OpenAI acquires cloud startup Ona to secure enterprise AI agents
··5 min read
The acquisition of the German infrastructure provider addresses a critical governance bottleneck for deploying long-running coding agents inside corporate networks.
OpenAI's acquisition of the Kiel-based cloud infrastructure startup Ona signals a fundamental shift in how the technology industry is packaging artificial intelligence for corporate buyers. The deal, announced in early June 2026, moves the competitive battleground away from raw model intelligence and towards secure deployment environments.
For heavily regulated industries across the DACH region and the broader European market, the capability of an artificial intelligence model is secondary to how it is governed. Enterprise technology leaders want AI to perform complex, multi-day engineering tasks, such as codebase modernisation, legacy system migration, or exhaustive security audits. However, risk-averse chief information officers will not allow autonomous software agents to roam their proprietary networks without strict boundaries and verifiable audit trails.
By integrating Ona's technology, OpenAI is addressing this bottleneck directly. The acquisition acknowledges that raw processing power is useless to industrial and financial enterprises without robust, secure deployment infrastructure.
The transition to persistent agents
The acquisition highlights a necessary transition from simple, prompt-based AI copilots to persistent, autonomous agents. Historically, coding assistants have been tethered to a single developer session. If a software engineer closed their laptop or lost their internet connection, the AI stopped working.
Ona, which originally operated as the developer tool company Gitpod before pivoting to AI infrastructure, solves this problem by providing secure, sandboxed cloud workspaces. These micro-environments allow OpenAI's Codex agents to execute multi-step tasks over hours or days without human intervention. Through automatic checkpointing and persistent memory, an agent can compile code, run test suites, and generate pull requests overnight, presenting the finished work when the human developer logs back in.
Crucially for enterprise adoption, these workspaces offer what Ona calls customer-controlled execution. The data, credentials, and access controls remain firmly within the customer's own virtual private cloud. The orchestration and the underlying model are provided by OpenAI, but the actual execution happens inside a fenced-in corporate environment where security teams can monitor audit trails, scope credentials, and enforce strict role-based access.
This infrastructure layer is the missing piece for Codex, which has experienced rapid adoption across global enterprises. The platform now supports more than five million weekly active users, representing a 400 percent increase since the start of the year. As tasks stretch from quick code completions to complex system refactoring, the need for a durable and secure execution space has become urgent for OpenAI's commercial viability.
Securing the enterprise market
The regional landscape in Europe has long demanded robust data sovereignty and strict deployment infrastructure. The integration of Ona's 80-person engineering team, led by chief executive Johannes Landgraf, demonstrates OpenAI's intent to satisfy these regional compliance standards while scaling its global enterprise footprint. German cloud providers and infrastructure startups have historically excelled at navigating these regulatory constraints, making Ona a strategic asset for a US-based company looking to deepen its penetration into European enterprise accounts.
This move also intensifies the competition with rival artificial intelligence providers. Companies such as Anthropic have made their own aggressive pushes into the enterprise market with tools like Claude Code, competing for the same lucrative corporate development contracts. OpenAI's decision to buy its way into persistent cloud infrastructure rather than build it from scratch underscores the pressure to lock down the enterprise execution layer quickly.
The broader industry is learning that an autonomous agent is only as useful as the trust it commands. It is not enough for a model to write perfect code if the execution environment introduces critical vulnerabilities, exposes proprietary intellectual property, or violates strict data protection policies.
As the Ona team integrates with the Codex group, pending standard regulatory approvals, the value proposition of enterprise AI is being fundamentally rewritten. Software vendors will no longer win corporate contracts based solely on benchmark scores or theoretical reasoning capabilities. The next phase of enterprise AI adoption will be won by the platforms that can prove their autonomous tools operate safely, persistently, and strictly within the boundaries set by their corporate customers.
Sources
Forbes. "OpenAI Buys Ona To Run Codex Agents Inside Enterprise Clouds." June 2026.
Tech Jacks Solutions. "OpenAI Acquires Ona to Lock In Enterprise Cloud Execution." June 2026.
Digital Applied. "Why OpenAI Bought Ona: Cloud Execution for AI Agents." June 2026.
InfoWorld. "OpenAI buys Ona to help rein in AI agents." June 2026.
ETIH EdTech News. "OpenAI to acquire Ona for long-running Codex agents." June 2026.
