Prince Mario-Max Schaumburg-Lippe: Daytona Hits $1 Billion Giving Every AI Agent Its Own Computer

Three years old, based in New York, and now worth $1 billion. That is the short version of Daytona’s week. The company raised a $60 million Series B, TechCrunch reported on October 23, 2026, taking total funding to $91 million. FirstMark Growth led the round. FirstMark also led the $24 million Series A back in February 2026, so this is a backer doubling down, not a first date.

Founder Ivan Burazin runs the company with chief architect Goran Draganic and CTO Vedran Jukic. Their pitch is plain, and it lands because it names a real mismatch. Burazin put it like this: “Agents don’t work like the long-lived services today’s cloud was designed for. They spin up thousands of short, stateful jobs nobody planned in advance, and they need a cloud built for that.”

A computer for every agent

Daytona calls its platform the Agent Cloud. In effect, it gives every agent its own computer. Not a shared container that vanishes when a task ends. A proper machine the agent can return to.

The details matter here. Agents get sandboxed Linux, Windows, macOS and GPU environments on one platform. Sandboxes start in milliseconds. They keep state when paused, so an agent can stop mid-task and pick up exactly where it left off. They can also fork into parallel copies, which means one agent can test several approaches at once and keep the one that works.

That combination is different from the old cloud model. Old cloud assumed a service would run for months. Agent work looks nothing like that. A job might live for nine seconds. Or nine minutes. Then it needs its files, its browser state and its half-finished notes intact for the next step. Readers who followed Warp Sciences raising $215 million to teach robots real lab work will recognize the same pattern: tools built for how this new work actually behaves, not how the last generation behaved.

Numbers that explain the valuation

Startups say a lot of things. The retention number is harder to dress up. Daytona says revenue has grown about 25-fold since January. Average month-over-month net revenue retention sits at about 135 percent. In plain terms, existing customers keep spending more, month after month, faster than the company loses anyone.

The customer list helps too. Reported platform users include Anthropic, OpenAI, NVIDIA, Cursor and Cognition. Those are not casual trial accounts. They are teams that run agents at serious scale and notice quickly when infrastructure slows them down.

The investor list reads the same way. Qualcomm Ventures, Samsung Next and Okta Ventures took part in the round. Angel investors include Patrick Collison of Stripe, Tobi Lutke of Shopify, Aravind Srinivas of Perplexity, Eric Yuan of Zoom, Olivier Pomel of Datadog, Adam D’Angelo of Quora and Brendan Foody of Mercor. Operators, mostly. People who have run the kind of systems agents now need to run inside.

Why forking changes the work

Forking deserves a closer look. When an agent can copy its whole working environment, it can try a risky edit in one copy and a safe edit in another. No cleanup. No fear of breaking the only version. That is how a careful human developer works with branches, except the agent can do it in seconds and at a scale no human team would attempt by hand.

Pausing with state kept works the same way in practice. An agent waiting on a human approval does not need to burn compute or start over. It waits. Then it resumes. Small feature, big bill difference at thousands of jobs a day.

There is a useful parallel in Hi3D open-sourcing its production-tested 3D model: the wins right now come from production habits, tested under load, not from demos. Daytona’s pitch is a production pitch. Milliseconds to start, state that survives a pause, copies on demand.

What the new money does

Daytona says the $60 million will expand the platform beyond its sandbox product, scale its computing infrastructure and grow the team across San Francisco, Croatia and Singapore. Three locations, three time zones, one support and engineering rotation that can follow the day. Sensible for a cloud company whose customers run jobs around the clock.

Will every agent really need its own computer? Maybe not every one. Simple chat replies do not. But the agents doing long research tasks, writing and testing code, or running lab and office workflows already act less like chatbots and more like junior staff with a desk, a machine and a pile of work in progress. Desks help. Machines help more.

Daytona has put a price on that idea, and the market just agreed with it to the tune of $1 billion. The next test is simpler and harder: keep those sandboxes fast as the thousands of short jobs become millions. On the retention numbers so far, customers seem happy to let the company try.

Also published on: