Hello everyone and welcome back to theCUBE's live coverage of VAST Forward 2026 here in Salt Lake City. I'm your host, Rebecca Knight, alongside Dave Vellante, my co-host and analyst. We are so excited to welcome to the show, Renen Hallak, the founder and CEO of VAST. Welcome, Renen.
Thank you. Wrapping up a good day.
Really good show. Really good show.
So you were on the main stage this morning and you began with a Hebrew saying of, you must know where you came from and know where you are going. And that is apropos because we're here at your first ever customer conference. Would you reflect a little bit on where VAST has been and where it's going, especially when you started it in 2016 at the same year OpenAI was founded and the Google DeepMind. Talk a little bit about what you're seeing in this moment now.
I think we've been building towards this moment for the last 10 years. It's hard, 10 years is a long time, but from the very beginning we were always hopeful that AI will happen. It's happened a lot faster than we expected it to, and it's going to accelerate into the next decade. But that's what we live for. That's why we exist. We built a new software infrastructure layer for these new workloads. And the whole stack is getting redone and we have our part in that stack that we're trying to do.
You said in your keynote, you're not like AI people, you're data people, something to that effect. But we always talk about how we're in the early innings of AI, but in a way, we're not. I mean, you talked about 2016 and Google DeepMind and the neural networks and then transformer model. So when you saw all that happening, connect it to what you were thinking at the time, you were like, this is all going to need data. And the data plus AI was in your vision.
It's all going to need a lot of data because historically computers didn't understand any of this stuff and now they can figure out pictures and genomes and sound and natural language. And so much larger amounts of information and much faster access. And much faster access just for training in order to build these models. And once you want to start inference, then it needs even faster access because you have trillions of agents eventually all needing to access all of this data in parallel. And it's not one hop inference, it's multi hop for reasoning. And then it's reinforced learning across inference and training as we fine tune these models and each agent will fine tune its own and every one of those things is another exponent. And when you add 2, 3, 4, 5 exponents, then you realize the old stuff was not built for it, it won't be able to do it. You need a new architecture, you need a new system. And that's why we're becoming more relevant as we progress on this journey.
And by the way, thank you for commenting on my LinkedIn post. I wrote a post about Danny Hillis who I met in 1986, who was the creator of The Connection Machine and Thinking Machines. And I remember at the time thinking, wow, this is amazing stuff. And then what happened was the economics of microprocessors just wiped out that whole big iron industry. And now we're in a new moment. It's actually the economics of GPUs and buy more and make more, Jensen's law, the cost per token declines. Those are all tailwinds for you. The exact opposite of what Danny Hillis focused or realized.
Extremely lucky in that we started when we did. If we were to start two years earlier, we wouldn't have been able to build this. If we were to start two years later, we would've been late to this generative AI wave that's hitting us now. And I think that's the difference. Danny Hillis and his team at Thinking Machines, they had the right ambition. They had the right target. They were very, very early. And I remember seeing an interview with him where he said they chose the name Thinking Machines because they wanted something they would never outgrow. And that's how I feel exactly. That's why this is our mission. We want to see it through. And I think we're at a point in history where maybe that's possible now.
Okay, so the OS for the Thinking Machines.
When I think of great operating systems, I think of MBS, I think of DEC VMS, which nobody's ever heard of, but I think you're probably too young to even remember. Digital Equipment Corporation. But they were a great operating system. And Unix and Linux, these are the operating system. What do they do? They orchestrate resources. So that's a traditional operating system. What is an AI operating system and why VAST?
So why VAST? Every one of the operating systems that we've had before was for a revolution and a technological revolution that had a specific defining characteristic. You guys both have an Apple operating system in your desktop that was compute oriented. Internet was networking oriented. AI is data. Especially as we move from training to everything else, data is at the core of enabling AI and making it work. And that's why VAST. What is the operating system? It's my analogy for that software infrastructure layer. We're not hardware people, we're not AI model people, but we want to provide that middle of the sandwich software infrastructure stack that includes storage, that includes database, that includes scheduling of jobs and orchestrating compute. It includes security. All of the parts that you need to make it simple, such that regular enterprises can start writing and using AI apps to make it secure, such that they can comply with regulation. It's the software layer that makes a new technology widespread. And that's what we're trying to do for this next wave.
So you're making a very persuasive argument for why it needs to exist, but what would you say to the skeptic that really is asking, but why VAST? I mean, couldn't a bigger, more established company or shouldn't a bigger, more established company be the one to create this AI operating system?
