructure that they want to buy over the next three to five years. I was talking to some of the Gartner analysts, and they're doing some primary research on this, but north of, I think there's an IDC report out that's talking about north of 58% of buying behavior preference in the future is to consume it as a service. This is a key area. You'll see some of the SaaS-like metrics and subscription-like metrics in Tim's presentation that we already have today. As that will evolve and as we announce new product offerings and as customers prefer to buy that way, we want to be packaging our motion in that way.
That means an assessment-led consultative sale, talking about people, process, technology, and architecture, that then says, "Hey, we can drive this through an API best-of-breed approach, and delivering you a modern data experience." You can use our Pure1 platform to manage it. Public and private cloud seamlessly across multiple clouds in the future. We're going to package a commercial offering that a petabyte's a petabyte. If you want to buy a petabyte from us, and you start it off on-prem, but as your cloud strategy evolves going forward, move that petabyte over to Cloud Block Store and have the economics of that work. It de-risks their environment, ensures the revenue stream comes with us, and the other folks can't do it. It really is a nice advantage.
We haven't put a stake in the ground relative to our long-term model that we shared on what percentage of our business looks like this, but we've got a lot more progress underneath the covers on this than I think we're getting credit for. Again, you'll see almost all of our products that we come out with will have a subscription element. This Pure as-a-Service offering will become our default go-to-market motion, going forward. This is not something that changes next quarter. We're not ripping out everything and doing it differently. It's an incremental approach based on how customers want to buy. It's unique and differentiated, compared to what our partners or our competitors can do technically and would want to do commercially. We realize we can't do this alone.
The business that we have with Cisco is incredibly meaningful to us and continues to grow nicely, that converged offering. Dheeraj is out talking about it's hyperconverged. VMware is talking about it's hyperconverged. We think what people want is simplicity and automation. If we can deliver that through partnerships with Cisco and with NVIDIA that drive efficiency and has lower total cost of operations that delivers similar simplicity, that ultimately is a buying requirement. We're very focused on continuing to evolve our relationships there. The data protection part of our business continues to thrive. Rapid Restore adds value, completely redefines it, from a disk-to-disk to tape, to flash-to-flash to cloud. You can see all of these folks on the floor with a lot of successes and enthusiasm underneath it.
We talked about in the last earnings call, both Splunk and Elastic, the log analytics and security processing is a natural fit for what our systems can do. Taking that to market helps them sell broader solutions to their customers and obviously is a win for us as well. Then what we're perhaps most excited about, although the revenue impact of this will be next year and the year after, but lighting up, like I said, the big three marketplaces and public cloud vendors, I think is going to be a really nice opportunity for us going forward. Underneath, we have routes to market that really map to the segments that we're focused on, and we're investing into those as well. Partnering here is strategic.
I think Charlie hit on this pretty solidly, and for those of you that were earlier, but being able to package a very simple message into our teams to be able to go repeat is really important as you move into a portfolio motion. You want to be use case driven and very simple in your articulation that resonates with economic buyers and senior executives as much as it does down at an administrator level. I'm really excited about this positioning and packaging. I think it's got a lot of runway for us as we look to evolve our product portfolio organically and inorganically in the future. It's a nice wrapper, and it's very different than what the other teams are doing. Lots of training and enablement on this.
It's not a significant departure of what we've been doing, but words matter and concise approach, I think, and precision matters, and we're excited about this, and you'll see us do a lot more around it. Ultimately, it's about making sure our customers are successful. To be both the innovator, which people know us for now, I think pretty clearly, but also the safe choice, really allows us to deliver a broader suite of products and offerings more strategically to the biggest buyers that are out there. I feel really, really good about the investments that we've made, the progress that we're seeing, and super excited about the product innovation that we're bringing to market over the next few years. With that, I know we're going to do Q&A later. Is it Tim or Matt?
Okay. Customer panel. Okay, good. All right. Thank you, guys.
Thank you.
Good afternoon. I'm Robson Grieve, and I'm CMO. Joined the company about 9 months ago, getting close to 10 months. This afternoon, I'm lucky enough to get to host our customer panel. I'm going to have each one of the attendees introduce themselves here in just a second, and we'll go through a few questions so we can get to know their experiences with Pure a little bit better and take advantage of this time. I made the sprint down the hallway here, so I'm definitely getting my steps today, as we all are in this giant convention center, but it's been a fun day. I would love it if each of you can start by just introducing yourself and your company and what you do.
Sure. Ludwig Gamache. I work for Element AI as the head of IT. Started there about two and a half years ago, when the company was quite small. We are now close to 500 employees, including 120-ish researchers, PhDs. We develop and help companies with their AI journey, in the sense where we'll help them from the early days, trying to have them understand what is AI and how they can benefit from it, up to ongoing support.
Excellent.
I'm Michael Wilkes. I'm the head of operations for SAP Sales Cloud. SAP Sales Cloud was a company that was purchased called CallidusCloud, by SAP. We do commissions for sales, we have lots of super large databases doing lots of IOPS and things of that nature. Just really heavy loads that Pure is pretty much handling all of.
Awesome.
All right. My name is Brian Graham. I work for a company called Nielsen. We do market research and measurement. I'm the head of global infrastructure. We have multiple data centers, a lot of Pure, and it's super helpful with the workload.
Awesome. I've got a few questions for you guys, and would love it if each of you can just jump in as you have thoughts about them. I'd love to start, Brian, with you and just hear a little bit about what a modern ops team looks like and how has the life of an operating group inside a big company changed over the years?
Yeah. It's changed in a very positive way. We used to be very reactive years back. Platforms like Pure and others, not just Pure, other platforms as well, are helping us be a bit more predictive. Allow us to get to issues before they bubble up into what we call sev-ones, and that means you have to get on the phone. It's awful. Because of that, I'm very happy with the direction the industry's going in Pure specifically. We're able to optimize our operations staff because if you don't have to react as much and you can be more planful, you have a smaller staff, which allows us to not necessarily reduce spend, but allows us to get more talented folks. That's basically, in a nutshell, the transition that we've gone through over the last five or six years.
Yeah. Have you guys gone through similar transitions in your operating teams, or have you been maybe, I guess, a little bit more born in the cloud kind of operations?
On our side, we're mostly on-prem. We do have products in the cloud and so on, but for the learning or in deep learning, in AI, we are on-prem, for many reasons. Because we were started about two and a half years ago, we started with the mindset that we want to be the shadow IT of IT, basically. We want to be the organization, the IT team that people will go see from the get-go and not then look at other options. We wanted to be very helpful, provide the best infrastructure possible, so that people would not try to look at other things. As soon as it bubbles up, it's really hard to manage.
Yeah.
Yeah, we went from being on-prem to making the transition to in the cloud. For us, as CallidusCloud was, we ran with really just a skeleton crew. We haven't really had to expand on that at all. We're literally managing nine petabytes of data with two people. It's been super good for us to actually transition to the cloud. Using Pure, we were able to do it very fast.
Yeah. You, Mike, your company has gone through, over your life cycle with Pure, a very significant commitment where you're essentially operating almost all Pure in your storage layer.
100%.
Talk to us about that decision and how you arrived at the idea that that was the right thing for you to do and what advantages that's brought.
Sure. My CEO at the time wanted nothing but the best. He wanted the fastest and unfortunately, he played golf with the EMC guys. We ended up with a bunch of VMAXs for some reason. They weren't working. They weren't working as they were supposed to. We gave Pure a try and over probably two years, we replaced all the VMAXs that we had with Pure arrays. I think we have a total of 27 of them now, running various different workloads, all of them outperforming what we're actually expecting to get out of them currently. It was a very fast transition, but it was based off of my CEO just wanting the best.
Yeah. Ludwig, in your context, you're doing a very intense AI work with our products. How did you do it before you found Pure? How would you do it without us? What would be your alternative? What advantage would just Pure have over that?
Well, that would be a big challenge, I would start with that. I don't know where I would go, to be honest with you. We tested many things. In the end, we have chosen Pure because it could really deliver what we needed. If you tell me you have to pick up something else, I would have nightmares. Yeah. When we started this whole journey, small company, small amount of processing power and so on. When I had the first discussion with Pure is I will be a bad customer because I cannot tell you what will be my workload, I cannot tell you how much data I need. I cannot tell you how fast I will grow. I don't know. We're a small startup.
I was giving a lot of problems to Pure in a sense, because I did not know, it was hard to know what to expect. It definitely lived up to our expectations because we're still using the same box. We scaled the amount of processing power that we have by 100 times or so. That's where we stand now.
Yeah. Brian and Mike, are you guys getting pulled into AI as a primary function that you're trying to support inside your development teams?
A little bit on our side.
Yeah.
My engineering team is in a different area for me, so I don't really see a whole lot of that on my side.
Yeah. Okay.
The AI strategy at Nielsen is being developed now. We're not getting pulled in yet. There's a lot going on, but right now it's still, like I said.
Yeah
trying to figure out how to hone that, and then at that point, there will be a lot more involvement.
