OpenAI is acquiring Ona, the secure cloud execution and orchestration company formerly known as Gitpod, to give its Codex coding agent persistent, customer-controlled workspaces where tasks can run for hours or days after a developer closes their laptop. The deal, announced June 11, 2026, is the clearest signal yet that the coding-agent race is shifting from "which model writes better code" to "which platform can safely run autonomous work inside an enterprise."
What happened
On June 11, OpenAI said it had agreed to acquire Ona and fold the team into Codex. Ona was founded in 2020 as Gitpod, building browser-based cloud development environments for more than two million developers. It rebranded to Ona in 2025 and pivoted toward secure execution infrastructure for AI agents. Its technology provisions isolated, persistent cloud sandboxes that remain alive across sessions, with customer-controlled security boundaries, scoped credentials, activity logging, and review workflows.
The acquisition is subject to regulatory approvals and customary closing conditions. Until it closes, OpenAI and Ona remain independent.
Why OpenAI wanted Ona
Codex has grown fast — more than five million weekly users, up 400% since the start of 2026, according to OpenAI's announcement. But fast growth exposed a structural limit: most Codex work still happened inside a local session or a short-lived OpenAI-hosted sandbox. That is fine for quick completions or one-shot explanations. It is not fine for the tasks enterprises actually want agents to handle: multi-day code migrations, overnight security sweeps, long-running test suites, or large refactoring projects that span many files and many hours.
Ona solves the persistence problem. Its sandboxes can keep running while the user is offline, resume on a different device, and operate inside the customer's own cloud so security and compliance teams retain control. Hash-based program blocking and automatic teardown of unused environments add further enterprise guardrails. That combination is the difference between a coding assistant and a production agent.
OpenAI's framing was explicit. "Agents need more than intelligence; they need a trusted workspace," Ona CEO Johannes Landgraf said in the announcement. OpenAI Core Products Lead Thibault Sottiaux added that the goal is to help organizations deploy agents "with confidence" at "the highest standards of trust and scale."
Why it matters for the market
This acquisition reframes how buyers should evaluate coding-agent vendors. Model benchmarks still matter, but they are no longer sufficient. Procurement teams will increasingly ask about runtime isolation, permission scoping, audit trails, review workflows, and where data actually lives. The competitive battlefield is moving from the prompt layer to the infrastructure layer.
It also tightens the race with Anthropic. Claude Code has become a leader for long-running terminal-based tasks, and Anthropic has built out its own enterprise hooks. Anthropic had just launched Claude Fable 5 on June 9, only to see it suspended globally on June 12 after a US government export-control directive. OpenAI's timing with Ona — two days after Fable 5's launch — looks deliberate. While Anthropic is fighting a regulatory fire, OpenAI is strengthening the enterprise execution layer underneath Codex.
The parallel to watch is distribution. Anthropic's June 15 billing change splits interactive and programmatic Claude usage into separate credit pools, a move aimed at reining in third-party agent harnesses. OpenAI, by contrast, is expanding Codex through Oracle's enterprise channel and now Ona's infrastructure. One lab is tightening the leash on how its model is consumed; the other is building more on-ramps.
The broader context: OpenAI's busy June
The Ona deal is part of a concentrated run of corporate moves from OpenAI:
- June 8: OpenAI announced a confidential draft S-1 filing with the SEC, saying it expected the news to leak and that it had not decided on timing. The company framed going public as a "complicated set of tradeoffs."
- June 10: OpenAI and Oracle announced a partnership that will let Oracle Cloud Infrastructure customers apply eligible Universal Credits toward OpenAI models and Codex, removing a separate procurement path.
- June 11: The Ona acquisition.
Taken together, the three moves point in the same direction: OpenAI is building the enterprise distribution, governance, and infrastructure layers it needs before a public-market debut. A trillion-dollar valuation story requires more than consumer ChatGPT growth. It requires convincing CFOs, CIOs, and CISOs that OpenAI agents can run safely inside their environments.
What to watch next
Three questions will determine whether the Ona acquisition changes the market or becomes a footnote:
- Integration speed. How quickly does Ona's technology appear inside Codex, and in what form? A separate enterprise SKU would be slower than baked-in persistent workspaces.
- Customer control versus convenience. Enterprises want agents inside their own clouds, but they also want low friction. OpenAI will have to balance those tensions.
- Competitive response. Anthropic, GitHub Copilot, Cursor, Replit, and others all have cloud-execution ambitions. Expect a flurry of feature announcements and possibly acquisitions in response.
The bottom line
OpenAI did not buy Ona to improve code completions. It bought the runtime layer that turns Codex from a helpful assistant into a service that can work unattended inside a company's infrastructure. For developers and engineering leaders, the practical takeaway is that the next generation of coding tools will be judged less by how well they autocomplete a function and more by how safely they can refactor, test, patch, and migrate code while you sleep.
Sources
- OpenAI, "OpenAI to acquire Ona," June 11, 2026: https://openai.com/index/openai-to-acquire-ona/
- OpenAI, "Access OpenAI models and Codex through your Oracle cloud commitment," June 10, 2026: https://openai.com/index/openai-on-oracle-cloud/
- OpenAI, "Confidential submission of draft S-1 to the SEC," June 8, 2026: https://openai.com/index/openai-submits-confidential-s-1/
Last verified: June 15, 2026 UTC.
This article was produced with AI assistance and follows the Shaam Blog how-we-work disclosure and correction standards.
Discussion
0 comments