AI teams are starting to pay for a new workflow: giving autonomous agents a secure “computer” where they can run code, use tools, and complete tasks without exposing core systems. US venture firm FirstMark is doubling down on that need, leading Croatian company Daytona’s $60 million Series B (~EUR 56m) at a reported $1 billion valuation.
Daytona builds infrastructure for AI agents, focused on isolated sandboxes for code execution and programmable environments that can be spun up quickly. The company positions this as an “Agent Cloud” rather than a general-purpose cloud service. It has reported startup times of roughly 60-90 milliseconds, a performance claim that speaks to agent workloads where latency and rapid provisioning affect how reliably agents can chain actions.
The financing is a continuation of an existing investor conviction. FirstMark previously led Daytona’s $24 million Series A and has now led the Series B as well. Daytona has raised $91 million across seed, Series A, and Series B rounds, with the latest financing also including major strategic and technology investors.
Why this round fits the current infrastructure cycle
This is a with-trend infrastructure bet. As more companies move from “chat” to “do” and deploy agents that execute code, open tickets, query systems, and run multi-step workflows, they run into two practical blockers:
- Security and blast radius: letting an agent run arbitrary code on shared infrastructure is a governance problem. Isolated environments reduce risk by separating agent execution from production systems.
- Operational repeatability: production agent workloads need consistent environments, tool access, and auditability. Purpose-built sandboxes can standardise what agents can touch and how they are monitored.
Daytona’s pitch is that agent compute should be provisioned on demand, with isolation as a default. That positioning can create switching costs if the product becomes embedded in how teams package tools, manage credentials, and define safe execution policies for agents.
Commercial reality: who buys, and how it expands
The most likely buyers are AI platform teams, developer productivity teams, and security-minded engineering leaders who want agents to execute tasks without turning every run into a bespoke DevOps project. In practice, adoption often starts as an enablement layer for a small number of internal agent use cases, then expands when teams standardise on a single runtime for multiple agents.
Retention tends to be driven by three factors:
- Implementation depth: once sandboxes are integrated with identity, secrets management, CI/CD, and internal tooling, replacing the platform is operationally painful.
- Trust and policy: security review cycles are long. A platform that passes review and proves containment becomes the path of least resistance for new agent projects.
- Usage-based expansion: if pricing is tied to environments spun up, runtime, or workloads executed, growth can track the number of agents and tasks, not just seats.
Competition is still forming. Daytona is operating in a specialised, emerging category focused on “computers for AI agents”. While the company frames this as distinct from conventional cloud services, it is adjacent to developer infrastructure, sandboxing, and security tooling. The sources do not quantify fragmentation or establish a clear incumbent, but the direction of travel is consistent: more autonomous execution means more demand for controlled compute.
What funding is likely to support
Daytona has not detailed a full use-of-proceeds plan in the available reporting. Based on the category and stage, likely focus areas include scaling go-to-market capacity, hardening the platform for enterprise requirements (policy controls, audit trails, integrations), and broadening partnerships with strategic and technology investors (inference).
For Croatia’s tech ecosystem, the reported $1 billion valuation also marks a milestone: Daytona is described as the country’s third unicorn.
What this enables
- Faster rollout of production-grade agent execution environments with isolation by default
- A more standardised runtime layer for agent workflows across teams and use cases
- Deeper integrations with enterprise identity, secrets, and developer toolchains
What to watch
- Whether Daytona can translate developer-led adoption into repeatable enterprise sales cycles
- How pricing maps to agent usage as workloads move from pilots to always-on operations
- Competitive pressure from adjacent developer infrastructure and security platforms as the agent stack matures