Inside OpenAI's Ona Acquisition: Persistent Cloud Agents and the 5M User Codex Surge

From Gitpod to Ona: The Rebrand and Platform Capabilities

Ona is not a new company. It was originally founded in 2020 as Gitpod, a browser-based cloud development environment that attracted over 2 million developers . On September 2, 2025, the company rebranded to Ona and made a deliberate pivot: it moved from an IDE-centric product to what it calls "mission control for your personal team of autonomous software engineering agents" .

This was a response to the rise of powerful large language models capable of writing, testing, and debugging code . The vision was to transform the developer's role from a hands-on coder into a "software conductor" — a strategist who plans, delegates, and reviews the work of AI agents .

Key Platform Capabilities

Ona's platform is built around three core components designed to manage the entire AI agent lifecycle securely .

• Sandboxed, VPC-native environments: Agents run in ephemeral, API-first cloud environments within a customer's own VPC, with full OS-level isolation and kernel-level governance. Ona maintains SOC 2 compliance, making it viable for Fortune 500 code migrations and CVE remediation at scale .
• Persistent, async agent execution: The core model is "task in, pull request out." Engineers submit bug fixes, feature requests, or code migrations, and Ona agents work autonomously for hours or days, reviewing and deploying code before returning a pull request .
• High-performance infrastructure: Environments support up to 32 vCPUs, 128 GB of RAM, and 200 GB of disk, with GPU support available for larger workloads — a significant upgrade over Gitpod Classic's 8-core limit .
• Full VS Code in the browser and mobile access: Developers can check progress and review results from any device, including phones, through a full VS Code interface running in the browser .

On its own engineering teams, Ona's agents reportedly co-authored 60% of merged pull requests and contributed 72% of merged code . The company said its agent session count had grown 13x in 2026 alone .

Why Ona: Codex Needs a Persistent Home

The strategic rationale behind the deal is straightforward. Codex has evolved from a coding assistant into an agent that knowledge workers and developers use for tasks spanning hours or days — research, analysis, workflow automation, and app building. Until now, these agents stopped working the moment a user closed their laptop .

Ona's technology solves this by providing secure, persistent cloud environments where agents can continue running, accessing the tools and context they need over time . OpenAI said the integration will "expand Codex beyond work tied to a single device or active session and help more organizations deploy agents securely in production" .

Thibault Sottiaux, OpenAI's core products lead, said the deal will make Codex "easier to deploy securely across production workflows for customers operating at the highest standards of trust and scale." Ona CEO Johannes Landgraf wrote on LinkedIn that the acquisition broadened the company's "life's work" .

The Context: Codex Surpasses 5 Million Weekly Users

Nine days before the Ona announcement, on June 2, 2026, OpenAI revealed that Codex had crossed 5 million weekly active users — up more than 6× since the desktop app launched in February .

Two dynamics stand out. First, knowledge workers — analysts, marketers, designers, and operators — now represent roughly 20% of Codex users and are growing more than three times faster than developers . Second, Codex is increasingly a tool for white-collar workflows, not just code. Users create reports, spreadsheets, presentations, contracts, and lightweight tools that previously required engineering support .

At the same "Intelligence at Work" livestream, OpenAI launched six role-specific plugins, a hosted app-publishing feature called Codex Sites, and an Annotations system for editing individual sections of documents . Codex now integrates with over 90 partner applications, including Canva, Figma, Slack, Notion, and Spotify .

The explosive user growth made the infrastructure question urgent: if millions of people are deploying agents for long-running work, those agents need somewhere to live. Ona is that place.

Enterprise Strategy: Security, Persistence, and Production

OpenAI's core enterprise challenge is moving AI agents from prototyping sandboxes into production environments where security, compliance, and governance are non-negotiable. Ona fits directly into that playbook.

Ona's customer-controlled VPC model means enterprises keep their data and infrastructure boundaries intact while Codex agents do the work. Agents get persistent, secure environments with audit logging and kernel-level isolation — exactly what regulated industries require .

This deal also signals how OpenAI is positioning Codex as a platform, not just a tool. By buying the orchestration layer, OpenAI ensures it can offer an integrated experience: the model, the agent, and now the persistent cloud environment where work actually gets done.

The 2026 Acquisition Sprint

The Ona deal is not an isolated move. OpenAI dramatically accelerated its M&A engine in 2026. By late March, it had already completed six acquisitions in Q1 2026 alone — nearly matching the eight deals it closed across all of 2025 .

According to Crunchbase data, OpenAI has acquired 17 companies over the past three years, spanning developer tools, security, healthcare, media, and cloud infrastructure . Notable deals include:

• io Products ($6.5 billion, 2025) — enterprise deployment and hardware design
• Statsig ($1.1 billion, 2025) — feature management and experimentation
• Torch (January 2026, ~$60–$100 million) — healthcare technology and unified medical memory
• Astral (March 2026) — open-source Python developer tools
• Promptfoo (March 2026) — red-teaming and AI agent security testing
• Hiro (2026) — personal finance infrastructure
• Convogo (January 2026) — GenAI solutions and predictive analytics consultancy

Each deal fills a specific gap across the AI value chain, from hardware design to model training tools to enterprise deployment . The Ona acquisition adds cloud execution infrastructure — a critical layer as agents become more autonomous and persistent.

The IPO Filing Looms

On June 8, 2026 — three days before the Ona announcement — OpenAI confidentially submitted a draft S-1 registration statement to the SEC for an initial public offering . The confidential filing, permitted under the JOBS Act, lets OpenAI work through SEC review privately before making its financials public.

The IPO context sharpens the rationale for the deal. When a company is about to face public market scrutiny, acquiring infrastructure that demonstrates enterprise-grade security, recurring usage, and production capability is not just strategic — it is existential. Ona gives OpenAI a concrete answer to the question every public-market investor will ask: How do your agents actually run in a real enterprise?

The picture is clear: OpenAI is racing to build a complete, defensible enterprise platform ahead of going public. Ona is the latest — and one of the most strategically telling — pieces of that puzzle.