Got it. Okay. We talk a lot about Evergreen and that Evergreen upgrade process, and you guys all have experience with products from other companies that we get compared to sometimes. Can you contrast your upgrade experience and your process of updating the product with Evergreen versus what you've had in the past?
On my side, I can't because we started.
Yeah. You started
the Evergreen version that we have.
Yeah.
We're in FlashBlade. What I can say is that in the past, I mean, upgrades have always been a huge challenge for all the organizations that I worked with.
Yeah.
Yeah, for us, the Evergreen program and the upgrade process that we go through is ridiculously easy. We just upgraded our X70s to X90s, and we didn't even take an outage in order to do it, which is just massive for us.
Yeah.
That's unheard of. We've never done that before. When we moved off of Dell EMC and onto this, it took us a year to get off, and it's not even getting onto Pure. Those machines literally sat there for a year just trying to decommission them.
Wow.
It's so easy. I don't want to think about doing anything else but that. I really don't like doing lift and shifts with the existing piece of equipment to another piece of equipment. It's a year-long process.
Yeah.
Unless you're with Pure, in which it's 15 minutes, and I didn't know it just happened.
The team came and told you they had done the upgrade?
Yeah.
That's great.
That's very similar to our process or our experience, I should say. The Evergreen, I was very skeptical of it when it was pitched. I'm like, "Yeah. We'll see.
Yeah.
It is exactly as advertised. There's no disruption. It would take a year to do a traditional migration. I guess traditional is Evergreen now. The legacy migration would take a year, and you'd have a dual run rate.
Yeah
support on one frame, and you just bought another frame. We had an emergency patch that Pure called us to apply, by the way, which is important to note because of your predictive analytics that you do on the support side. We were able to do it in the middle of the day. It was that important that we had to do it immediately, and we did it in the middle of the day successfully.
Yeah
without anybody knowing about it. The Evergreen program is very beneficial.
Yeah. The predictive support, one of our customers at Mercedes-AMG Petronas, the Formula 1 team, mentioned in an earlier session that sometimes you'd get a call from their shipping department that something new has arrived, and it will be a part from Pure that our predictive support has decided they would probably need it and has initiated that cycle. How often do you get predictive interactions with us where we are reaching out to you? You obviously have had at least that scenario.
Well, this is going to be a strange answer, but it's positive. I don't know. Now, the reason that's odd is because maybe I should, maybe I shouldn't, but the reality is, it's working so well. The parts are coming. I'm sure they're coming in. They're getting replaced.
Yeah.
Nothing has escalated, so I don't know how often it happens. That's what I'm telling you. The frames aren't failing because of what you're doing.
What you're doing is running smooth with my support staff, so I don't hear about it.
Yeah. Same here.
On my side, I know for a fact that I did not have any part that failed. I didn't have to replace anything. We did have calls about behavior at the software level.
The interesting thing is that the first call we received from Pure was like, "Okay, what did you guys do? You got more out of the box than our marketing paper. You did something that is even better." After that, all the other calls were mostly things to verify. There was no real failure or anything like that.
Got it. When you get pulled into the C-level strategy meetings, what are the technology topics that they want to talk about and that they're asking for your advice or feedback or they're thinking on?
On my side, I'm being pulled into a lot of I'm obviously a Pure supporter, but I'm being pulled into a lot of conversations, not necessarily with my C-levels, but with various different groups within SAP, to talk about my experiences with Pure so that we can look at different stacks for our HANA-based products and everything else. Also with Rackspace, I'm being pulled in on a weekly basis to talk about where I want us to go.
Yeah. Architecture discussions.
Yeah
big deal.
Absolutely. Yeah.
Yeah.
I'm laughing because basically what I'm asked is to make sure that it never goes to the execs. That's easy. Make it work.
Yeah.
Make it available. Make sure that our researchers can work. This is as much as we want to interfere with my work.
I mean, we're two and a half years old as a company. We are growing client acquisition and so on. The execs have things to do that are not necessarily IT related.
Yeah.
They ask me to make it work and make sure that everybody is happy about the infrastructure.
Yeah.
Most of the strategy I get pulled into is discussing cloud, how we're going to use it, how we're going to get there, things of that nature.
Yeah. Do you foresee cloud being a major theme for all of you going forward?
Oh, yeah.
Yeah. It's in our blood.
Yeah.
It's our whole business.
Yeah.
For different reasons on our side.
Yeah.
Mostly cost related.
Yeah.
For the deep learning part, where you train the models, I think we'll stay on prem. With the infrastructure that we have, it's very costly to go in the cloud. For the product deployment, that's another thing. Product deployment, that's really where the cloud will help us.
Yeah. Do you guys see a long-term hybrid future that you're going to be working through for years to come? As you described, there are certain things you just can't move for cost and performance reasons. Will that exist across all IT environments?
We have a cloud-first strategy. Time will tell.
Yeah
whether it's going to be a hybrid or an all-cloud strategy.
Yeah.
Right now, it's all cloud.
Yeah.
Yeah, we're on the same page. We have an initiative to get out of our data centers where we can, and use the hyperscalers where possible. It doesn't really work for our system.
Yeah.
Our databases are too big for it to be cost effective.
Yeah. Hybrid will be a part of your life for a while.
Quite a while.
Yeah. You guys all got to see our keynote this morning, I hope.
I know, Ludwig, you had a very good seat for it.
Yeah.
What thoughts or ideas were in there that you were most excited about and most interested in? Yours obviously was when Ludwig was on the stage.
Yeah.
I don't know. There were a lot of things. On my side, maybe what I'm looking at is the next generation of NAND chips, how it will be integrated into the Pure products. On our side, it's really performance that we need. We need to get the most out of the hardware. The next generation will help us either to drive the cost down or give us more processing power. That's maybe the thing I was the most interested in.
Yeah. Okay.
For me, it's the exact opposite. I'm super happy that you have the C Class coming out.
Yeah.
Not all of my workloads need to be fast.
Yeah.
In fact, a fair amount of my workloads, well, just don't need to be fast.
Yeah.
Right now, all of our workloads, no matter what it is, even if it's archived reports, are sitting on Pure Storage flash arrays. Getting that off of flash arrays and allowing that to become just database or just application level and slowing it down just a little bit really excites me.
Yeah.
The simple fact that we'll be able to do that's one of the things I've wished you guys have done for a long because the ease of use for the system, if you can take that same level of, I guess, well, ease of use, right? If you don't need to train somebody, if you can just plug it in, make it work, if you have the same level of support and Evergreen with just something slower, that's a huge win for us.
Awesome.
Yeah, I'm on the same page. I love the platform that we're using today, but it's overkill to some extent now. It reduces a lot of problems at the same time, but I don't think that the, I guess tier is not the right word, but the tier 2.
is still going to be faster than anything we've had anyhow.
Yes.
It's not going to matter much, and it will allow me to deploy more Pure and replace our aging fleet because the cost is going to go down. For the same amount of spend, which everybody likes to keep that run rate flat every year.
Right
The same amount of spend is going to translate to more devices for us.
More capacity.
More capacity. Yeah.
Yeah.
It's huge.
Yeah.
I'm also interested in the Pure1 Meta, and I can't remember the other name, although I was listening. Pure1 Forecast or something. What was the other?
VM Analytics.
No, we have VM Analytics-
Okay
which is good, but it was almost like a capacity planning.
Oh, yeah. Workload Planner.
Okay.
Workload Planner.
I'm interested about those two tools as well.
Yeah. Do you work with your development teams to try to plan capacity? How do you go through that? Is that-
We do try to work with them.
leave to you.
We do try to work with them. It's generally up to us to gather that data, look at trends, and see what we can do to help them.
Yeah.
They don't always know either when a new client's going to come on board. If we can just see trends going up, we'll prepare some purchases.
Yeah.
Yeah, that's basically how we are. We just deal with the trends. We find our engineering department doesn't really have a grasp on how big actual customers are.
Yeah.
So.
Yeah. Cloud, for you guys thinking about hybrid and your cloud transformation, how did that Cloud Block Store message from AWS and from us land, and how do you see that being a part of your transition?
For us, it is. We're looking at AWS in order for us to essentially spread out to regions where data privacy is an issue, and so we're not running into data being moved between countries and things along those lines. It's impossible for us to set up data centers in every single one of these countries that's doing this. Amazon already has. Being able to use the exact same workload as we do in our existing data centers and migrate that to a cloud-based environment works out really well for us.
Yeah.
For me, I'm going to have to dive into it a little bit and start sharing it.
Yeah
some of my peers and see what the feel is around adoption.
Yeah, got it. What big trends are on your mind? What are the big ideas or big things that you want to learn about this year, that you want your teams to be thinking about?
Trends in general about storage?
In technology, in your operating environment, in the way that you run your team.
