Daytona Raises $60M Series B at $1B to give every agent its own computer
Daytona raised a $60M Series B at a $1 billion valuation, led by FirstMark Growth, bringing total funding to $91M. The angel list reads like a who's who of companies that now run agents: Patrick Collison, Tobi LΓΌtke, Aravind Srinivas, Eric Yuan, Olivier Pomel, Adam D'Angelo and Brendan Foody, with Qualcomm Ventures, Samsung Next and Okta Ventures joining. Daytona's post is dated October 20 and TechCrunch covered it on October 23. https://www.daytona.io/dotfiles/daytona-raises-60m-series-b-to-build-the-agent-cloud
The pitch is that an agent's computer has little in common with a cloud instance. An instance boots in minutes and runs until you kill it. An agent's machine has to start in milliseconds, pause with state intact, fork into a thousand parallel copies, rewind, and vanish when the job ends. Daytona calls it the Agent Cloud and says it is the only provider with Linux, Windows, macOS and GPU sandboxes on one platform, so the same agent can tune a CUDA kernel, test a Mac app and drive a Windows desktop.
The growth numbers are the reason for the valuation. Revenue is up about 25 times since January, with average month-over-month net revenue retention around 135 percent, so customers from six months ago spend nearly three times as much today. Customers include model builders like Meta and Cohere and the RL-data shops, Handshake, Snorkel, Surge AI and Turing. Writer ran its own sandbox platform until it passed 20,000 sandboxes a day, then moved to Daytona and provisioned 300,000 in one month.
The customer list tells you where the money is. The biggest buyers of agent sandboxes are not app developers, they are the labs and data vendors running RL environments by the million. The same week OpenAI described training its frontier models inside layered containers with no internet, Daytona got a billion-dollar price for selling that exact primitive. Sandbox was a feature a year ago. Now it is a cloud category, and the next fight is whether the labs keep buying it or build their own.
← Back to all articles
The pitch is that an agent's computer has little in common with a cloud instance. An instance boots in minutes and runs until you kill it. An agent's machine has to start in milliseconds, pause with state intact, fork into a thousand parallel copies, rewind, and vanish when the job ends. Daytona calls it the Agent Cloud and says it is the only provider with Linux, Windows, macOS and GPU sandboxes on one platform, so the same agent can tune a CUDA kernel, test a Mac app and drive a Windows desktop.
The growth numbers are the reason for the valuation. Revenue is up about 25 times since January, with average month-over-month net revenue retention around 135 percent, so customers from six months ago spend nearly three times as much today. Customers include model builders like Meta and Cohere and the RL-data shops, Handshake, Snorkel, Surge AI and Turing. Writer ran its own sandbox platform until it passed 20,000 sandboxes a day, then moved to Daytona and provisioned 300,000 in one month.
The customer list tells you where the money is. The biggest buyers of agent sandboxes are not app developers, they are the labs and data vendors running RL environments by the million. The same week OpenAI described training its frontier models inside layered containers with no internet, Daytona got a billion-dollar price for selling that exact primitive. Sandbox was a feature a year ago. Now it is a cloud category, and the next fight is whether the labs keep buying it or build their own.
Comments