VAST Forward 2026 | Jeff Denworth, VAST

Clips
More from VAST Forward 2026

Jeff Denworth

Co-Founder

VAST Data

Vast Data marks 10-year milestone with ambitious AI operating system vision

Businesses rushing to implement artificial intelligence are unfortunately running into fragmentation across storage, data management and compute. It’s a dilemma that Vast Data Inc. is eager to correct, using its proprietary AI Operating System platform.Following a decade of engineering around first-principles distributed computing, Vast is using its tenth anniversary to hail the movement of AI-native infrastructure from experimental to foundational. At the center of that transformation, AI itself is the catalyst driving a fundamental rethink of the data and compute stack, according to Jeff Denworth (pictured), co-founder of Vast

play_circle_outline Financials: over $500M ARR, billions deferred revenue, accounting profitable, cashflow independence.

play_circle_outline Market opportunity: VAST targets AI-related TAM roughly $200 billion annually.

play_circle_outline Customers: AI cloud platforms and labs (CoreWeave, Lambda etc.) plus enterprises at varied AI maturity.

play_circle_outline VAST: From Universal Storage to an AI Operating System — Global Scheduler, Filesystem, Index, and Device Drivers

play_circle_outline Technical differentiation: first-principles architecture, emphasis on distributed computing, ACID/state and geo-distribution.

play_circle_outline Ambition to reinvent entire data processing stack, not just storage.

Info
Transcript

Jeff Denworth, VAST

Jeff Denworth

Co-Founder VAST Data

In this exclusive from VAST Forward in Salt Lake City, Jeff Denworth, co-founder of VAST Data, joins theCUBE’s Rebecca Knight and Dave Vellante to mark a decade of disruption – and to signal a far bigger ambition ahead. With more than $500 million in ARR and billions in deferred revenue commitments, Denworth describes a company riding an exponential curve, propelled by AI’s insatiable demand for flash, compute and globally distributed data infrastructure. What began as a storage rethink has evolved into what VAST now calls an “AI Operating System” – a softwar... Read more

explore Keep Exploring

How close is the company to going public, and how do its ARR, deferred revenue, and profitability compare with those of major tech software IPOs? add

What is the company's total addressable market (TAM), which markets does it encompass, and how large is it (in dollars)? add

What types of customers does the company serve, and how do they differ? add

What is an operating system, and why would you describe this product as an "AI operating system"? add

What makes VAST different / why should someone choose VAST? add

What are the defining characteristics of VAST that make it special? add

bolt Powered by CUBE AI

Jeff Denworth, VAST

search

Rebecca Knight

>> Hello everyone and welcome to theCUBE's live coverage of VAST Forward 2026 here in Salt Lake. I'm your host, Rebecca Knight alongside Dave Vellante, my co-host and analyst. I would like to welcome Fresh From the Main Stage, Jeff Denworth. He is the co-founder of VAST. Welcome, Jeff.

Jeff Denworth

>> Thank you.

Dave Vellante

>> 10 years. Wow.

Jeff Denworth

>> It's a decade.

Dave Vellante

>> Nearly a dozen.

Rebecca Knight

>> You're entering your tween years as a company.

Dave Vellante

>> True. Double digits.

Rebecca Knight

>> Well, that's actually, let's start there because this really feels like a moment of truth for VAST. I'm curious to hear you talk a little bit about what's different now at this company versus, I mean, obviously at your start 10 years ago, but even just a year ago, in terms of the life of VAST and also where we are in this moment of AI transformation.

Jeff Denworth

>> I did this for... We just had our company kickoff yesterday. Our one-day company kickoff, which things are pretty compressed in terms of time at VAST. And there's a quote that Hemingway writes around, it's from the sun also rises around going bankrupt. Somebody says to the main character like, "How long does it take to go bankrupt?" And the answer is, well, very gradually and then suddenly. And I think about that with respect to VAST in the moment that we're in right now and the idea of exponential curves. When we started the company, we were super small, couldn't even see us on the map of the technology space, like a little spec. Then we grew and grew and grew and still wasn't very big. And then when you start growing, compound interest is the eighth wonder of the world. And we're now way bigger than we had expected to be even a year ago. And we're seeing that in all sorts of different dimensions. We have people that are coming to us, not to sound like too full of ourselves, but we declared sold out about four weeks ago and people were like hostile about the fact that they couldn't get in here, not a lot, but like enough where it was like, this is a little strange. We look at what's going on right now in the market with respect to like flash capacity and compute infrastructure. And what we've realized at the beginning of this year in particular was that we have such an outsized impact on the market that we're moving like significant parts of the industry right now. And these things come up on you all of a sudden where you don't actually realize it until you're in it. And then the interesting thing that we've been thinking about and trying to conceptualize is, well, if you're on an exponential curve, where does it go from here? And what does the next five to 10 years look like? And we have no clue, but we're ready for the ride.