Yeah, they should, but they don't know how. I think the answer to your second question is yes. The answer to your first question is I don't think so. And it's really interesting because we're competing with old companies. Young companies can usually innovate, but at relatively small scale. Old companies usually can't innovate anymore. They can't start from scratch. They're good at iterating on what they have, but that's basically it. And the reason is an economical reason. There's a trade-off between high growth and profitability in the business world. When you're young, you focus on growth, on innovation, nobody cares if you're profitable. Once you get to a certain size, that's the only thing they care about. And then you have to start cutting costs and R&D is the first thing that you cut. And then you don't have the engineers that know how to build new things and then you can't build anything new. And it's a vicious cycle. Our advantage is that we broke that trade-off between very high growth. We've been tripling consistently year over year as we're now in the very, very high numbers. I have a CFO now, she won't let me-
Come on. Wait, wait, don't stop there, Renen.
I will say that in parallel to tripling year over year, we are building a very efficient business and we're above 90% gross margin. We've been free cash flow positive forever. We're generating more than $100 million a quarter in cash now. And we're now profitable from a GAAP perspective as well. And so the fact that we can do both of those things means that we're of a certain size that can tackle very large problems, but we're still innovative, we're still hungry, we're still builders at heart. And that puts us in a unique position, not just based on what we've built over the last decade, but based on the culture and the model of our company to continue to innovate over the next decade as this AI era unravels.
So you mentioned culture, which I have to go into-
Oh, before you go there. I have to follow up on the 90% gross margin. On the marginal economics of software are under attack right now, SaaS companies, we saw their gross margin go from 90% down to maybe mid 70s because of a cloud cost of goods sold. Now they're facing more challenges because they got to pay for tokens. You probably, as you become more multi cloud, super cloud, I would call it, maybe absorb some of those costs. I understand, but do you see the economics of your business as being more durable and more attractive than some of the SaaS companies that are facing costs?
SaaS companies basically built the last generation of applications.
And so if I look at this stack, we're in a different layer of the stack and we're in the new side, not the old side. And so we're building that software infrastructure layer for these new AI workloads. What the SaaS companies built is the old enterprise workloads, and I think there's fear that they are going to be disrupted by these new applications. When you have agentic workloads, when you have AI applications, they will overtake some of these older SaaS applications. Obviously, it's partially true-
Yes. Some will, some won't.
But we're underneath the covers, and so we're building all of these new AI applications, all of these new AI models, all of these new AI workloads at scale are getting built on top of our infrastructure. And so we're riding that wave rather than getting disrupted by this wave.
So the operating system gets pulled along as those machines get smarter and smarter and smarter.
And as they get bigger and bigger and bigger, this is the biggest infrastructure deployment that we've ever seen. And you need the software to orchestrate all of it as it gets to that level of speed.
So you're not directly attacking or disrupting some layer of the stack, it's the architectural change that you're a key component of that is disrupting the computing industry.
Everything is changing. It's new power, it's new chips, it's new software, it's new applications. And we are that layer in the middle. I'll say that... Well, two things. One, if you look at GPU attach rate, how many GPUs are we attached to? It's in the millions. And if you do the math, millions means it's almost all of them. And so we are definitely riding that wave and hopefully contributing to the speed at which GPUs can be deployed. Second thing I will say is if you look at just amount of capacity, we are today or this year for sure going to be the biggest in terms of data under management. And we're tripling year over year. And so AI is driving the workloads towards us.
Let's talk culture. We have to because it's something you talked about on the main stage this morning and this great company that is always there for its customers and its employees. You talked a little bit about your mission, always trying to learn new things. We follow principles, not rules. We have unreasonably high expectations.
Yes. But I imagine that filters through the organization too. I mean, if the boss has unreasonably high expectations. So how do you maintain this, particularly as you are getting bigger? And I know that you talked about how you have not made sacrifices to be able to be independent in this way, but you also said big companies don't have as much leeway to innovate, to devote the time and resources to innovate. So how do you make sure that you are maintaining that kind of culture with the people that you hire and how you approach the work?
I think large companies aren't measured in headcount, they're measured in mindset. And we never want to be a large company. We never want these hierarchy and levels and chain of command, and all of these things that came from military organizations, they don't have any room here at VAST. We're a flat organization. We don't have a hierarchy. I tell people, the customer's at the top, the team supporting the customers right underneath and everybody else is supporting them, including myself. We need a certain type of person that loves challenges, that can't wait for that impossible task so that they can prove to the world that we can do it even though everybody else thinks that we can't. The interview process definitely tries to filter out the people who won't be a good fit, but the ones that end up getting a little bit addicted to this cycle of the way that we operate, the speed that we operate, the intensity that we operate, it's very different than other places. And what you said about following principles, not rules. I think, well, A, we hate rules. And so I tell everybody just do what makes sense for the customer. Do what makes sense for the company long-term. Don't try to copy anybody else. And when we find those principles, they lead us in a direction that ends up being very different than the way other companies are built. And what that means is that anyone who's been here for a few years ends up not being able to work anywhere else.
