To Bank of America's Global Tech Conference here in San Francisco. Delighted to have you all over here. I'm Wamshi Mohan. I cover IT hardware and supply chain for the bank. I'm delighted to welcome NetApp to the stage today. We have NetApp CEO George Kurian, who you're all familiar with. George obviously has had a long history in the technology industry, NetApp, 10 years coming up or just happened. And so congratulations. That's a great milestone. And you've seen a lot of cycles in your share of the time over here. Always love to hear your perspective on where things are. Before we get started, I do have to read. We also have Jeriel Ong from Investor Relations at the back. If you have any follow-up questions, you can definitely reach out. I'll quickly read a safe harbor, and then we can get into Q&A.
Today's discussion may include forward-looking statements regarding NetApp's future performance, which is subject to risk and uncertainty. Actual results may differ materially from the statements made today for a variety of reasons described in NetApp's most recent 10-K and 10-Q filed with the SEC and available on their website at netapp.com. NetApp disclaims any obligation to update information in any forward-looking statement for any reason. I think I can kind of become the master of reading these in one breath. I'll pat myself on the back for that. Welcome, George. Thank you so much for being here. Really appreciate it. Maybe to kick it off, you just reported a pretty strong end to your fiscal year. You got it the next year. What are you seeing at the macro level?
Maybe just to help us think through the way that you thought about this upcoming year, given all the uncertainties, what were some of the puts and takes that you weighed in that?
Yeah. First of all, thank you for having me. And thanks for attending this session. We finished a strong fiscal year 2025, where we executed on all the areas of growth that we had shared with our investors. All-flash arrays grew 14% year on year to $4.1 billion. We gained 280 basis points of share, according to IDC Research. Cloud storage grew 43% year on year. Our Keystone Storage as a Service business grew on a TCV basis, 53% year on year. And our AI business grew fivefold year on year. So we're super strongly positioned. We have refreshed our entire portfolio of systems entering the new year. And we feel really good about our position in the market. When we guided the new year, we looked at our traditional kind of multi-year growth rate sequentially.
We also compared some of the signals we saw from the market in the second half of Q3 and really Q4. We saw a couple of areas where we thought it was right to be more prudent. One was U.S. public sector, which is in the low teens part of our business, but the federal part of that, which is about 75% of that low teens number. Rounded out, about 10% of the business, where we saw increased inspection of deals, elongating deal cycles. In the end, we converted all the transactions that we had in our pipeline. We thought that when you look at how you forecast the year, you typically look at a bottoms-up pipeline and a conversion rate of that pipeline. We naturally, in these circumstances, bring down the probability of conversion.
We also saw some caution in the European markets that were particularly manufacturing and export-oriented, just people saying, "Hey, I want to wait and see what the tariff situation looks like before I make large-scale purchasing commitments." Those are the two reasons for caution in our Q1 outlook. Through the course of the year, you will see that sequential growth rates are within the ballpark of what we've done in the last 10 years, maybe a little bit higher weighted in the second half than the first half. We talked about visibility to some larger AI deals that we're working on, as well as additional capacity that we have put into the field that will take some time to ramp.
No, that's great context. Maybe just talk about this AI deal opportunity that you're talking about in the back half of the year. What kind of products? What kind of customers? What's the opportunity?
Yeah. These are our flash and object storage products. We have had wins with the same players before. It is not like we are trying to get a design win. We already designed in. We have been successful at a range of clients. The end clients are large enterprises or large public sector organizations that want to build sovereign AI environments. Our technology is really, really strong for a whole range of reasons, which I do not need to elaborate. We have good line of sight. These are complex transactions. Working through the agreements with clients to get that closed.
How are you, George? How are you thinking about NetApp's overall portfolio? Are there any areas that you think you need to refocus on? I know we sort of went through a phase a few years ago around cloud ops and then refocusing the company to some degree, maybe away from cloud ops in particular. On all flash, you guys have been very steady. You've shown a tremendous amount of growth, well above market, well above peers. Sustainability, confidence of that, anything you need to do in the portfolio to maintain that level of growth, how are you thinking about it?
