Interview: Ivan Burazin of Daytona on building the cloud where AI agents work

Ivan Burazin is the co-founder and chief executive of Daytona, a San Francisco-based company that rents out isolated virtual computers, known in the trade as sandboxes, in which AI agents write and run code, browse the web and operate business software. He founded it in 2023 with Goran Draganić and Vedran Jukić, initially to automate software development environments for human engineers. Before that he co-founded Codeanywhere, an early browser-based development environment, and started the Shift developer conference, which Infobip bought. Daytona says its customers include Anthropic, OpenAI, Nvidia, Cursor, Cognition, Meta and Cohere.

On 20 October Daytona said it had raised a $60m Series B at a $1bn valuation, led by FirstMark. Qualcomm Ventures, Samsung Next and Okta Ventures joined as strategic investors, alongside founders including Patrick Collison of Stripe, Tobi Lütke of Shopify, Aravind Srinivas of Perplexity, Adam D'Angelo of Quora, Olivier Pomel of Datadog and Eric Yuan of Zoom. Mr Burazin told Fortune that about $50m came from existing investors and that the revenue run rate grew 25-fold between January and September. Total funding now stands at about $91m, and most of the new money is earmarked for go-to-market hiring. The round matters because it prices a layer of the AI stack that barely existed two years ago: the compute where agents act, as distinct from the GPU clouds where models are trained and served.

Answers are drawn from Ivan Burazin's public remarks; sources are linked below.

What does Daytona actually sell, if not a sandbox in the usual sense?

…it can be misleading ‘cause people usually think about sandboxes as a demo or a test environment versus a production-grade environment. But what Daytona does, if you think of the laptop that you have in front of you or the computer that’s over there, or, my wife is an architect, so she has like a Windows with a 3D graphics card inside to do 3D rendering. Like, as humans, we have different computers or different compositions of computers. And our belief is strongly that agents today and going forward will need all these different compositions of computers to do different types of tasks. And so we offer that basically through an API.

You walked away from a product that had customers and revenue. How did you reach that conviction?

Also, even if we'd acquired all those companies, that's a great business, but you still have to think about your second and third act. Versus agents, once we got pilled that there will be vastly more agents than humans, the TAM is just vastly bigger. There were competitors, but no one owned the market, and the market was growing and could be vastly bigger. So this made more sense.

And one thing we'd learned the hard way across multiple companies is that not focusing is the worst thing for us. It's very hard to let go of revenue and some success. So this time I told my co-founder: if we do it, we do it all the way, a pure cut. So we cut it off and sent any customer we had to one of our competitors. Here you go.

Who are the customers, and what are they running?

What I can say for sure: among the top five most valuable companies, two are our customers. Then in the Fortune 500 we have a bunch of household names that aren't tech companies, using us internally, and then all the way down through Series D, C, B, A, seed, and inception. As for use cases, we have two major ones and one minor, newer one. The two major ones are background agents, an agent or service where a human interacts with the agent, and the agent calls our system. Think Harvey, Gamma, Lovable, not saying those are our customers, where a human interacts with an agent and the agent goes off and does something. The second is researchers, for reinforcement learning or evals, who use us instead of a Kubernetes cluster to run their CPU workloads.

Why should a company deploying agents care about isolation?

Companies are going to manage thousands of digital employees, and each one needs a safe place to fail. Daytona ensures one rule breaking agent doesn’t spill into company wide chaos.

There are many sandbox providers. Who is the real competitor?

There are obviously a lot of companies doing sandboxes, I think there are like a hundred of them today. But the number-one competitor is Kubernetes. We see this from startups, but more from large companies and enterprise. When a team or engineer says, "I need a machine for my agent," the most logical thing is to spin up a Kubernetes cluster and give it a pod. But Kubernetes was made for something different.

What we've seen in the last thirteen or fourteen months is that people usually come back and say, "My infrastructure is now hindering my feature set and product growth." We've heard this multiple times, and a lot of people have come over to us, or I assume to other competitors as well, because of that.

Investors focus on GPUs. Why do you think the pressure point will be CPUs?

We internally sort of believe that the number of dollars spent on CPUs will probably be higher than the dollars spent on GPUs in the foreseeable future, which seems sort of counterintuitive today.

I strongly believe that there will be a CPU crunch. The amount of CPU machines that we will need if we want to keep growing at this trajectory actually doesn’t exist right now.

Will you build your own data centres?

…So we run on co-location providers, bare metal machines. Data centers, we technically can run on that or our own data center. Like, that’s how we architected it. Today, from a gross profit margin perspective, it doesn’t make sense for us to get in that. You have to raise a large amount of capital, a large amount of risk for, single-digit percentage points. So today, that doesn’t make sense, but we are fundamentally architected so that we can do that if we want.

Is the end state an AWS for agents?

There will be an equivalent, everyone says like an AWS for AI agents, but… it might look more like Stripe than AWS, in a sense. So there will be a cloud built out specifically for agents. And so that cloud will have sandboxes, and it will have web search, and it’ll have, databases like SQLite or Neon or whatever, specifically for agent and other things. We are not at the end of the new infrastructure primitives for agents. There are more coming.

So people think like, “Oh, there’s nothing else. This it.” There are more. Like, we have some ideas about the next ones. We don’t have time to do them, but there are definitely more primitives that are being built out for agents, and there will be, I think, a cloud that runs all that together.