OpenAI promises Dots for Europe 'in a few weeks' as Berlin DevDay pushes platform lock-in

OpenAI's AI assistant Dots should roll out more widely in Europe "in a few weeks" following talks with regulators, Romain Huet, the company's head of developer experience, told developers at DevDay Exchange 2026 in Berlin. The date is a promise, not a release. Personal Pro accounts in the European Economic Area, Switzerland, and the UK still cannot access the service, while certain business accounts already can.

The Berlin event mostly restaged announcements OpenAI had already made in San Francisco. The new material was thin: a timeline hint for Dots, fresh growth figures for the Codex coding tool, and a question-and-answer session where management explained what the company has chosen not to build. The venue itself carried a message. Europe is a market OpenAI is still negotiating its way into, one regulator conversation at a time.

How fast is Codex actually growing in Germany?

Weekly active Codex users in Germany have grown twentyfold since January 2026, according to OpenAI's own figures. That number deserves scrutiny. In June, OpenAI reported a more than sevenfold increase over the same baseline, and in both cases the company declined to share absolute user counts. A twentyfold rise from an undisclosed base could mean anything from explosive adoption to a modest climb off a small starting point. The claim is OpenAI's, and it has not been independently verified.

One usage detail adds context. OpenAI says nearly one in five desktop users taps skills for documents, presentations, or spreadsheets — meaning they use the coding tool for general office work rather than programming. That blurs the line between a developer product and a productivity app, and it suggests OpenAI is measuring engagement across a broader user base than coders alone. Twentyfold growth, in plain words, is a strong relative figure attached to an undisclosed absolute one.

Why is there still no video understanding?

Asked directly during the Q&A why OpenAI still offers no model that understands video sequences the way Google's Gemini does, management gave a demand-side answer: customers have barely asked for it. The company has therefore not prioritized the feature beyond single-frame analysis. The gap is a competitive one. Google markets video understanding as a core Gemini capability, and OpenAI has effectively conceded that ground, at least for now.

The panel also revisited the shutdown of the Sora video generator in March. Managers again blamed extreme resource demands, and said customers mainly want coding models and agents instead. OpenAI did not rule out reviving Sora later. The stated reason is cost; the strategic reason, given everything else presented in Berlin, is that OpenAI is spending its compute where its priorities now sit.

Those priorities carry their own infrastructure bill. Because agents run on their own virtual computers, and AI-generated code is growing fast, demand for traditional processor power — CPUs, not just AI accelerators — is rising. OpenAI said it maintains a dedicated team that combs through its internal codebases every week specifically to cut CPU load. Agent infrastructure is not cheap to run, and the company is actively managing that cost.

What is the real pitch to developers?

Beyond product updates, the event's substance was strategic: OpenAI wants developers locked into its platform, not just its models. The Agents API means OpenAI runs the agent infrastructure itself. Plugins and web apps give developers more channels to sell their own products inside OpenAI's environment. The trade on offer is less technical work in exchange for dependence.

The distribution argument rests on a large number: OpenAI says ChatGPT has 1.2 billion weekly users. That figure, like the Codex growth metric, comes from the company itself. The model resembles earlier app stores. Developers build the apps; OpenAI supplies the infrastructure and the distribution channel — and, by extension, the terms.

There is an open-source element, and it is narrower than it sounds. The underlying runtime, Codex Harness, remains open source, so developers can read the code and run it themselves. But anyone who relies on OpenAI's managed services, platform-specific integrations, and customer acquisition through ChatGPT will likely find switching ecosystems expensive later. Openness at the runtime layer does not undo lock-in at the infrastructure and distribution layers.

Why does Berlin matter beyond the product news?

The regulatory backdrop gives the Dots timeline its weight. Personal Pro accounts across the European Economic Area, Switzerland, and the UK remain locked out of the assistant while business accounts get selective access — a pattern that suggests ongoing negotiation with European regulators rather than a technical delay. "In a few weeks" is Huet's phrasing, and OpenAI has not published a fixed date. Readers should treat the timeline as intent, not commitment.

The stakes run in both directions. For European users, Dots access is a question of when a flagship assistant becomes available under EU rules. For OpenAI, Europe represents both a large user population and a regulatory environment that has already shaped how its products launch there. The DevDay format itself — a restaged San Francisco event with a few European specifics — signals a company calibrating its rollout to that environment.

The strategy underneath is the story. OpenAI's message in Berlin was not only that its models keep improving. The company wants developers to build their businesses directly on OpenAI infrastructure, with the Agents API, plugins, web apps, and ChatGPT's stated 1.2 billion weekly users as the bait. If the app-store analogy holds, the next phase of OpenAI's growth will be measured less in benchmark scores and more in how many businesses find it too costly to leave.