We feel really good about the portfolio. We, in the course of the last fiscal year, sold our Spot and Cloud Checker business, which is a $95 million headwind on a year-on-year compare with FY2026 versus FY2025. Cloud storage, the hyperscaler native cloud storage, is now 75% of total cloud revenue, which is a really strong number. It's growing at north of 40%. We feel really good about that. Our all-flash business is two-thirds of our hybrid cloud business. The disk-based business is about one-third. It has been declining at a 7% CAGR. We expected that as people moved to flash. That should asymptote over time because an increasing portion of that business is nearline hard drives, which today flash cannot economically replace for most rational clients.
That's interesting, given what we heard a little while ago. I guess when you think about the opportunity with sort of the asymptote or maybe the ex all-flash array portfolio, what do you think is the steady state in the, or maybe not steady state, but just over the next year or two? How should people think about modeling that piece of the portfolio?
Listen, I think we said at our last financial analyst day, our goal is to drive revenues in the mid to high single digits and to grow earnings per share in double digits. We are in that ballpark for 2025. And we want to do that for 2026 and 2027. It's too early to tell. That's what we're focused on.
Okay, great. Maybe on all-flash, as you think about this portfolio mix shift that's going towards all-flash and enterprises, can you talk about, has it been performance needs? What kind of workloads that are running that are driving that continued shift towards all-flash? And sort of the relative economics between all-flash and your hybrid portfolio for you.
Yeah, I think maybe I can just go back to the days before flash, right?
Sure.
Where there was memory, which was used as a cache like DRAM, and then there were hard drives. There were three classes of drives: 15K, which were kind of these are the number of revolutions per minute that the drive did. 15K was high performance. 10K was sort of general purpose. 7,200 RPM or nearline was capacity focused. From a market perspective, from a revenue perspective, it was, let's just say, 15% 15K drives, 35%-40% 10K, and the remainder was nearline. From a capacity perspective, obviously, nearline dominates. When you saw flash come into the market, 15K was a no-brainer. It was really easy to replace that with flash because it was all about performance. Flash was just superior straight up. You saw a rapid clip in transition there. With 10K, the questions became a little bit more nuanced.
You had to see flash costs getting close enough to 10K. Then some of the benefits of flash was it's more consistent in terms of performance, so less optimization required. Now we are in probably the third inning of a nine-inning ballgame around the 10K install base getting replaced with flash. Nearline is more complicated. In nearline, really, performance has no value. It's just cold storage. People are using it for backup and data protection. Flash has to get a lot closer to that nearline economics to make it really work. I think there are clients who have data center constraints in some environments and things like that, where there are unique circumstances. From a broad-based change, it's got to get economic.
Okay, that's helpful. Maybe to talk a little bit about profitability and margins. When you think about your product growth margins over the last, call it, a decade or so, your time over there, there's been periods of volatility in that. Some of it was driven by COVID and supply chain things and all kinds of expedites and other things that you had to manage through at that point. Also, flash pricing has been pretty volatile over that period. A, what's your opinion on how you look at sort of the state of supply demand for the flash market as you look at it, or call it the next six months? I know it's harder to predict longer term, but for the next six months, are you doing anything in terms of pre-buys, or where do you stand with respect to any strategic purchases?
How should we think about that margin trajectory, I guess, putting all those together?
Yeah, I think, first of all, in our long-term model, we said, "Hey, we try to operate the business between mid to high upper 50s." We are not trying to get to the 60% range in product gross margin rate. I think if you look at NetApp, we have operated a company with a very fixed operating cost structure. OPEX CAGR over the last many years has been 2%. And so what we really see is if you can drive revenue and profitable gross margin dollar growth, there is enormous leverage in conversion of the incremental gross margin dollars into operating margins and earnings. We're always making trade-offs across that. I think to get specifically to your question, overall operating margins for the company are fueled by kind of really strong numbers in cloud. We'll talk more about that if you'd like. Strong support growth margins.
