When OpenAI launched Codex last year, it was pitched as an AI coding assistant, a tool that could write, debug, and explain code in natural language. But what started as a developer-sidekick has quietly transformed into something far bigger: an autonomous agent platform. And with last week's acquisition of German cloud startup Ona, OpenAI just made it official.
On June 11, 2026, OpenAI announced it would acquire Ona, a Berlin-based startup that provides secure cloud execution environments for AI agents. The deal, terms undisclosed but reportedly in the hundreds of millions, brings Ona's orchestration technology directly into the Codex ecosystem. The result? Codex can now run persistent, long-duration tasks without human supervision.
Think of Ona as a cloud runtime purpose-built for AI agents. Before this acquisition, Codex could generate code and answer questions, but it couldn't hold state reliably across long sessions or execute multi-step workflows in the background. Ona's architecture changes that: agents can now spin up cloud environments, run complex pipelines, and report back, all while the developer sleeps.
This is a bigger deal than it sounds. Codex usage has surged 400% in the past year, and OpenAI revealed something surprising: knowledge workers now make up roughly one-fifth of Codex's user base and are growing three times faster than developers. That's not a coding tool anymore; that's an enterprise automation platform wearing a developer interface as a disguise.
Ona's investor list reads like a who's-who of enterprise cloud: General Catalyst, Crane Venture Partners, Speedinvest, and Vertex Ventures US, alongside individual backers like Shopify CEO Tobi Lütke and Datadog CEO Olivier Pomel. The startup already served two million developers and had deep integrations with organizations that also work with OpenAI, making the acquisition a natural next step.
What makes this particularly interesting is the customer-controlled execution model Ona brings. Enterprises have been wary of letting AI agents run wild inside their infrastructure. Ona's architecture gives companies secure, isolated cloud environments where agents operate under strict guardrails, approving code before deployment, sandboxing file access, and logging every action for audit trails. For CIOs worried about "rogue agents," this is the safety valve that makes autonomous coding palatable.
The acquisition also intensifies the arms race in what's being called "full-stack agent infrastructure." Anthropic has its own agent tooling with Claude. Google's Gemini is pushing deeper into long-running tasks. Meta's Llama ecosystem is going open-source with agent frameworks. But OpenAI just leapfrogged them all by buying the actual runtime, not just the model, but the secure cloud infrastructure that models run inside of.
For developers, this means Codex agents that don't stop when you close your laptop. They'll monitor production systems overnight, triage bugs as they appear, and deploy fixes by morning. For knowledge workers, it means AI that can research, draft, iterate, and deliver without constant hand-holding. And for OpenAI, it means a moat that's increasingly hard to cross, because models come and go, but infrastructure integrations stick around for a decade.
Ona is still a relatively small team, around 60 people, but their technology is about to become the backbone of what Codex does next. OpenAI CEO Sam Altman described the acquisition as "bringing a full runtime to every Codex agent," adding that "the model is the brain, Ona is the hands that get things done."
In a week where the Forbes AI 50 list dropped and AI companies race toward IPOs, this acquisition might be the quietest power move of 2026. It doesn't make headlines like a new model release, but it changes the battlefield: from who has the best model to who has the most capable, trustworthy, and persistent AI workforce.