Dave Vellante

>> But you're still not that big.

Jeff Denworth

>> In terms of what?

Dave Vellante

>> I mean, in the grand scheme of the industry, you're pretty tiny.

Jeff Denworth

>> I think so.

Dave Vellante

>> I mean, you're like a lever because you're in all the right spots, but I mean, I don't know, the last time... At one point last year you guys talked about a run rate number.

Jeff Denworth

>> Yeah.

Dave Vellante

>> What can you tell us today? You probably can't say anything.

Jeff Denworth

>> We have like a legal team now.

Dave Vellante

>> But what was the last public statement?

Jeff Denworth

>> I think I'm allowed to say a few things. So we're at a run right now of over a billion bucks.

Dave Vellante

>> Okay.

Jeff Denworth

>> Excuse me, half a billion of ARR. And so there's a ton of deferred revenue. One of the ways we sell is customers make large commitments to us and the deferred revenue is in the many billions of dollars. And so I look at it in terms of scale and perspective. You think like, okay, well, what were some of the biggest tech software IPOs in history? And we're of that level where we could come out and make like a very, very significant splash in the market with one big distinction, which is if you look at the core business fundamentals, most big tech IPOs are underpinned by this idea or question mark about sustainability, like how long can this run for? When are they going to become profitable? When are they going to stop burning cash? And we're already to the point where we haven't been burning cash for almost half a decade and we've been like accounting wise profitable this year, which means that we can just keep sustaining this for as long as we want to.

Dave Vellante

>> But in terms of the market share of the macro, I mean, I don't know. Your TAM, it's bigger than the storage TAM, you would agree?

Jeff Denworth

>> Yes.

Dave Vellante

>> Okay. It's like the AI TAM.

Jeff Denworth

>> We look at it in two ways. One is we take a view of, well, what's the market that VAST is going after? And it's storage, it's database, it's compute infrastructure from a software perspective. And then, well, who are your competitors in that? And that's kind of like everybody.

Dave Vellante

>> Yeah.

Jeff Denworth

>> But then the question is like, who are we disrupting? And AI becomes the agitator in the market that is the thing that invites us into the door to do our thing. And so that TAM, by my rough estimation, is about $200 billion. And I don't think we sell a drop of it if customers don't need to change how they're doing things, but because AI is such an agitator in a good way in terms of like enabling business productivity and gains and things like that, that our TAM is roughly like a $200 billion year TAM.

Dave Vellante

>> Right. There are competitors that are envious, but not because you're necessarily disrupting them directly. You're getting dragged into the AI.

Jeff Denworth

>> That was one of the big realizations I made the other day. I looked at a few of our competitors who have had no market share gains where they've operated in the last five years, like one, 2% growth rates, which is like not even tracking with inflation. And we came out of nowhere and we're selling today billions of dollars a year in terms of software contracts. We had hardware that were dragging either in public cloud or private cloud. It's a big business. And those other vendors are still selling, which tells me that we're operating in a space that most other companies aren't. And in certain cases, we have like almost full market share, but yeah, it's like all the new stuff.

Dave Vellante

>> As a co-founder and somebody who's... Renen talked about principles, not rules. Is one of those principles that efficiency or do you ever think, wow, to really scale this thing, we need to scale our go to market. You clearly have product market fit, you have go to market fit. You could scale that and actually not be as profitable or even cashflow negative. Would you ever consider doing that?