Product growth margins should tick up through the course of the year. It does not need to reach 60% for us to reach the targets we outlined because we see cloud gross margins ticking up from 79% crossing 80% this fiscal year.
Yeah, maybe that's a good segue into just talking about sort of what you think the growth profile of that business looks like for you. Maybe it'll be helpful to give some context. I don't know. Everyone's as familiar with that part of the business. Just how much of that is consumption versus subscription? Maybe elaborating on your margin comment about exceeding 80% over there. As you utilize these assets more and there's more revenue, I doubt you're adding a ton of incremental costs. I would think that the margins can have further room to expand as well. Would love to get your take on that.
Yeah, thank you. The cloud storage business, the cloud business began with the work we did with the hyperscalers, Microsoft, Amazon, and Google, where we took our software and first made it available in the cloud marketplaces. This was in 2013. As we saw growing customer adoption, worked with the hyperscalers to natively embed it in the hyperscaler environments. We are now part of every Amazon data center, every Google data center, every Microsoft data center. We report that as a segment called public cloud. It is an ARR business. We had to go through a few transitions in that business. The first was a move from third-party to first-party. Obviously, being embedded as a first-party service sold by the hyperscalers natively integrated into their billing, operating environment, security architectures is sort of exalted high ground.
Our clients, as we made the first-party technology available, shifted from third-party to first-party. The second was the hyperscalers used to be subscription businesses. When we started, we had to make our technologies available alongside their business model. We were in subscription. When they switched to consumption, we had to switch to consumption. Today, 80% plus of our business is consumption. Subscription is meaningless, which is why we now report it on a revenue basis because the concept of ARR has limited relevance now. Cloud storage, which is the natively integrated solutions we sell with the hyperscalers, are now 75% of our business. They have grown, as you could, we reported, north of 40% year on year, an acceleration in terms of growth rate in 2025 versus 2024. We feel really, really good about our position there.
I think on the margin profile, you hit it exactly right. We started to work with Microsoft in particular, having deployed systems in a whole range of their data centers. Those systems had a depreciation life that is coming off. The first of those systems went in five years ago. You will see depreciation coming off consistently. In all of the hyperscalers, we also have mostly software solutions now as a growing percentage of our mix. The combination of revenue scale plus depreciation expense coming off the P&L drove an 1,100 basis point improvement from the start of last fiscal to the finish of last fiscal. It should continue to tick up. Too early for me to give you a new range or gross margins, but we are going to monitor the situation.
We feel really, really good about our ability to continue to drive both the growth rates as well as the contribution margin from that business.
Okay, that's helpful. Can we just talk a little bit about sort of the market share gains that you experienced over the last year? What would you say was the core driver behind that? You just sound like you just refreshed all of your product portfolio. As you think about the rate and pace of that, can investors think about an accelerating trend over there in terms of market share or at least sustaining sort of what you did last year?
Yeah, listen, I think that the bets that we made over many years that clients are now starting to realize is super important to them are starting to play out in our favor. We said, "Hey, you want to have a trusted unified data foundation because you need to unify all your data so that you can actually use it with AI." Every one of our competitors sold silos, right? Pure has three silos, Dell has nine or seven, HP has three or four, IBM has three or four. We were always the people that said, "You want one. You want one. You want one." That kind of historic kind of truth that we have told the market is playing out in our favor.
In fact, everybody else is trying to copy what we are doing by saying, "Oh, we can unify your data in a bunch of different ways," none of which is real. The second is our capacity to do hybrid workloads and to be able to offer private clouds with all the features that, listen, Amazon, Microsoft, and Google all chose us as their storage partner. That means we got awesome tech. I'll just leave it there. That's playing out on-prem because clients are now beginning to realize, "Hey, these guys have a really good multi-tenant architecture." I think on AI, which is a material part of our growth this year, listen, I think that data management is super important in AI workloads because people want to version their data. They want to implement security and access control. They want to automate data pipelines.