Wow. Maybe because I'm in AI or the company I work for is in AI, I see a lot of things towards that direction. Operationally, on our side, this morning, I talked about the role-based access control system managed by a model. To me, this is something that will help IT in general, but there are many other areas where AI will be able to help us. Pure, from the get-go, started to use AI to do predictions and so on, which is great, but I think there will be more and more of those that we'll see, not necessarily only in storage, but in server management and so on. Even internally, there are teams that are ready to tackle some of the problems we have to see if there is a possibility to look at AI that can solve some of these problems.
Yeah.
I think there is a trend there.
Got it.
From our side, more or less, I'm looking for my teams to start doing less. What I mean by that is I want them to do more, but I want them to do it more efficiently.
Yeah.
We're always pushing our teams to do more with less. That's not really technology-based, but it's the reality that we live in, trying to maintain all of the workloads that are coming through as the business grows without actually growing headcount.
Yeah.
From my point of view, what helps us out with that is all the tools that you guys give us. Like I said earlier, I'm doing nine petabytes with two people.
Incredible
that's hard to do.
Yeah.
Yeah, mine's along the same lines. I want to do less so they can focus on advancing the environment rather than maintaining it. The trends that we've seen this past year is more API-driven operations. You always hear about API and DevOps and all that.
Yeah
There's a place for the API-driven legacy infrastructure operations, if you will. That's where I really want to see some increase this year.
Yeah.
Yeah, that's good.
The concept of experience, we obviously talked a lot about that in our keynote. It seems to feed into or be a really good match with where you guys are in terms of trying to do more with less and trying to have your team focus more on the innovation parts of the business. Is that thought landing with you in terms of Pure being something that can be a leverage point as you go that direction?
Absolutely.
Yeah.
Yeah.
You have been so far.
Yeah.
It's just getting better based on what I've heard today.
Yeah.
Great.
I have the same experience. You guys have been a great partner for us, going above and beyond, and you're always pushing us to be better, too.
Yeah, I would relate to that. To me, the relation with Pure is really a partnership. It's not a sales experience, it's a partnership experience. It's completely different than other transactions, and as I said it early on, because we did not necessarily know a lot in the early days, so it's very important for us to find a partner, not somebody that would sell us a box.
Yeah. Okay, one last question for you guys. Our Pure as-a-Service offering or subscription offering, is that something that you're going to look to use as you expand or as you think about a more dynamic cloud-driven kind of operation? How do you see that fitting into your procurement model for our category?
My sales guy's going to be in trouble because he hasn't told me that.
Yeah, mine did. I think it has a place. I couldn't tell you where it is right now, but there's a lot of flexibility around you may need to do some restructuring in order to move to the cloud, which is our strategy. You don't want to invest a bunch of capital into something. Maybe you can use the service to get you over that hump and then just not use the service any longer. That's what's exciting to me. I think that's where the opportunity is for us.
Yeah.
On my side, we had already started the discussions around that. I know there are possibilities. It's great because there are different options. With our sales rep, we're looking at all of those for many reasons. We could go in many directions, including still purchasing the FlashBlade. The OpEx version of it's interesting. It's something that we'll definitely look at.
Yeah. What % of your IT budget is in subscription products now versus in CapEx?
I can't tell you the percentage. It's just more. It's significantly more.
Growing.
I said cloud first, but it seems like there's a lot of SaaS first, subscription first.
Yeah
approaches to everything now.
Yeah.
Yeah, we're pretty much the same way. I'd say that I'm probably 50/50 CapEx versus OpEx, just with various different systems.
Yeah.
I think we're more CapEx heavy still because we're on-prem and so on. The cloud for us is not really an option for the learning part. In terms of percentage, I would have a hard time to do this. I would say three-quarter, one quarter, but I'm not entirely sure. I would have to run some numbers.
Yeah.
What I believe at this point.
Awesome. All right. Well, thank you. You guys have been great. Do we want to take a minute, take any questions, Matt, or are we good, or we have time for now?
Probably we should wrap up.
Okay, awesome. Thank you. Appreciate it.
Thank you.
Thank you.
As we're transitioning here, I'd like to thank Michael, Brian, and Ludwig for their commentary and taking the time to speak to the investor community. Such a powerful statement when you have customers speak about what they're seeing with Pure. We get the question a lot, "What makes Pure different? You're selling storage arrays just like sort of your competitors. What is different about Pure?" I think you heard today, it's about the efficiency, it's about the ease of use, it's about the ability to upgrade in a day versus a year from our competition. Again, it's something that I and the rest of the management team talk a lot about as we spend time with the investment community.
To hear that from our customers, I hope you feel that it is something that's real, and one of the reasons we've been successful as we have, from a company perspective. I am the last guy for the day or the last presentation for the day. Hopefully, you've enjoyed the day. Hopefully, you've enjoyed hearing from Charlie in terms of his vision about where we have been over the last 10 years, but more importantly, where we're going for the next 10 years. It was important to us to give you a flavor of our general managers in terms of the products they're building, and obviously a lot of excitement both in our FlashArray product as well as our FlashBlade product. Just as importantly, our software products in terms of how it all fits together. Hopefully that was helpful.
As always, great to hear Hat in terms of his discussion from a go-to-market perspective. Again, just as importantly, listening to the voice of our customer, which we do all day, every day at the company. My name is Tim Riitters. I am the CFO here at Pure. For this last session today, I've got about 20 minutes. I have just 3 key topics. We're going to do a very, very quick overview and flyby of our Q2 results. We reported results about 3 weeks ago. I'm going to spend the bulk of my time talking about what we really talk about and call Pure being positioned for success. It's what we almost think about as an investment thesis as we're talking to investors in our travels.
Finally, I'm going to do a little bit of discussion around some modeling things and some just technical updates to think about as you start tuning and thinking about next year's financial model. Again, very quickly, this is going to go really quick. We have one slide. You guys have all seen these results before, I think it's worth sort of reviewing them one more time. You know the 28% top-line growth, you know being essentially right around non-GAAP profitability in a still seasonally lower part of our year. We generated positive free cash flow as well as are growing our deferred revenue balances. There are two other metrics I'd like to sort of make sure that we don't lose sight of as we think about Q2, and then obviously, for the rest of the year as well.
28% revenue growth, yes, 41% revenue growth in our support and subscription business. You're going to hear me talk a lot more about that as we get into the next series of slides. As you heard from many of our speakers today, delivering products in a software-centric, a subscription-centric manner is something that we definitely are thinking about here and focusing on at Everpure. Something I'm very proud of, that you're seeing those results already in terms of the 41% year-on-year growth. Gross margin. We've been talking about gross margins at the company and how we are really differentiated from our peers in the industry, somewhere between a 10-15-point spread relative to the competition. It's how we innovate. It's the technology. It's the software at its core.
It's also the value and the differentiation that our customers put in that product that allows us to deliver those results. Again, I'm pleased to say that this last quarter, we put up all-time high gross margins on a non-GAAP basis for the company, almost 69.5%. Again, a lot of good sort of results in Q2, but a couple of key ones that I want to make sure that we continue to emphasize and reinforce as you think about Everpure. All right. You saw this earlier today from Charlie, again, what we're terming Everpure positioned for success. This document, these five tenets, really were established by us at the start of this year as we thought a little bit more about why would investors want to follow us along on the journey over the next five and even 10 years.
It really comes down to these fundamental tenets that, as I said, Charlie articulated early on in his presentation. I'm going to spend a little bit more time in depth talking about each one of these items. Whether it's chasing after a large and very attractive, lucrative market, whether it's taking share consistently in that market, whether it's winning and innovating in the cloud, expanding our subscription-centric business model, and doing it all with a compelling financial profile. Hopefully, as I walk you through the next series of slides, you'll agree that Everpure is really a special asset to be thinking about in the years ahead. First of all, large market opportunity. You've seen these slides before.
We went public, we declared a $24 billion TAM. Over the last several years, as we've added more features and capabilities in the business, we've raised that TAM to $50 billion. As you heard today, we've launched a whole host of new products across both our FlashArray and FlashBlade business, as well as, again, the software businesses or software products that I talked about a little bit earlier. These new products we're very excited about. They allow us to capture just that much more of this TAM and doing it at speed. Again, as you heard, many of these are GA and selling already today. We're excited about it. It's a $50 billion market. That is a lot of opportunity.
The other thing I'd make sure that we sort of emphasize on this slide is that we're focusing on what we call the markets that matter here at Pure. A $50 billion TAM, there's a lot of different things within that TAM. If you look at where a lot of the growth, we believe as a company is going to happen over the course of the next several years, it's in those markets that matter down in the lower right-hand corner. Obviously, the largest enterprises are going to continue to be voracious users of data, which bode well for us to the extent we succeed in getting into those markets, as you heard Hat talk about a little bit earlier today. Cloud and SaaS businesses, those businesses, Cloud and SaaS is where the economy is going for the future.
To be a supplier and a valued and preferred supplier to many of the cloud companies. You heard CallidusCloud, obviously, up on stage here. A great success story, 100% Pure, which again speaks to the sort of differentiation that we've demonstrated in these new technology stacks. Obviously, our win rates or deal sizes are much better in those markets as well, as you heard Hat talk about a little bit earlier. Finally, next generation AI and analytics, Element AI. Element AI on stage, one of many new AI companies that we've had the pleasure of working with. These companies see the tremendous benefit and the tremendous differentiation that Pure delivers to their business. The second tenet, growing and taking share. I think you've probably seen this slide three or four or five times today, so my apologies again.
