Croatian-founded Daytona, which aims to build infrastructure for large-scale agent workloads, raised a $24M Series A led by FirstMark
Daytona is building agent-native infrastructure to support large-scale, stateful agent workloads through fast, flexible computing environments designed for long-running execution.
Context & Ripple Effects
Daytona’s round arrives as funding in the coverage is reaching both agent applications and the infrastructure beneath them. Cerrion’s production-line video agents and Crogl’s cybersecurity agent illustrate workloads that must act on operational data rather than merely generate text.
The $24M Series A gives Daytona dedicated capital to pursue environments for long-running, stateful agent execution. That makes it a stack-level complement to vertical agent platforms such as Nitra’s medical-practice automation platform.
First-order effects
- Daytona gains $24M in Series A financing, led by FirstMark, to develop its agent-native infrastructure for large-scale, stateful workloads.
- FirstMark becomes a lead investor in a company targeting the execution layer required by agents that run over extended periods.
Second-order effects
- Agent builders with persistent or multi-step workloads gain another potential infrastructure supplier rather than having to assemble flexible execution environments themselves.
- Infrastructure providers serving agent developers face a clearer requirement to support stateful, long-running execution, not only short-lived model inference.
Third-order effects
- If agent applications continue moving into operational settings, differentiation in AI infrastructure may shift toward orchestration and durable execution environments alongside raw compute access.
- Funding across both vertical agents and their underlying runtime layer suggests the agent market could separate into specialized application and infrastructure vendors, though adoption will determine how durable that split becomes.
The trend: This is one data point in the build-out of an agentic data-center stack tailored to persistent software agents rather than one-off AI requests.