What the OpenAI Ona acquisition means for Codex enterprise agents
The OpenAI Ona acquisition is a move to embed secure cloud execution into Codex so enterprise AI agents can run long, complex workflows inside customer‑controlled environments. OpenAI has agreed to buy cloud development infrastructure company Ona and integrate its orchestration technology with Codex, the firm’s agent platform for software and knowledge work. This will let Codex agents continue working for hours or days inside persistent cloud environments instead of being tied to a single device or short user session. OpenAI reports that more than five million people now use Codex each week, a 400% rise from earlier in 2026, as organisations apply it to research, analysis, software development and workflow automation. By adding Ona’s persistent, reproducible environments, OpenAI aims to shift Codex from helpful assistant to production‑grade engine for AI agent workflows across enterprise systems.
From short sessions to long-running secure cloud execution
Codex originally focused on helping developers write code, but OpenAI is now pushing it toward end‑to‑end assignments that can span days. Many current agents stop when a laptop closes or a browser session times out. Ona’s secure cloud execution technology is designed to remove that limit by giving agents a persistent workspace in the cloud where they can keep running independently. Sarah Friar, Chief Financial Officer at OpenAI, wrote that the deal will "bring secure cloud infrastructure into Codex so agents can keep working for longer — over hours or days, not just minutes." In practice, this means agents can run tests, resolve software issues, address vulnerabilities, modernize applications and complete complex workflows while users come back periodically to check progress, give feedback and approve results from any device.
Customer-controlled enterprise cloud infrastructure and governance
A central promise of the OpenAI Ona acquisition is customer‑controlled enterprise cloud infrastructure for Codex enterprise agents. Instead of running agents solely in vendor‑managed environments, organisations will be able to execute Codex agents inside their own cloud accounts. OpenAI will supply models and orchestration, while customers define where workloads run, which internal systems agents can access, how credentials are scoped, what activity is logged and when tasks are routed for human review. Ona has already supported around two million developers moving work from local machines into reproducible cloud setups, experience that transfers directly to governance, auditing and compliance for AI agent workflows. This approach targets companies moving from isolated pilots into production, where questions about security boundaries, access control and change management become as important as model accuracy and response quality.
Competing in the enterprise cloud race for agentic AI
By folding Ona into Codex, OpenAI signals that success in agentic AI will depend on more than model quality; it will require a complete operational stack. Enterprise buyers increasingly compare AI platforms on their ability to integrate with existing cloud estates, respect internal governance rules and support long‑running workloads without giving up control. Ona’s customer‑controlled execution model positions OpenAI to compete directly with enterprise cloud platforms that already offer secure environments for AI workloads. Thibault Sottiaux, Core Products Lead at OpenAI, said enterprises "want powerful agents that can do real work while meeting the security and control requirements of their environments." As AI agents shift from experiments to production systems that touch real code, data and processes, this deal helps OpenAI present Codex as a serious option for mission‑critical automation.
Sovereign, secure environments and the future of AI agent workflows
The OpenAI Ona acquisition reflects a broader move toward sovereign and secure cloud environments for AI workloads, where customers keep control over infrastructure and data while vendors provide specialised AI capabilities. For Codex, that means giving enterprises a trusted workspace for agents that can live across the full software lifecycle, from development and testing through to maintenance and modernization. Ona’s focus on reproducible environments, access controls and collaboration aligns with demands from teams that need AI agents embedded in production workflows, not isolated sandboxes. Johannes Landgraf, Ona’s co‑founder and CEO, summed this up by saying that "agents need more than intelligence; they need a trusted workspace." If regulators approve the deal, Ona’s team will join OpenAI’s Codex organisation to build out these secure, persistent execution paths for enterprise customers.