I wanted to bring it all together and all home. I mean, we get this question a lot. Pure, you've been at this for 10 years, is it in a matter of time until someone catches up? I hope you feel that today, that innovation lead is widening and that differentiation is widening. Again, you've heard it from our customers that were up here just a few moments ago. For those of you that have been following the company for a while, we talk about differentiation not only at the innovation layer, but also at the business model layer and taking care of our customers. We like the moat that we have as a company, because if you're going to compete against us, it's important that you have to compete against all three of these areas.
You have to be prepared to disrupt your business model. You have to be prepared to significantly innovate on a technological level, and you got to take care of your customers. As Hat, I think, said earlier, with an NPS of 86.5, 86.6 versus the rest of the competition at roughly 20% or roughly 20 points, it's no secret why we've delighted customers and won and grown the way we have. You can see it in the numbers. Moving from, think about this, a decade ago, Everpure wasn't even a company. We weren't really even selling. In less than 10 years, we're on track to be the fourth largest storage provider in the world. Again, really phenomenal track record that we are particularly proud of at the company. We're doing it consistently taking share and growing faster.
I won't sort of belabor this graph on the right other than to say 28% when everybody else is shrinking speaks to the fact that the formula is working and the recipe is working. All right, subscription progression. We get this question a lot. We get this question a lot about what is sort of Pure's plan to sort of think about the new world that we live in terms of what's happening with software and what's happening with services and subscription. You heard Hat talk a little bit about it earlier. He talked about serving products and solutions in a way that our customers want to consume them. I affectionately call this slide the have it your way slide. Our goal at Pure is not to drive any specific way of consumption.
Rather, it's to have the offerings and the products and the portfolios and the capabilities to serve people if they're CapEx centric, if they're utility-based centric, if they're software-based centric. I want to take the next sort of build to talk a little bit about how that's evolved over the last, call it two or three years. Obviously, we got going early in days at Everpure delivering a software-enabled product via a platform, via infrastructure. That's what I would call our traditional business model. It's what we've been doing the lion's share of for several years now. Again, as I think one of our GMs talked about, 90% of our engineers are software engineers. We've always thought about our company as a software company, and yet we needed a platform and a form factor in which to sort of deliver it.
In the early days, our own sort of hardware made sense from a delivery perspective. Over time, we've moved to subscription. For those of you that were in the room or listened to the last year's analyst call, last year was the announcement of our ES2 subscription business and what we're rebranding today as Pure as-a-Service. It was a recognition in listening to our customers that people wanted the same capabilities that we deliver in our FlashArray//X and FlashBlade and FlashArray//C products, but doing it in an as you want to consume. Almost think about it as a utility-based business, a utility-based product. That's, again, what we launched a year ago. This year at Accelerate, we've launched a whole host of multi-cloud capabilities.
I think Matt Burr said it best where the cloud is just another hardware target. That's the way we've thought about this as we've developed, is that certainly, and as you heard on stage from our customers, the cloud is obviously real. It's something that's on our minds of many of the people that are managing storage today. It's important that we had software layers that allowed us to ensure that we could sort of go against that target, and you saw a lot of announcements today about that. Over time, we get the question software only.
While we don't have a stated plan today to do a software-only overall product and sort of engineer against the myriad of combination of white boxes that are out there, we consistently have conversations with the Global 50 and other large data consumers about opportunities to deliver our software on top of their own corporate standard from a hardware perspective. Again, those conversations have taken place in the past, they take place today, and they'll take place in the future. Again, as you look left to right, I think again, the key point before I go to the next slide is really think about Pure is as you want. Have it your way. We will continue to sell products across all four of these columns or silos over time. Okay. Evergreen strategy is working.
Again, I'll refer back to our colleagues or our friends here from the various companies you just heard about. They all had an opinion on Evergreen. They all talked about how profoundly different it was in their business to be able to do a non-disruptive upgrade in a day and not a year. To do it and have the confidence that they don't have to take all of their business down while they do these upgrades. That has been a fundamental tenet of who we've been as a company since we got started, again, almost a decade ago. That strategy is working. For those of you that have listened to the last couple of earnings calls, it's been about 3 or 4 quarters now where we've called out the increase in deferred revenue, and in fact, the acceleration in deferred revenue.
Culminating in an overall deferred revenue growth of nearly 50% this last quarter. Why is that? That is really primarily a function of our customers understanding and recognizing that the Evergreen promise is working. They're continuing to renew and upgrade and update, and you'll hear me talk a little bit about that in the next couple of slides, at a very nice and a very strong rate. They're buying longer-term subscription contracts because they see the value in Evergreen. They've probably already experienced another array that they've bought. Then finally, some of these new as-a-service and as-a-utility offerings are coming to bear. When you put all of that together, the leading indicator of how we're doing along that dimension has manifested itself in the deferred revenue balance.
We get more questions, say, "I understand that, but give me a sense of what percentage of your business is recurring, what percentage of your business is subscription." What you're going to see in the next part of this slide, and then in the subsequent slides, is a little bit more color and context of where that is. Investors can appreciate the size and magnitude of our recurring business model. Next slide over here, is our progression over the last two years of our billings, the percentage of our billings that are subscription or support billings. As you can see, over the last two years, we've improved that ratio by 10 points. Again, I don't think people fully appreciate that a third, fully a third today of our billings as of the Q2 that we just closed, are subscription and recurring revenue.
Obviously, you're seeing that sort of manifest itself in the deferred balance. I think it's really important to get a perspective of that. Obviously, we believe over time, that has the ability to continue to increase, particularly as we offer more and more of these software and subscription licenses, and do more business along those lines. Again, wanted to give another proof point in terms of where we're headed and the nature of the business to sort of talk about the Evergreen strategy working. All right. Also about Evergreen. We took a step back, and we get a lot of questions from investors, candidly, and the analyst community, the product analyst community, think about the Gartner of the world, if you will, that ask a lot about how do I think about your business as a SaaS business, right?
You talk about Evergreen, you talk about high retention rates, you talk about this sort of subscription, you talk about stickiness of revenue, you talk about customer retention. Can you give me a little bit of more proof points so that I can think about the company in more of a SaaS-like manner? As we did our research and we did our homework, four key metrics came out that really are popular in the SaaS lexicon, if you will, today. Obviously, as people that follow a variety of companies, I'm sure you've seen these numbers in some of the other companies you cover. Specifically, cohort ACV growth. Taking a company or a set of customers that were acquired in one year and understanding over time how that ACV has grown over time, how that subscription residual and stickiness has stuck over multiple years.
Customer retention. How many people are here today of this customer base that were here a year ago? How many people have you lost over time? What are your statistics there? My favorite, net dollar retention. Very common sort of statistic in the SaaS world. What looks at overall growth year-on-year in ACV normalized for a cohort from last year. Think about all the customers you had at the end of last year, how much ACV did they have then? And how much ACV do they have a year forward? Very common SaaS measure as well. Finally, a sense of scale, a sense of size. Contract value under management. How many dollars of ACV or TCV exist in the business to sort of credential the size and scope of the business that we're talking about?
Again, these were the ones that kept coming back as we did our research and did our study in terms of how investors think about SaaS businesses. We want to give a flavor of where we stand from Everpure's perspective. We're going to go 1, 2, 3, 4 in the same order of the metrics I just showed you. Number 1, ACV by customer segment. We've been tracking about 5 or 6 ACV segments for the last several years now. If you see other SaaS companies, you'll see this classic sort of waterfall build schedule. What I'll say here is that you can see these are average annual increases in ACV. This is not something that goes from year 1 to year 6 and grows 50%. This is something that's growing in the third sort of stack, for example, 50% on average every year.
Very, very attractive economics relative to many of the SaaS companies that we take a look at in terms of building that ACV cohort. Again, what is this a statement on? Unlike SaaS companies that are selling maybe by the seat, we're selling by the array, or we're selling by the terabyte. As you heard up on stage from our customers today, once they see the technology and see what it can do, you can believe that they're going to put that in more and more and more of their workloads. You're seeing that as they buy more product and put more subscription under contract. Very, very nice average annual growth rates in ACV. Customer retention, 97%. Very, very attractive number for us as well. Again, this simply measures year-over-year change in terms of overall customer churn in terms of number of customers in the business.
Again, as I said earlier, my favorite one, net dollar retention, 140%. What does that mean? If I had $100 of ACV for customers that were with me a year ago, you fast-forward a year, that same customer group now has $140 of ACV under management versus 100. The reason I say that this is one of my favorite ones is that if you look at the SaaS universe, this puts us probably somewhere in the top third to the top quartile of excellent SaaS companies in terms of their net dollar retention on an ACV basis. It compares favorably to companies like Zoom. It compares favorably to companies like Okta, Atlassian. Again, really strong measure of the success that we've had in terms of landing and expanding and retaining customers for the long haul.