We do all of that extremely well. We have been focused on enterprise AI. That's starting to play out. We have a really strong portfolio there.
No, that's super helpful context. Maybe it'd be good to get your perspective given your position in the market around what are you seeing with workload movement, given that you have exposure both to on-prem and public cloud? Are you seeing much of containerization being used as something to go test out different environments? How much are you seeing maybe repatriation of workloads from public cloud? Any trends that look interesting to you that you could talk about?
Yeah, I think, first of all, enterprises are smarter now about where to use cloud and where to stay on-prem. I think before it was kind of a black and white discussion. Some people, "Oh, cloud's bad," and others, "Oh, I'm all cloud." Frankly, I thought both of those were completely misinformed discussions. Now, through COVID, people have realized, "Hey, cloud's really good for certain things. Hey, it's expensive." Now there's a good balance. That's one. I think with regard to proof of concepts, more and more people are comfortable using cloud. Data science people, enterprise workloads, they're like, "Hey, I don't know if this thing's going to work.
Let's run it on the cloud for a few weeks or months, see if the thing works, and then I can stand up a real environment. I think with AI experiments, listen, I don't think there's any difference than traditional environments where there are clients who are clearly concerned about putting their data on public clouds. I think we have seen it's early to say that this is a trend, but clearly, the posture of our government has caused European clients to be a lot more concerned about that. I can tell you, if the Canadians don't like you, you really have a problem. We got some work to do there.
Yeah, I think, listen, there's just people getting nervous about, "Hey, are we really friends, or are you guys really not going to be our friends?" I don't think it affects anything in the near term, but that is causing people to have some—we're doing more explanation that, "Hey, we've been in your country for 25 years.
Right. Yeah, no, interesting times for sure. Maybe to step back a little bit, George, if you just think about the aggregate demand environment, I mean, you noted a few things right at the beginning around some softness in public sector. How are you navigating what's happening with tariffs? How do you think about the impact of that, both from a direct and maybe indirect perspective?
Yeah, I think, first of all, the most important thing you need to do operating in a volatile environment is control what you can control. That is what we tell our teams. I think on the supply chain, over the last several years, we have reduced our exposure to China to de minimis. It is literally like 1% of our cost of goods sold. We have built modular systems that are like Lego blocks so that if one piece of silicon is problematic, we can swap in another piece of silicon. That gives us diverse supply base. It gives us some degree of protection from things like tariffs. Obviously, you cannot just swap it out instantly, but it gives us a lot of flexibility in doing that.
From a system perspective, listen, we've always believed in using industry-standard silicon and using standard componentry with contract manufacturers so we can move our supply base where we want. We manufacture in Asia for Asia. We manufacture in Europe for Europe. We manufacture in a combination of Mexico and the U.S. for the U.S. The country of origin designation is conferred on when the final assembly of the product and the final test happens. We are able to do that with a lot of strategic flexibility. We can make that in the U.S. if we need to because we load our software and run final test here. We can do that out of Mexico. We can do that in a lot of different places.
From a supply chain perspective, we said, "Hey, I think our direct exposure is 50 basis points on calls." It is not a big number. In terms of other elements, obviously, we are talking to our technology providers to make sure that we understand what their exposure is and what they are doing to mitigate tariffs. We have a variety of planning scenarios for exemptions, no exemptions, all of that. We are well-planned for that so we can move things around. I think if you ask me, the most important impact of tariffs is really uncertainty and what that causes in our clients. In terms of direct impact to us, we got it under control.
