OpenAI acquires Ona to give Codex persistent cloud environments for long-running agents

OpenAI announced on June 11, 2026 that it plans to acquire Ona, a cloud infrastructure company formerly known as Gitpod, to expand its Codex AI coding agent with secure, customer-controlled execution environments. The deal—terms undisclosed and subject to regulatory approval—will bring Ona's engineering team directly into OpenAI's Codex group.

What's new

Ona provides persistent cloud environments purpose-built for AI agent workloads. Unlike ephemeral compute that spins down when a session ends, Ona's platform keeps agents running inside a customer's own cloud infrastructure. Data, credentials, and audit trails remain under organizational control—a critical requirement for enterprise deployments.

With the acquisition, Codex will gain:

• Secure persistent workspaces: agents continue long-running tasks after a developer disconnects
• Customer-controlled execution: code and data stay within the customer's cloud boundary, not OpenAI's
• Enterprise audit trails: activity logs remain with the customer

Ona was rebranded from Gitpod in late 2025. Gitpod had served roughly 2 million developers by moving local development environments to the cloud before repositioning itself around AI agent infrastructure.

Context

Codex has grown rapidly in 2026. According to OpenAI, more than 5 million people use Codex each week—up approximately 400 percent since earlier in the year. The coding agent can already handle multi-step programming tasks, but until now its agents ran in OpenAI-managed environments, a sticking point for organizations in regulated industries that need data residency guarantees.

The Ona acquisition follows a broader pattern among AI labs acquiring infrastructure companies to close gaps between frontier models and production enterprise use. Microsoft has Azure, Google has its own cloud, and Amazon runs AWS. OpenAI, as an independent lab, has lacked a native secure execution layer. Ona fills that gap specifically for agentic workloads.

Why it matters

AI agents operating on long-horizon coding tasks need something stateless cloud functions cannot provide: a stable, auditable workspace that persists across sessions. Ona's architecture was built for exactly that. By folding it into Codex, OpenAI is targeting the enterprise segment most hesitant to adopt agentic AI—organizations that need contractual guarantees about where their code and credentials actually live.

It also positions Codex more directly against GitHub Copilot, which benefits from Azure's enterprise cloud infrastructure, and against Cursor and Replit, which are building their own agent runtimes. Persistent, customer-controlled execution environments may become a significant differentiator as coding agents shift from autocomplete tools to autonomous systems running complex overnight workloads.

This acquisition signals that OpenAI views Codex not merely as a developer productivity tool but as the entry point for AI agents operating inside enterprise software stacks—and that a trusted execution layer is table stakes for that ambition.