As I said earlier, the last measure we'll offer up is kind of, okay, give me a sense of how big this business is. These are all great statistics, are you operating at say, scale? The reality is we are operating at a total contract value, or TCV, of in excess of $1 billion. In other words, we have $1 billion of contract under management that is active right now in terms of the support and subscription business. To have that value, that sizable value, and growing and increasing on these NDR basis or on these increases in ACV, it's a very healthy business for us, and we're excited about it. Before we go to the next slide, I'd like to sort of pause and just step back.
We were talking about this as a management team as we were putting this slide together. If you took the Pure Storage logo off and just put this down on a table and looked at it, you'd say, "Wow, it's a very attractive SaaS business, a high-growth SaaS business that's doing some very, very powerful things," right, that we feel proud about, that you can compare to very good quality names in the SaaS space that you're all familiar with. Obviously, it gives us great pleasure to share this additional insight, to talk about what's going on in the SaaS, in our metrics that look a lot SaaS-like, very frankly, and are on a range that, as I said, put us somewhere between the top quartile, somewhere between the third and the top quartile of all SaaS businesses. Hopefully, you can appreciate that the Evergreen strategy is working.
It's what we set out to do many years ago. We're very pleased with the progress and the momentum that we've seen, and you can see the statistics here. All right, last tenet that we talked about in terms of the top five in our investment thesis, compelling financial profile. I want to be careful, this is not a guidance exercise. Typically, we don't offer guidance in these venues, and that's no different in this venue today. We did want to step back and just understand where we are on our journey from $1 billion to over $2 billion. For those of you that were at our Analyst Day a few years back now, we crafted a trajectory as a team to grow from $1 billion to over $2 billion over the span of about four years.
Again, as you can see on the left-hand side, we're tracking very nicely to that. Really right on target here for the year three of that journey. We've been doing it in a very sort of measured and principled manner from an operating profit perspective. Again, for those of you that have been along for a while, Everpure, as it was growing, it was scaling substantially, was in a net operating loss mode. That was okay. We understood that it was important to invest. We wanted to demonstrate to the investment community that we are being principled and thoughtful about how we are driving sort of profitability in the business.
A couple of years back, we achieved profitability, or sorry, a year back, we achieved full-year profitability and tracking very, very nicely in that range again and sort of driving leverage in the business. Again, last part of the financial thesis, if you will, really doing all the things that we talked about earlier, but doing it with a very compelling financial profile, which we're proud about as a company. All right. One more thing before I get to the end, talking about modeling things. I alluded to it at the beginning of my presentation. I'll talk about it more deeply here in this slide. Is gross margins. Roughly running somewhere between 10-15 points higher than our competition, and really have been differentiated and operating at our long-term model for quite some time now.
Our long-term model, 63%-68%, as you can see here, we've been operating at the midpoint or above for all of that period, 13 quarters. In fact, if you continue to bring the tape backwards, if you will, it'd be 15 quarters. You'd have to go back 15 quarters before you saw us under 65% from a gross margin perspective. It's time to raise the gross margin level for the business. It's time to raise that operating target between 65% and 70%. Moving it up roughly about two points from a midpoint perspective.
Again, we feel comfortable as a management team that given the innovation that we're driving, given our approach to how we software engineer against all of the oncoming classes of NAND and memory that's coming, in addition to moving to more of a software-centric business, which by definition is stronger gross margins, really a positive signal to the investment community that we think about ourselves differently. We've demonstrated that we can operate in this range, and we're excited about raising that range over time as we think about the business going forward. Last slide on long-term operating target model. As you can see, we've raised the gross margin targets. We have not raised the R&D, S&M, and G&A targets. We are continuing to hold those where they're at. There's a significant opportunity out in front of us. You heard Charlie earlier today talk about the next decade.
We want to keep the targets where they are at this point in time. It gives us a little bit more room to operate as we continue to grow the business. Obviously, for those of you aware, I think everybody's aware here, I'm leaving Pure in the fall. I want to make sure that we give leeway to our new CFO coming in terms of how they think about the business. The management team is very committed to continue to drive profitability, leverage in the business as we grow a multibillion-dollar business. All right. That concludes the second part of my presentation. The last part of my presentation is some details from a modeling perspective.
Again, these are in the weeds, if you will, but I think it's important to start talking about them now so you can start thinking about how, if relevant, it's time to tune any of the models as you think about next year. First of all, moving to a 13-week fiscal quarter. It's quite common in the infrastructure industry to move to 13-week quarters. The reason you do that as a business scales is you like consistency, you like to de-risk anything that goes on in the quarter. With that consistency, when you can always end a quarter on the same day, quarter after quarter and year after year, it just brings a whole sense of new predictability and structure to the business. Again, companies like Cisco do this, for example.
We felt it was the right time to do it, particularly for this year, because the time between our old fiscal year, Friday, January 31st, and our new fiscal year, and Sunday, February 2nd, was close enough, and there were really no business days between the two, that it was time to make that move now. What that means is that we will end our fiscal year that we're in, fiscal year 2020, as of February 2nd, 2020, versus Friday, January 31st. Again, because those are not business days, we anticipate no changes in our financials, so the full-year guidance for the year remains unchanged. Going forward, every quarter and every year-end will end on a Sunday from here forward. Okay? Second modeling item, non-GAAP tax rates. We anticipate the non-GAAP tax rate starting next year will be somewhere between 19% and 21%.
To give you a little bit of backstory on that, we've been operating, when you look at when it's time to release your valuation allowance and start recording taxes, you look at your trailing 12-month performance over time, and once you get to a point where on a non-GAAP basis you're profitable for a period of time on a trailing 12-month basis, it's time to release that reserve from a non-GAAP tax basis and start recording expenses. You're going to see that happen here at the start of the year. That's effective for our Q1 and going forward. We will do that on what's called a static rate. There's a couple of different ways you can employ tax calculations. I personally like the static rate. I think the investment community does as well because it's just as the name implies, it's a static.
We establish a rate for a medium-term time, could be anywhere between 3 and 5 years, and we continue to record at that rate, subject to no significant changes in the business. We're going to do that. We will update the investment community if anything changes from a rate perspective because we're in the process of narrowing down exactly our rate. Once we get going, you can count on a static rate on a non-GAAP basis quarter in and quarter out. Last couple of things I'll say on the tax rate.
Number one, this does not mean we are a cash taxpayer in the U.S. For those of you doing DCF modelings, just be careful because we'll see a tax expense on a non-GAAP basis, but we won't have cash taxes in your DCF. We will also not be recording taxes on a GAAP basis for quite some time because we still get the SBC deduction as well. Again, some modeling trivia in these two items, but I think super important as you think about the next year and you start getting your models and estimates going forward and start building those out. That's it on the third part. I'll wrap where we started. I'll wrap where Charlie started. Again, first of all, a big thank you. Thank you for attending today. Thank you for hearing and listening to what we have to say.
Thank you for supporting the company. Hopefully, you heard today about the innovation we're driving, the significant innovation we're driving, that we've driven in the past and will continue to drive, the momentum we're seeing from a go-to-market perspective, and I think just as importantly, the excitement and the true passion you hear from our customers who, once they use this technology, they buy more and more and more, which you're starting to see in our metrics. Again, we're excited about the future, and we'll turn it over now to Q&A. Yes.
Hello again. Sure. Who's got the mics?
Yeah. Do you want to grab the mics?
No, for the audience.
For the audience. There's one here.
You are?
Yeah, I'm wired. Yeah.
Hey, good afternoon.
Mic.
Hey, good afternoon, Karl Ackerman from Cowen. Tim, if I could just go back to the financial outlook for a moment. I think your implied guide for calendar 2020 or fiscal 2021 was high teens revenue growth, over $2 billion from the $1.7 billion plus today, plus 6%-10% operating profit, so 8% at the midpoint. I guess I'm curious as how you think about the industry growth rate or the TAM growth rate for storage in calendar 2020. What do you think would drive your OpEx toward the lower high end? What would happen to OpEx that would drive you toward the lower higher end of that operating margin range?
Yeah. Maybe I'll step in, and then I need you, Charlie-
Yeah, sure.
if you want to give your perspective as well. I want to be very careful. Again, yes, we put the slide up in terms of our momentum in the business, and I wanted to be very deliberate saying we're not specifically offering guidance. We're talking about our trajectory. We're excited about our opportunity to grow bigger than, go beyond $2 billion. I wouldn't take that as a specific guide, specifically. I know obviously there are consensus numbers out there. I want to just put that caveat in to begin with. Then Charlie, I don't know if you've got perspective on investments and where we would operate within a range.
Right. We've been fairly consistent, that to the extent that we see medium to long-term growth being in the 30-plus percent range, that we would continue investing both in sales and in R&D to drive that higher growth. Personally, I'm growth oriented. I'm growth oriented. That's what I want to see in the business. As companies do grow, if we start to see that our medium to long-term potential is less than that, then obviously we'll use leverage in the business to drive more to the bottom line. With 70% gross margins or 65%-70% gross margins, we do have a lot of leverage in the business. Growth costs us a lot in this market, as you can see from our sales and marketing number. We exist in a very competitive market with existing incumbents, so it's expensive to compete with them.
