Analysis
Daytona, a San Francisco startup that rents out cloud sandboxes where AI agents actually execute code, has raised a $60 million Series B at a $1 billion valuation, Fortune reported. FirstMark led the round.
About $50 million of the round came from existing backers, and Daytona has now raised roughly $91 million total. Individual investors include Stripe's Patrick Collison, Shopify's Tobi Lutke, Perplexity's Aravind Srinivas, Quora's Adam D'Angelo and Zoom's Eric Yuan, alongside strategic checks from Qualcomm Ventures, Samsung Next and Okta Ventures.
What does Daytona actually sell?
Founded in 2023 by CEO Ivan Burazin along with Goran Draganic and Vedran Jukic, Daytona started by automating development environments for large enterprises. It has since repositioned around a narrower, faster-growing problem: giving AI agents a safe, disposable virtual machine to run code in. Burazin says Daytona can configure a new sandbox in 60 milliseconds, and describes the split this way: "The inference clouds is where the agent thinks, and our cloud is where the agent works." Customers named in the Fortune report include Anthropic, OpenAI, Nvidia, Cursor, Cognition, Meta and Cohere -- a list that covers most of the frontier-lab and AI-coding ecosystem at once.
The numbers, and the catch
Burazin told Fortune that Daytona's revenue run rate grew 25-fold between January and September, with average month-over-month growth of 50.9%. Those are startup-stage multiples on what is still an early base -- the company hasn't disclosed absolute revenue, and a 25x multiple off a small starting number is a very different claim than 25x off real scale. That's the gap between the headline and the diligence: fast percentage growth in AI-agent infrastructure right now partly reflects the fact that almost every lab and coding startup is still building out agent infrastructure for the first time, not necessarily that Daytona has locked in durable share.
Competitive landscape
Daytona draws a deliberate line between itself and both the hyperscalers (AWS, Azure, Google Cloud Platform) and the "neoclouds" that rent out GPU capacity, like CoreWeave and Nebius. CoreWeave's own debt load is becoming a sector-wide worry as investors question whether AI infrastructure spending can keep outrunning its financing costs -- a risk Daytona's lighter, sandbox-rental model is structurally different from, since it isn't buying and depreciating GPU fleets. That contrast is part of the pitch to investors: exposure to the AI buildout without the balance-sheet risk of the compute layer itself. It's the same logic behind deals like Nscale and Anthropic's reported $45 billion compute arrangement, where capital is flowing toward whoever can make agent workloads run cheaply and reliably, not just toward raw GPU supply.
What to watch
Burazin's biggest claim is structural: he estimates that Meta's Muse personal-agent product alone would need $2.8 billion worth of sandboxes to serve 100 million users, implying the agent-execution layer scales into a multibillion-dollar market on its own. FirstMark partner Matt Turck frames it the same way, saying in a statement that "the next infrastructure shift is from human-centric cloud primitives to agent-native ones." That's a real thesis, but it's also exactly the pitch every infrastructure round in this cycle makes. The bigger risk for Daytona's next round is retention, not growth: whether Anthropic, OpenAI and Cursor keep paying for sandboxes once their own platform teams build equivalent tooling in-house, the standard exposure for any vendor selling infrastructure to companies large enough to eventually build it themselves. Daytona says the new capital goes mostly toward go-to-market hiring rather than R&D, suggesting the company sees distribution, not the underlying technology, as its next bottleneck.