Jeff Denworth

>> I think the question is what underpins that principle and what is underneath that is a thrust towards independence. And so we never want to be in a position where we're dependent upon any other party to give us money to operate our business, right? And so the interesting thing is that the cash balance has to grow as the company grows because the safety net needs to be larger. And so they're directly correlated. And efficiency isn't the goal, independence is the goal. So we can do what we want to do. We can do what customers ask us to do and we're not subject to the whims of external forces, I'll say.

Rebecca Knight

>> Well, so actually, let's talk about the customers because you had your sales kickoff yesterday. You are running on fumes, Jeff.

Jeff Denworth

>> I'm okay.

Rebecca Knight

>> It's your first ever customer event today. I'm curious about the conversations that you're having with customers and where they are maybe further along than you expected and where they're perhaps more stuck than you'd like.

Jeff Denworth

>> Oh, I think it's interesting to understand our customer dichotomy before we get into that. The company was born with an idea that we could power AI pipelines and there weren't any existence. So we basically just started helping people that did large scale computing, like high performance computing centers like traders and life scientists and things like that, with an idea that AI would come. And when it did come in 2022, what we realized is that that would first manifest in some of the world's largest AI cloud data centers where AI labs were those original customers. And so if you think about platforms like CoreWeave and Lambda and Crusoe and Nscale, these are all like built on VAST platforms and they are servicing the needs of the largest AI labs. And there's a bunch of them here. There's a bunch of clouds here, there's a bunch of cloud customers here, but then there's this whole other part of our business that's getting the enterprises ready for their kind of like large scale deployment of AI, AI agents in the market. And those are much earlier in the process of figuring things out. And so it's a whole spectrum. I wish I could say there's like one archetype for everything that there's not.

Dave Vellante

>> All right. So let's get into some of the details of vision and announcements. Let me start with, what is an operating system and what's the difference between an operating system that orchestrates things like memory and IO and an AI operating system?

Jeff Denworth

>> Okay. So let's have the conversation about what an operating system is before we get to VAST.

Dave Vellante

>> Great.

Jeff Denworth

>> An operating system is a few things. You have a run time that you can run your applications on. You have a file system that you keep kind of like your unstructured data in. Sometimes you have an index for that file system that's in the form of a database or something like that. You have a set of security policies and you have device drivers that basically make the hardware that's underneath it run. So we've been building for the past 10 years and last year was the moment where we got to the point where we realized that what we were trying to call this thing didn't marry up to what it actually was, right? And so started as a storage product called a universal storage. Then we said, okay, no, it's not that. It's an AI data platform and data platforms are pretty popular. People could kind of like, you know, you could understand those words. And then we started building geo distributed pipelines on the platform and we had these SDKs and you could just run AI models natively within it. And we took a step back and we said, "Okay, you've got this global compute scheduler on top of this global file system with a global index and we build when we're allowed to all the way down to the device driver level so that we can get the maximum efficiency out of hardware." And at that point we ran out of ideas. We didn't know what else to call it, so we just called it the AI operating system. Now there are other people that use the term, I mean, it's kind of like so generic that it's tough to figure out what it is. We have customers that call their cloud platforms AI operating system. You have NVIDIA calls some of their stuff-

Dave Vellante

>> I know they call their-

Jeff Denworth

>> Dynamo was an AI operator. Does it work with training? No, it's more for inference, but it's just a kind of a general term and it does require some explanation, but anything that was like specific, like let's say Dynamo, people like, "What's a dynamo?" Okay, well I still have to explain it anyway, so why don't we get through it.

Dave Vellante

>> Okay.

Jeff Denworth

>> So then VAST is like... So why us? And I think the thing that I talked about on stage is that I think you can define VAST by two different characteristics besides Renen's frugality, which creates efficiency in the company and things like that. Kidding. First, I think of us as being very first principles minded. And when we started the company, we spent a year just working on architecture and an engineer's inclination when they get a bunch of venture capital is just go start coding. And we resisted that in a way where we could just spend a year thinking. It was almost like a think tank, about seven guys in a room, probably pretty stinky. Some of them were here, hopefully not stinking up this room. And the idea was, well, let's envision the first 10 years of the product and let's envision it for all the forms that we wanted it to be. And it wasn't just a storage conversation. We were building asset data structures for distributed database transactions. We're building geo distribution tech capabilities. And then Rome is never built in a day, even if you have the architecture and the plans. So we started building and we built in a way where we can monetize the first aspects of the product. The second big characteristic... So the first one was first principle is like, let's just solve for the most fundamental problems in distributed computing so that we can just get this thing that pays dividends as we expand the capabilities. And that brings us to the second part, which is second characteristic of VAST that I think is really special is one of ambition, right? We're on a march not just to reinvent how people store and access their data, but basically to reinvent the entire data processing stack. And so that involves us thinking about and endeavoring upon basically computing with the entire data industry and the entire compute industry and the entire storage industry, all with a single software product that we think kind of defies conventional limits in a way where customers just get gravitated to the gains that they get, not only from each individual use case, but when everything comes together and you kind of see it magically working altogether. And so first principles thinking coupled with real ambition, like wanting to not be constrained by anything is how we define ourselves. And that gets infused into this AI operating system we built.

Rebecca Knight

>> So before we get to the announcements, which I want to get into, but I have to ask you about culture because it is something that Renen talked about and you just talking about first principles and this ambition really gets at the culture of the company and this mission drivenness of you and Renen and the seven stinky guys and now the big company that you are today.

Jeff Denworth

>> I'm going to get in trouble for that one.

Rebecca Knight

>> Sorry, sorry. Okay. We'll... Anyway, how do you maintain that? Because I mean, it's one thing for two founders to get together and say, and bounce ideas off each other and have this immense sense of urgency and these high expectations for themselves and for each other. But then as you're building the company, it is harder to maintain.

Jeff Denworth

>> It is.

Rebecca Knight

>> And you might make things harder for yourselves, in fact, if you have those expectations.

Jeff Denworth

>> Well, I don't know quite how to answer that question other than to say, number one, we don't actively have culture as a program. In fact, we have a very, very small HR team that basically helps us with recruiting, but it's not to impose a set of behaviors or rules on the team, right? And that is by design. We don't want to get bureaucratic. And then number two is we spend almost no time talking about culture, but we do try to operate according to those principles that Renen talked about and once a year we get together. Last year, it was his 42nd birthday. Nobody realized this until the end of this talk, but he gave us 42 principles for the company and the 42nd was about the secret of the universe, which is also 42. On his 42nd birthday, it was just like totally a meta thing. So anyway, so yeah, it's just-

Rebecca Knight

>> It's really embedded. Yeah. And how you operate, but it's not something you talk about. Interesting.

Jeff Denworth

>> Not much. Not much.

Rebecca Knight

>> Okay. All right. Well then let's get onto the announcements.

Dave Vellante

>> And before we do.

Rebecca Knight

>> Yeah. Sorry, sorry.

Dave Vellante

>> To me, the architecture is what is most fascinating because what you just described is you didn't set out to build a faster box. I wrote a piece years ago Thinking Outside The Box asked.

Jeff Denworth

>> No incrementalism.

Dave Vellante

>> Yeah. And not a lot of storage companies think about acid properties when they're starting to-

Jeff Denworth

>> I agree.

Dave Vellante

>> Right? And so that to me says a lot. The whole stack is reforming. Thank you. NVIDIA's started it all. But when you think about networking, totally changing. When you think about the data stack has to totally change. But not a lot of companies other than VAST are really thinking about that, how storage has to change. And my question is-

Jeff Denworth

>> I think there's a reason for that, by the way.

Dave Vellante

>> So explain that. But my real question is, how far can you go? Because there's this whole intelligence layer emerging, and it's unclear. Is it going to be the... Everybody's saying SaaS is dead, it's not, but maybe parts of it are. The LLM vendors clearly have a say there. You guys have ambition. There's this system of intelligence that you're feeding, how far do you go there? But why is it, to your point, that not a lot of people are thinking about that in the storage space, and how far do you go up that stack?

Jeff Denworth

>> I'm going to try to not sound arrogant here, but I think it's in part because we've cornered some of the market. And as a result, there's not a lot of space to innovate by other companies that don't play or don't get to go to the party that we do. And so as I think about it, the fact that we are welcomed to work with some of the world's largest model builders, it puts the fuel in our tank to go and build to what they're looking for in ways that we don't see a lot of other companies even having these conversations, much less thinking about co-engineering with them. Like this year is a big year for us where we're building a customer engineering team, which is not a customer team, it's not an engineering team, it's not a suite. We have 15 different types of teams. Now we're building one just for AI builders and AI labs that we can co-invent technologies together to scale up intelligence next order or two of magnitude. And that's a very special place for us to be. I got a text message from one of the world's largest AI builders yesterday, and he just said it's go time for something that he's working on that's really special. And that just inspires us so much to work so much harder.