Once growth slows down, then you don't have that kind of growth that you're investing in anymore. We do have a lot of leverage in the business, but in the meantime, we're investing to grow.
One more, if I may. Matt or Hat for you. On FlashArray//C, the workload opportunity is significant, I think, for tier 2 workloads. Those continue to be run on 5,400, 7,200 hard drives today. Given the prior commentary today that indicates enterprise customer spend has been moving at a perhaps slower pace of adoption, I'm curious how you think about the adoption of this new offering. Particularly, also how we as analysts try and compare this from a raw and effective cost per terabyte or per petabyte basis. Thanks.
Yeah, just to clarify on the enterprise adoption, the 21% growth was new customer acquisition, not actual bookings growth or revenue growth. We still have great purchase metrics. You see the ASPs that we offered up are larger in that segment. We feel great about the growth rate of our enterprise business. I just want to clarify that, and I'll have Keith talk about the other.
Yeah, I think we believe the opportunity for this product is massive. Whenever you bring something, a capability that's new to the market that no one's had before, you're just essentially selling into the greenfield opportunity. In this case, we're selling something that everybody knows they want. They love Flash. They just haven't been able to afford it at this new tier of applications. We're really bringing it down to this other tier of applications. We've tried to be really thoughtful about creating two very distinct products between X and C that have fundamentally different performance and cost points, so they don't really overlap with one another. They really serve different needs. I think we're very bullish about it.
Alex?
Oh, sorry. Go.
Alex Kurz with KeyBank. On quotas for next year, when you start thinking about fiscal 2021, and with the new product out, right, the RAC. I would think that most reps would feel like they can get to the numbers faster now with this product, assuming that pricing in the market can match what a tier 2, tier 3 storage array can price out at. I wonder from a quota-setting perspective, you got FlashBlade now. In different use cases, we've got the C, and we've got the core product. What would that quota look like next year versus what it was the last couple of years?
Yeah. We always want to create quotas that reps can hit. You want to try to drive toward a participation where there are data-driven quotas. It's not just peanut butter. A lot of other legacy providers will actually just say, "Hey, take the number that you had last year, add 15% or 20% to it and you're good." We've never done that. We look at the specific market opportunity in any given territory, based upon what opportunities are in the pipeline and what we anticipate what price we can sell, and we build a bottoms-up quota. With the goal toward everybody having an equal shot at getting to their plan. That drives morale, that drives productivity, et cetera. Philosophically, mathematically, we've always approached it that way. Next year will be no different. Obviously, putting more products in their bag gives them a heck of a lot of confidence.
They have customers, we have customers asking us every single day to get into tier 2, asking us to get into the cloud. A lot of the innovations that we've brought to market are in the wheelhouse of where customers are already asking for help. I think that as we build the territories and the quotas, it's more about narrowing the span of control, and focusing the investments in the segments that matter, and then creating a data-driven quota based upon the product opportunity.
For the bell curve Pure rep, could this be a third of their book?
At a normalized rate, the C array?
My view is it's too early to talk about that.
Okay
To think about that. We've got big opportunities still in FlashArray and FlashBlade as well.
Yeah. I'll just repeat. Having introduced a lot of products over time, even though I joke with our sales force, do you think you can sell a cheaper FlashArray? Because that's what it is, right? Every new product has its own learning curve as you bring it to market. Given that this is just going GA, we'll have our learning curve to go through. Yeah, we'll know more in a couple of quarters.
Yeah. Hey, it's Jason Ader with William Blair. Kix, did you answer the effective price per gigabyte? I didn't hear you answer that. Maybe I missed it, for the C.
We haven't put a market price per gig out, just like we don't really publish one for X either. When you think about the difference between the two, you should think about C as roughly 40% cheaper and going from there in the first introduction. The goal was to establish fundamentally different price points. If it was only 5% or 10% difference, there's no point in having another product line. The other thing I would also just highlight is that C is pretty big, and that was intentional, right? We want to go after large customers and large-scale consolidation deals, not to sell little 30K, 50K systems. The smallest C is a petabyte usable and will be many hundreds of thousands of dollars, and so this is something that's more in the sweet spot of our enterprise sales force.
Okay, just following up on that, I think you talked about it in the keynote maybe, but which are the top workloads that you would call out for C?
Yeah. We see several workloads. There's just the consolidation of tier 2, in the classic VMware environment. If you think about a customer that went in and implemented VMware, they probably virtualized their less important applications before their higher-end applications, right? There's a lot of VMware we don't play under today, that today might go to a second-tier disk array or even in a hyper-converged environment. That opens those up for us. There's second-tier DR, where we can go in and provide this as a replication target for DR use cases and allow more applications to afford DR. There's test dev, and there's a really great snapshot integration story, so we can sell into the test dev use case, but then refresh data from FlashArray//X's every day. You're always testing with the most current data.
Lots of different use cases and then larger scale, lower cost analytics type use cases, Splunk, things of that nature too.
Could we get a mic over this side? Oh, you got it. Great.
I got the mic. Simon Leopold with Raymond James. I wanted to see if we could talk a little bit about how, as you're pursuing more of the Global 2000s and moving upmarket, what that introduces in terms of change for the company. The two variables I'm thinking about, one is you're competing against customers who have broader portfolios, and I wonder about, do those customers seek broader portfolios? It's part one. Part two is I think about enterprise spending in the most recent quarter. The big IT-exposed suppliers talked about the big multinational companies as softening. If you had less exposure, that was a good thing. Maybe if you could talk about how this is changing the business dynamics. Thanks.
Okay. Yeah. A, they do want to buy more products from a single supplier. They're not looking for more vendors. That creates an opportunity and a challenge for us. The opportunity is we're best of breed investing in this and doing it dramatically better than everybody else, right? We've had a lot of success over the last several years of landing our systems, whether they're FlashBlade or FlashStack or FlashArray, in those G2K customers. It's still a significant portion of our business. They've been asking for us to expand our portfolio. The opportunity with FlashArray C and the other products that we have, I think, does fit into their desire.
One of the reasons why we built the company and invested what we did in distribution is that we believe that our business practices were going to be so much better than the legacy planned obsolescence model of tech refreshes, that we'd be able to land, and then we'd be able to expand. I think all the repeat purchase metrics support that. That's why I'm excited about the portfolio expansion because our sellers can deliver that, and it's what customers want to buy, particularly in the large enterprise. Relative to softening that the other vendors are talking to, it's not super positive, it's not super negative. It really is kind of middle of the road, and that's been consistent for us.
We did mention a couple of quarters ago that the shape of our funnel is changing, and you've seen the number of cumulative $5 million customers and $1 million customers, et cetera. Those do have different characteristics in any economy. They just get higher exposure, higher visibility. There's more work to do to get them across the line. Not as systematic as a commercial motion, where the math is just really, really predictable. We want those larger transactions. These are eight-figure deals. These are meaningful, but they're harder to call. I think it did affect how we've managed the business going forward from an instrumentation perspective and guided. I think it's middle of the road. We're not seeing it as a net negative or pressure in our business, or a net positive, but stable.
Great. Aaron Rakers with Wells Fargo. Two questions, if I can, real quick as well. It's back on the portfolio motion and how we think about the timing of some of this stuff. The first question is with regard to the C series product. I'm a little bit confused because there's not much availability of QLC NAND out in the market today. Are you selling a product that actually has TLC on it at first?
Do you expect a bigger customer adoption, Aaron, or should we really think about quad-level cell availability as being the determinant of opening up the market opportunity? If so, when does that actually happen in your mind? Yeah, I'll speak to that. If you look at the architecture on FlashArray//C, what we launched was our QLC-optimized product end-to-end. To do that, we built a capacity-optimized direct flash module. That direct flash module actually can take both QLC and TLC flash. Yeah. I showed on stage one of the QLC modules. We're actively working towards shipping that. We wanted to get going in the market as fast as we could, so we're making the first versions of C available with TLC flash. We've changed the direct flash software to optimize the performance, so it performs exactly like QLC.
We're going to introduce it at the price point of QLC. Essentially, the message to the customer is you're buying the QLC product of next year. We may choose to ship some TLC flash to get it to market quickly, and then sometime next year, we're going to just kind of transparently move over to QLC. May not even announce it. Right? Okay, perfect. The second question on the product portfolio, I don't think it was asked. DirectMemory Module. I think you can deploy those in either four or eight bundles in an array. You guys collect a ton of analytics on your customer base, your install base. How much of your customer base do you actually think you can go in now, sell into those products? Can you give me any framework of, let's just say, four bundles?