Okay, that's helpful. Maybe one question we get around just we spoke about the relevance of storage within AI, but a lot of growth is happening either directly within hyperscalers, the CapEx spend, or tier two CSPs. That's kind of where the concentration has been. Why is the reason that we have not seen traditional players really have within the storage industry pick up a lot of that market share? Is it the economics of it? Is it the technology of it? How would you describe it? Is there an appetite or interest to want to address that?
Yeah, I think, first of all, the vast majority of the work that has gone on to date has been about training foundation models. In a foundation model training environment, you have a large amount of compute and networking and a limited amount of storage. The storage is really used as a buffer for the large language model to write what's called a checkpoint. The checkpoint essentially is that, "Hey, I'm going to run through my processing of public data. If I fail, at some point, I can fall back to this checkpoint as opposed to going all the way to the beginning." You can actually calculate the amount of storage. It depends. You can do it's four to eight bytes times the number of parameters in an LLM times that's what you need for one run times the number of runs.
You could say, "Hey, I'm going to keep it for a few weeks." It's a deterministic amount. All it is is extremely high-performance storage with zero data management. It's as close to memory as you can get. That's a commodity environment. I think that we have our value in the hyperscaler environments to bring corporate data to connect with the hyperscalers: Bedrock, Vertex, OpenAI Studio, all of their tools. We're not trying to be the foundation of their model training because that'll just get squeezed out in economics over time. I think with regard to the tier two CSPs, we are successful with a bunch of them. We are cautious about the economic models of many of them, just like we were cautious and actually were right about the call on cloud.
Most of the competitors that we had in cloud thought the telcos were going to win. We said the hyperscalers were going to win. We were right, and they were wrong. We are appropriately prudent about the sustainability and the durability of some of the tier two CSPs. We talked about some large AI wins with tier two CSPs this past quarter, flagship national carriers in big Asian countries. We have had some big wins in the Middle East. We are really focused on, "Hey, are those businesses going to be around, or are they, 'Hey, I've got GPUs?'" When there is a shortage of GPUs and the second everybody has them, they are going to be out of business. We are deliberate about where we bet. We are super strong in enterprise. We got the data. We have unstructured data.
We know how to do hybrid cloud data management. We have far advanced kind of data security, multi-tenancy, and all of that, which is why we are outgrowing everybody else in the enterprise AI market.
I know we're almost out of time. I'm going to throw two quick ones at you, George. One is just around what are some of the things that you put in place to think through any slowness in demand, maybe a slowdown in enterprise spend that you've just kind of perceived a little bit of. If that sort of becomes a larger vector, what are some of the levers that NetApp can pull to manage through something like that? Any closing remarks from you to the investors out here?
Yeah, first of all, listen, we have been prudent operators of the business. We have grown OPEX at a 2% CAGR against revenue of 4, north of 4% CAGR over the last many years. The outlook for this fiscal year is to keep good controls on operating discipline. We are in a strong position entering the fiscal year. We feel like it's important for us to stay invested in capitalizing and using this year to continue to pull away from our competitors. I'm not planning any radical surgery to the operating cost structure of the company. We are obviously going to be careful about how much we spend on what. I think overall, what I would tell you is we finished FY2025 in a really strong position. We gained share in the all-flash business, which is now a materially large part of our overall hybrid cloud business.
It's two-thirds. The disk-based business has shrunk at a 7% CAGR for the last several years. That should asymptote as we exit fiscal year 2026 because the vast majority of that is nearline. The cloud business is now super focused. Cloud storage is 75% of the cloud business. It grew 44% year on year. That should be a durable growth rate for our cloud. That should drive a strong growth rate for our cloud business once we, if you take away the headwinds from Spot compare. It is growing at very, very good margins. If you combine that with disciplined operating environment, we should be able to drive growth, earnings leverage, and EPS through discipline capital allocation.
Fantastic. I think, unfortunately, we're out of time. So, George, thank you so much for being here. We really appreciate it.
Thank you.
Thank you.
Thank you very much.