Big language models get the attention. The machines that rent them somewhere to run are quietly becoming the better business.
At a $1B valuation, Daytona closed a $60M Series B led by FirstMark. The round also drew Alumni Ventures, Evergreen Fund, Okta Ventures, Qualcomm Ventures, Samsung Next and Torch Capital, plus individual cheques from Adam D’Angelo, Eric Yuan, Aravind Srinivas, Tobi Lutke and Patrick Collison. Pace Capital doubled its position, Upfront Ventures tripled and E2VC quadrupled theirs, bringing the total raised to $91M.
The product is deliberately unglamorous. One API launches agents into their own isolated environments, where they run code and call tools. Linux, Windows and macOS are all supported, capacity scales per task, and a paused environment resumes with the agent’s place intact.
Autonomy changes the load profile. An assistant answering questions needs inference capacity; thousands of agents taking actions need somewhere safe to run, fail and retry. Co-founder and chief executive Ivan Burazin frames the risk in headcount terms.
In his telling, firms will shortly be running thousands of digital employees, and every one of them requires somewhere safe to fail, so that one rogue agent cannot cascade into company-wide chaos. That is a security and reliability pitch rather than a performance one, and it lands with platform teams.
The customer list supports the thesis. Anthropic, OpenAI, Nvidia, Cursor and Cognition all run on it, and the wider customer base spans Meta, Cohere, Mercor, Handshake, Surge AI, Postman, n8n and Clay, for training, evaluation and deployment.
The growth is the loudest signal. Run rate rose 25-fold between January and September. The new money funds expansion beyond the original sandbox into a broader agent cloud, plus hiring in San Francisco, Croatia and Singapore.