What's the cost of that? What's the incremental ability to sell into that install base? Yeah. That turns out to be a question we can answer very precisely with our Pure1 data and the meta-analytics engine. We talked about that a bit on stage today, but we've gone through and essentially analyzed the global population of X70, X90 arrays. 80% of those arrays would get at least a 20% latency reduction by adding in direct memory. 40% of those arrays get between a 30% and 50% latency reduction. That means one and a half to two times faster. For that second tranche, that's a very strong value prop there. We're just going to go through. It's a very directed sales motion. We can go target each of those customers with a very specific offer around, here's the value of this.
Here's what it'll do to your array. We've modeled it for you. Here's what it'll cost. We think it's something we can really execute on quite well.
That's on-prem memory. That's incremental, right?
Absolutely. Correct.
Yeah.
It'll follow a motion that we've done when we went from 300s to 400s to M to all NVMe. We leverage the insights and the analytics from Pure1 to drive very focused campaigns to create a good conversion for the new product. It's a great opportunity for us there. Yeah.
Hi. Steve Fox with Cross Research. Two questions also. First of all, just on the cohort performance you put up in terms of the sales force. As we introduce all of these new products, is that a positive for cohort performance? In the interim, if we think about it over the 12 to 18 months, is there an incremental ramp that we have to think of that maybe slows that momentum? I had a follow-up.
Yeah, I don't think there's incremental pressure on that based upon training and development and certifications. Back to what Charlie was talking about, can the field sell a more cost-efficient sort of FlashArray for new workloads? These are the same buyers. It's the same motion customers are asking us for. If anything, I think it's a nice contributor to faster ramp over time. As we look at the ramp to productivity, there's multiple factors. There's product portfolio to be able to get after TAM. There's awareness, there's pipeline that's already there. There's obviously training and certification and enablement, et cetera. It is a contributor, I think, to the net positive that's there. It's early days, so when we look at those ramps and anything we can do to bend them through a combination of factors, we will.
Okay. I think I know the answer to this, but I just want to ask anyway, which is getting back to the QLC question. If you're initially using TLC on the C series, does that mean that there's a gross margin give up in the beginning as you ramp the FlashArray//C, or is it a neutral effect? Can you just talk about that?
There'll be a bit of pressure on the gross margin number. Again, relative to the overall business, that's blended in our overall gross margin guidance because, again, this is really the initial launch of the product.
Okay.
I would say that though that part of the algorithms that we use in the QLC algorithms uses the TLC more efficiently by having software that Remember, every flash product uses excess flash to deal with write wear and other things. This is capacity optimized, we're able to make better use, more efficient use of the flash than we do in our standard product. It does have lower cost.
Okay.
Ari with Goldman Sachs. Thanks for taking my question. I have two as well. First, on FlashArray//C, you talked about solving some of the issues with QLC with your software. Could you talk about how difficult that is to do competitively and whether you expect to have a lead, or how long of a lead?
Yeah.
For my second question, I think Prakash mentioned in his presentation that you're having conversations with customers about how they like your products, but they would like to see you do more stuff before they can get really interested. Could you talk about what those top things are? Is it mostly low-end with disk or hybrid, or is it file storage, or could you give us a list?
Right. Absolutely. Do you want to start, Keith?
Yeah. I think we've well proven over the last couple of years what an advantage talking raw to the flash has been for bringing out NVMe, and we did so at such a higher level of efficiency than our competitors that had to use off-the-shelf drives. We think we're going to have an even greater advantage here with the QLC transition. At the end of the day, QLC is much lower write endurance flash. What you have to do is you have to be much more careful in how you write to it.
What that means in a lot of the QLC SSDs, first generation that have come out, it means they're way over-provisioned. If you have a certain price delta, but then you start adding more flash to make up for it, pretty soon it floats back to the TLC price, and you add some SCM cache, and before long, there isn't much of a margin delta to exploit anymore. That's why with this product, we really wanted to design a product that was all about QLC, where we made a lot of software-level changes, how we write to the flash, how widely we stripe and protect on it, the ratio of CPU in the server to the flash. At a high level, we gave up performance to have much more longevity from a lower tier of flash.
That's really created two very nice price-performance places in our line card now. I think the other nice advantage to the business is that when you have a lower cost tier, you can protect price on your higher cost. Now if a customer really has a capacity-optimized use case, then in the past, they might have come to us and said, "Man, I really need the price at this." We can say, "Yeah, we can do that with C and preserve margin on X.
Okay. In terms, and Hap can, I think, weigh in on this as well, but in terms of what customers are asking us for, they are asking for the, and there is no other way to phrase it other than the Pure experience, in other parts of their portfolio. There is a reason for the 86.6 net promoter score, right? It goes to the actual experience with the product. You can get to the same destination driving a Volkswagen as driving a Tesla, right? The experience of driving a Tesla is really unlike any other car. Still, to this day. With us, the proof of this is in the driving experience, right? Our customers will tell you they sleep better at night. They have their weekends free with much less care when it is on a Pure workload, and it is just a lot less work for them overall.
They would like that experience in other areas. They'd like it on more file products. They'd like it on their tier 2 storage. They'd like to get rid of racks and racks of equipment that they can't do today because of the economics, right? Far, we've only really been relevant in primary workloads, and now we're going to be on the tier 2 of workloads, which is a very big estate, as well as on the so-called tier 0, and increasingly in the file portion of their estate as well.
Yeah. The only thing that I would add, is it's hard to capture how different it is, and people just initially purchase Pure as the land for a performance characteristic of a workload or for a TCO or efficiency to get out of the planned obsolescence and the tech refresh model. They fall in love with the simplicity and the automation that we provide them. You heard all three talk about, and I know you probably think these are plants, but please spend time on the floor with any of our customers or any of our partners, and you'll really recognize that there's a pervasive enthusiasm for what our simplicity and what our automation, what our process improvements mean to them. That 86.6 translates into repeat purchase rates and translates into higher productivity for our sales reps.
It is truly different, and it's hard to articulate that until you actually go get those experiences from them. To be able to do that now in places that are even more painful than their tier 1 production environments and their tier 2, I mean, there are football fields of disk arrays that are sort of supporting tier 2 workloads that are a total pain, that are a massive drain on power. They're very complex. They're non-performant. They give horrible SLAs to their customers. It's a pain point that I think our customers have really been looking for us to do it. As I think about our expansion opportunities, it's across block and tier 0, in file and object, with replication and other capabilities that we've got. You get more and more workloads.
We have a unified investment that we made with our acquisition of Compuverde that's going to be able to come in and provide a more addressable market, and being able to help them with their cloud strategy. They want that holistic experience. We're excited about it.
Hi, Tim Long here at Barclays. Also try to sneak in, too. First, could you talk a little about Cloud Block Store and CloudSnap and how you see the evolution and the ramp of those products? Related to that, I imagine it's pretty important to show traction with that customer base. Is there anything different that happens on the sales incentive side for that? Secondly, back again to investment and sales, it seems that you've, obviously, the new cohorts are doing better and your higher win rates and higher deal sizes in enterprise and cloud. When we think about where the incremental investment's coming, is it going to come in those two verticals, or where's really the focus going to be to help leverage the top line more than it currently is? Thank you.
When we think about the applicability of Cloud Block Store to our overall business, it's helped in tangible and intangible ways. There's obviously a revenue opportunity there. It's one that you have to remember is a subscription service, so it's going to ramp from what you guys see flow through at a different rate than, say, a CapEx sale. It's one where we're going in with a premium offering, a premium service, and we're offering something that isn't available in Amazon today. We think it has a very strong value prop. I'd also say it has an intangible in the sense that it really rubs off well on the on-prem purchase. We found that the set of conversations that we can be involved in now with customers around their cloud strategy has really driven on-prem sales as well.
That's particularly underscored by the Pure as-a-Service model, formerly known as ES2. With Pure as-a-Service, a customer can sign up for a contract with us, deploy everything on-prem on day one, and then have the flexibility to move to the cloud. If you think about the mentality from a customer who might be all on-prem today but has the mandate from their CIO to start moving to the cloud, this is a perfect solution for them because they don't have to sign up for another three to five-year array experience with someone else. They can have everything on-prem today, but they can know that they've taken a tangible step to be able to start going to the cloud. They can move their tier 1 applications to the cloud without having to re-architect them. That's been a great discussion in motion already.
Relative to sales investments, obviously it's expensive to build a B2B systems company in a market like this. That's why a lot of companies didn't do it, and they sold their companies to companies that had big distribution already and plug them in. We knew we had something that was so meaningfully different than what the other folks had and was going to be so impactful to their core revenue lines that an independent company needed to be built. We've been managing our sales capacity investments across geography and across segment and are working to continue to do that. We've seen nice contributions from non-North American markets to our growth rate outpacing the growth of the overall business. You need to be able to support global companies in a global way.
That initial investment is pretty significant, but you can get leverage from that over a longer period of time, so we'll see more leverage there. On the segment side, the easiest dollar to go pick up four or five years ago was a commercial dollar, because it was faster sales cycles, it was less competitive. They can move quicker in their own decision-making process. We're going to continue to have that. We've got a really nice business. We want to move much of that business to our partners to become more independent, and incrementally shift our resources into the largest enterprises and central governments and healthcare organizations globally. You'll see us incrementally shift the investments from commercial toward the large enterprise. We want to get that rate and pace right, because there are longer sales cycles.
