OpenAI Buys Ona To Run Codex Agents Inside Enterprise Clouds

OpenAI just bought the company that will let its AI coding agents slip inside your own data center. The acquisition of Ona, formerly known as Gitpod, marks a strategic pivot from building smarter models to controlling where and how those models execute.

OpenAI announced on June 13, 2026, that it is acquiring Ona, the startup that runs development environments in the cloud and previously operated under the name Gitpod. The deal gives OpenAI the infrastructure to deploy Codex agents directly inside a customer's own cloud, rather than forcing enterprises to send code out to OpenAI's servers. This is a major shift in the coding-agent race: model quality is no longer the only battlefield—the execution layer now matters just as much.

For years, OpenAI's Codex—the AI that powers GitHub Copilot and other coding assistants—ran exclusively on OpenAI's own cloud infrastructure. Enterprises worried about data privacy, compliance, and latency were hesitant to let an external model touch their proprietary code. The acquisition of Ona solves that problem by providing a secure, isolated environment where Codex agents can run inside a customer's virtual private cloud or on-premises infrastructure. Ona's technology, built originally for ephemeral developer environments, is now repurposed as a secure sandbox for AI agents.

The transaction comes at a time when the market for AI-powered coding assistants is heating up. Rivals like Replit, Cursor, and Sourcegraph's Cody have been gaining traction by offering agents that can access private codebases. Amazon's CodeWhisperer and Google's Gemini for Code also compete. But OpenAI's move is unique: instead of just offering an API, it now offers the entire execution runtime. Customers can deploy a Codex agent that reads, edits, and generates code without that code ever leaving their own network.

Ona was founded in 2018 as Gitpod and later rebranded to Ona in early 2026. The company had raised more than $40 million from investors including Spark Capital and Sequoia. Its platform allowed developers to spin up full development environments from a URL, which made it a natural fit for running AI agents that need to interact with codebases. The terms of the acquisition were not disclosed, but sources close to the deal estimate it at several hundred million dollars.

Industry analysts see this as OpenAI's answer to the growing demand for private, secure AI agents in regulated industries like finance, healthcare, and defense. "The real value in AI coding is not just writing code—it's doing it inside the context of an organization's existing systems," says Sarah Chen, an analyst at Gartner. "OpenAI's acquisition of Ona gives them the on-ramp to that enterprise market." The implication is clear: the next phase of the AI arms race is about deployment, not just capability.

Looking ahead, expect OpenAI to tightly integrate Ona's infrastructure into its enterprise offerings. The company is likely to announce a new product line—possibly called "Codex Enterprise" or "Agent Execution Environments"—later this year. Competitors will scramble to offer similar isolation capabilities. The acquisition also signals that OpenAI is serious about competing head-to-head with cloud providers like AWS, Azure, and Google Cloud, which have their own AI agent platforms. The battle for AI-powered software development has moved from the model to the metal.