Dave Vellante

>> Jensen talks about extreme co-design. A lot of his extreme co-design is within NVIDIA, but of course it's with the partners as well, no question.

Jeff Denworth

>> Yeah.

Dave Vellante

>> But you're talking about kind of a similar, the integrated stack-

Jeff Denworth

>> Sure....

Dave Vellante

>> and the partners. So how far can you go? Do you see VAST as feeding that system of intelligence or do you actually start to get into being that system of intelligence?

Jeff Denworth

>> We're never going to train a model.

Dave Vellante

>> Right. Sure.

Jeff Denworth

>> You never say never, but I don't think we're ever going to foundationally train a model. We want to be like the Windows for AI, right? You just put your apps on it works, boots up, makes that magical noise. You got the green field in the background as you look at your desktop. And imagine that's VAST, right? Everybody in the world, whether it's a cloud platform or a server that you've deployed or some edge robot, it's just VAST running everywhere, not running the intelligence, not being the intelligence, I should say.

Dave Vellante

>> But people don't pay for operating systems unless it comes with some hardware or it's free like-

Jeff Denworth

>> I'm sure that's true....

Dave Vellante

>> like Linux. Let's talk about that.

Jeff Denworth

>> There's a couple companies that sell operating systems that are pretty meaty, Amazon, Microsoft-

Dave Vellante

>> Yeah. Okay.

Jeff Denworth

>> Google, Oracle. The operating systems have changed.

Dave Vellante

>> But they sell other stuff with it and you do too.

Jeff Denworth

>> Right. There you go. So maybe the definition changes where-

Dave Vellante

>> Exactly....

Jeff Denworth

>> Linux has commoditized device drivers and things like that. We run on Linux as well. We don't actually develop anything at the OS level. We think the definition of an OS has been abstracted by cloud platforms and what we're trying to do is build something that's all encompassing without compromise.

Dave Vellante

>> It's more than an operating system. It really is.

Jeff Denworth

>> Okay. So now you're-

Rebecca Knight

>> We need another word. I think that's what's clear.

Jeff Denworth

>> Oh my God.

Rebecca Knight

>> I'm sorry.

Dave Vellante

>> You guys have this conversation internally, I'm sure.

Jeff Denworth

>> So Renen-

Dave Vellante

>> doesn't do it justice, but go ahead.

Jeff Denworth

>> I agree, but Renen has a great analogy, which is like, if you think about the PC and you think about the smartphone, you think about cloud, cloud is the new OS. All of these things democratized new hardware platforms so that organizations could consume them at mass scale. And so yeah, probably isn't the best word. I agree.

Dave Vellante

>> That's right. We can move beyond that.

Jeff Denworth

>> But it works.

Dave Vellante

>> So the announcements today, which ones excite you on be Polaris, I call that Supercloud. I don't mean to trivialize how hard it is to do that, but it's table stakes. You've got to run everywhere if you're VAST.

Jeff Denworth

>> I think so.

Dave Vellante

>> So that's an important announcement, but it's... Which one's exciting? And the C-node stuff was kind of really super interesting. To me, because I think this is the year that enterprises really start getting return, that embedding that governance and security into the platform to me was the most important. But I'd love your take.

Jeff Denworth

>> That's what I focus all my time on.

Dave Vellante

>> Yeah. I'll bet.

Jeff Denworth

>> I mean, it's trivial to take the great work that NVIDIA does from a library perspective, package it into some new systems, partner with the likes of Cisco and Super Micro, get that out into the market, and now we've got Accelerated VAST. And they've done some great work. I don't mean to trivialize that, but for us, it's a packaging exercise and integrating with some of their libraries so that we can realize that acceleration. The hard work that we're doing is rethinking things like agentic memory, because all of this just comes down to state. One of the early things that we talked about was just building state machines.