They are difficult to break into. You just want to keep blending it. I look at it, and we look at it more as a portfolio across geography and segment and getting the knobs tuned. We added, as you know, 40% or so capacity out in that we're now focused on getting productive. As that comes online in a year, in two years, in three years, that's leverage for the overall business. I don't think our strategy in terms of our investment thesis changes. It's just continuing to fine-tune the knobs across segments and geography.
The only thing I'll add to that is that, as Hat mentioned, we did add significant new capacity this year. We did that in anticipation, knowing that we were going to be delivering these products at this point in time, roughly at this point in time. We wanted to make sure that we had the training, the customer relationships with the new capacity that we could start to exploit it now rather than hiring the capacity after the product deliveries.
Yeah, you wanted to dovetail both of those sort of intersecting at the same time.
Right.
Really, really important.
Matthew Cabral from Credit Suisse. Hat, you spent some time on this in your presentation, but just, you touched on it a little bit earlier, too, channel independence. Can you just talk more about where you are in that process and in particular, as you think about the channel being able to sell themselves without you getting involved? I'm sorry, just related to that, you said yesterday you spent time with 1,400 channel providers. Just what's their biggest ask from you guys as they're looking to grow their Pure business?
Great question. We've segmented corporate under $500 million, commercial $500 million to $5 billion, enterprise $5 billion and plus, and then been building out an inside sales team to be able to drive accountability for the partners that are there to make sure they're representing Pure but more cost effectively and more efficiently. We're into that journey two and a half years. We've always had a lead qualifying engine in the inside since the inception, since I've gotten here for the last eight years or so. It's a natural progression for some of those employees to move into the inside sales. We're well along the journey in that corporate segment. We changed the programs last year, and that provided standardized discounting, standardized pricing, incentive structures, a lot of the learnings that Charlie brought from running commercial in one of his many jobs at Cisco.
To drive that velocity curve and that flywheel, you needed to put the infrastructure in place from a program, then you needed to put the systems in place to enable configuration, pricing and quoting, et cetera. We put a lot of that investment in for our 10s and 20s, the lower-end part of our product portfolio that hits that corporate market segment. They're totally independent. I think in the commercial segment, where they get into the larger systems of performance configuration, there's some additional systems-level work that we need to continue to work on there. That's just a march that we go through. I think that the dovetailing comment's a good one, which is put the program in place, put the pricing in place, put the systems in place, then incrementally move people up market on our own, and shift over to them.
That's what our traditional solution provider VAR. I would say if you look at the managed service provider segment or the GSI segment, these are areas that can be completely low touch. Rackspace, as an example, is differentiating their product offering based upon our technology capability, offering something new and going and selling it with their own sales force. I love that. To the extent that we can go create routes to market where they're completely independent based upon their own business model we're investing in. The GSIs have managed service contracts that buy a lot of equipment that right underneath it that they can go swap out, and effectively we're now supporting many other G2K companies that we didn't actually have to touch because they just got a better service level at a lower price point based upon their own in-house infrastructure.
That's kind of a sell to model versus a sell through and sell with that we're already seeing some leverage on that's low touch. It varies depending on the route to market and the segment that you're serving. I think we're well down the path of providing independence across the board. Remind me of the second question?
Just the ask from the 1,400 channel partners.
I always ask this question whenever I meet with a customer or whenever I meet with a partner, "What else can we be doing?" They're like, "Just don't change anything. We love the engagement with you." I'm like, "That's not that helpful." Like, "Tell us where we can do better." I just had a meeting with a very large New York-based solution provider, and he was asking what our services strategy was. I said, "Well, it's to enable you to capture revenue on the services line to go deliver a more consultative sell and a more holistic sell." That's an area that we've been investing in over the last year and a half. Still a very small part of our business. We don't want to be in that services business.
We want to enable our partner ecosystem to be able to go do it. He was asking for, "Well, give me the frameworks. Give me the playbooks. Give me the competency at an architect level to go help my team do the first couple of these, and then I'll take it from there." That's an area where it's still early. As you move into the large enterprise, more and more important to be able to provide advisory services. We've always done implementation services and data migration services and some of the things that just accelerate adoption of the product. When you're getting into more of a consultative multi-cloud strategy and how do you architect that, and how do you take advantage of all these, it's a higher-level services engagement that we're investing in. That'd be one.
Hi, David Rojek, Susquehanna. Kix, just a clarification on the C. You mentioned about a 40%, the C is priced 40% lower per gigabyte than the X today using the TLC. Does that mean when QLC becomes more broadly available, that means that pricing would become 50%-60% lower than X?
The 40% is where we're introducing it today, and it's what we expect to be the delta when we shift over to QLC next year. I think it's probably not wise to think about it beyond that. We do expect QLC will go down over time, but so will TLC, and so we'll have to see how the market for both these things ebbs and flows over time.
Great. No one asked about hyperconverged, so I thought I had to ask this question. Hat, you mentioned it in passing during your presentation. I'm just trying to understand philosophically where you stand. It does seem like it's the fastest-growing segment within storage. The largest storage player, it's their fastest-growing product. What is your position on this, and do you feel like you're comfortable over the next five years without hyperconverged in your portfolio?
Yeah, let me start, then I'll hand it over to Hat. I'll apologize ahead of time because it's a little bit of a geeky answer. It's not a simple answer. If you were to ask customers who appreciate hyperconverged what it is that they like about hyperconverged, it really splits into two areas, okay? One area is this entire concept where you can build your application on one box. What hyperconverged means is hyperconverged networking storage and compute. Although it's really more hyperconverged computing and storage. Then as they scale, they add another box, and they add another box. That's one part of hyperconverged, and that's specifically what a hyperconverged meant in the beginning. The other part of it is that we can manage that environment on one pane of glass. All right? One management environment for the whole thing.
That first part about add another box, add another box for scaling, that works for small to medium systems. It does not work for a data center scale environment. One way to think about it is hyperconverged is not hyperscale. We tend to work in large data center environments. When we talk about petabytes, we're talking about large-scale environments. That other part, being able to manage everything on one pane of glass, that's very valuable. That's an area where we're going to continue to pursue. Part of what we're talking about in terms of creating this modern data experience is not just managing it on one pane of glass, but having it be self-managing, so you don't even need it on the pane of glass.
Obviously integrating much more deeply into the upper layers, that is, the virtualization layer, the containerized layer, so that the customer manages their entire environment as one thing, is absolutely critical for us, and us are working with VMware, with other companies on that, so it is a single pane of glass, is absolutely critical for us. This idea of box by box, where it's highly constrained, symmetric scaling, rather than being able to scale each part of your estate independently, we really believe more in the second one than in the first one.
Yeah. I would just say that I think our religion is on growth and hitting customers' expectations. When we get engaged in a hyperconverged versus a converged sort of battle, they're still low single digits for us, and our win rates are consistent. The challenge for us is really a marketing and awareness challenge that we need to go address in addition to the technology areas, and so we're focused on that.
I'm being told in the back that we've got one last question. I think, isn't that what you told me?
Yeah.
I'm sorry, I don't read hand signals very well.
I'll tell you what, we'll do one more.
Okay.
We're all set.
All right.
Hi, this is Pinjalim from JPMorgan. You're of course doubling down on subscriptions with Pure as-a-Service and other Pure1 offerings as a pro, and I think I also saw a way to buy online maybe through Pure1. Would we see, today it's 33% of total billings. That's up six points or something, I think, from a year ago. That I am assuming will probably accelerate over time. First point is, would we see something like half of your business being subscription in two years or so? How much of a headwind on the flip side would that be on revenue? Should people start thinking about billings or RPO bookings or something like that?
Do you want to take that one?
Well, the answer is yes, we do expect to see subscription become a larger part of our overall revenue profile. Of course, then being able to provide additional guidance to analysts and our shareholders in terms of what that means for us on a going-forward basis is going to be critically important. We had alluded, I would say, in earlier earnings calls that we'd provide an update on this during Accelerate. Tim and I came to the conclusion that given that we'll have a new CFO, that maybe we can defer that until probably the beginning of the new fiscal year.
The bridge that we provided to this, to Charlie's point, I think over time, particularly as the % rises and rises, you'll see more disclosures along those lines. The bridge we wanted to provide today was some of the SaaS-like metrics. Again, as we look at our business, I think it's an underappreciated asset in terms of things like net dollar retention, things like the amount of TCV under contract, et cetera. It's a flavor of kind of the transition that you're seeing with our business, and to Charlie's point, as we get further down this path, obviously, there's the strong likelihood of more data over time. All right, now we're being told that we need to wrap up. Thank you very much. Appreciate your time and attention today. Thanks again. We'll be in touch.
Yeah. Thank you all for coming. I appreciate this is a lot of time out of your calendars. We really appreciate you taking the time and trouble to come and visit us. Thank you.
Thank you.