Dave Vellante

>> I was just going to say that. Yes.

Jeff Denworth

>> These are all nerdy topics.

Dave Vellante

>> This is what the guys at Thinking Machines and Kendall Square Research were doing.

Jeff Denworth

>> Yes.

Dave Vellante

>> They were building state machines.

Jeff Denworth

>> I read your blog.

Dave Vellante

>> Thank you.

Jeff Denworth

>> And I read it after I got on stage and showed the connection machine. So you and I are thinking about exactly the same stuff.

Dave Vellante

>> I was waiting for the shout-out, bro.

Jeff Denworth

>> I didn't see it. I was just walking over here. Sorry. So anyway, so what is state, right? Well, state is anything that's kind of like data oriented, right? And anything can be stateful if you decide to make it such, right? And two agents communicating with each other, if you log that and persisted, that's state that the system has access to. The memories that an agent has or the new learnings that a model has been fine-tuned with, all state. So VAST data, you could argue that we're a little bit biased here, looks at everything as a data problem, right? And if you view it as that, and you realize that you can put controls around data, and if you view communications also as data, then you can put policies in place to enforce how data is transacted through a system. And so policy engine is something that we're creating right now to basically put the guardrails on every single activity that happens in a distributed AI system. Nothing happens out of band from this, so it's complete control without performance compromise. And we think that this is going to be table stakes for more people explaining AI first more than... I think trust is kind of there, not entirely there, but being able to replay or understand what happened is the hard part right now.

Rebecca Knight

>> The auditing.

Jeff Denworth

>> And auditing and putting the right AI on top of the audit logs to really understand what's happening, that's a big part. And the second thing that we realize, and this gets us all the way to that Thinking Machines thing that we were working on or thinking about even 10 years ago was, well, if we're going to do this, well, let's make sure that we've also governed the state that exists within whatever our model's been fine-tuned with, either with Laura or approaches where models just fine tune themselves. And so here, we basically concluded that we have to build our own fine-tuning engine into the platform itself. And so now I can deploy the agents, I can deploy the models, I can fine tune them all in this kind of like continuously improving system. And if you do that and you have all the access to the state and all the logs and everything, then you have complete observability and complete omniscience about what's happening within your system. And so that was the big thing that we wanted to get out this year is like, we want to build this thing into a holistic offering that every enterprise can deploy to kind of safely execute AI wherever they want to, whether it's public cloud, private cloud, wherever. And we're basically bringing to market tools that only otherwise exist in some of the largest AI labs that are bringing out to market, but it's not preferential to a model. It's not preferential to a platform. Everybody can get access to this. And that's a big part of our, I would say one of the other lesser characteristics of as we really believe in democratization of infrastructure and AI infrastructure. And so deployability is big thing.

Rebecca Knight

>> But also part of that ambition. Jeff, thank you so much for coming on theCUBE.

Dave Vellante

>> I want to have you back. John and I want to do a-

Jeff Denworth

>> In New York....

Dave Vellante

>> long form, two hour podcast with you-

Jeff Denworth

>> Holy....

Dave Vellante

>> in our New York.

Jeff Denworth

>> Holy.

Dave Vellante

>> Maybe three hours. Seriously.

Rebecca Knight

>> You can do it. You're up for it.

Dave Vellante

>> I'm serious. You're in New York, but if you're ever-

Jeff Denworth

>> I'm rarely in New York.

Dave Vellante

>> I know. I know, but we'll find a time because we have to go deeper on this topic and to spend some real good quality time with you. It's the 15, 20, 30 minutes, not enough.

Jeff Denworth

>> Rebecca, first time I met you.

Rebecca Knight

>> Yeah.

Jeff Denworth

>> Dave, you and I go back like decades now.

Dave Vellante

>> Yeah, we sure do.

Jeff Denworth

>> So there's never enough that I can see of you. And it's always great to be here and great to talk to you.

Dave Vellante

>> Thank you.

Rebecca Knight

>> Excellent. Well, thank you so much. We'll look forward to many more conversations with Jeff Denworth down the road. Thank you. I'm Rebecca Knight for Dave Vellante. Stay tuned for more of theCUBE's live coverage of VAST Forward. You're watching theCUBE, the leader in enterprise tech news and analysis.