OpenAI models are coming to Google Cloud

OpenAI and Google Cloud say OpenAI's frontier models and Codex workflows are coming to Google Cloud's Model Garden. Google lists three of them, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, rolling out in private preview to select enterprise customers, with OpenAI pointing to general availability in the coming weeks. Purchases run through Google Cloud Marketplace, show up on a single Google invoice, and draw down existing cloud spend commitments.

That last part is the lever. Large companies sign multiyear cloud commitments and then hunt for ways to burn them down. Letting OpenAI usage count against a Google contract removes one of the biggest frictions between a Google shop and OpenAI's models. Google now claims to be the only hyperscaler offering all three leading closed-source model families on one platform.

We read this as a win for enterprise buyers first and for either company second. Model choice inside one governance boundary is what regulated customers have been asking for. Security teams get one identity system, one audit log and one data perimeter instead of a new vendor review for every model. Google says prompts and outputs are not used to train foundation models and that the OpenAI models run on dedicated Google infrastructure. That is a useful default.

Still, look at the fine print with care. Zero data retention and Confidential Compute are both described as coming soon, not shipping. The OpenAI post talks about general availability in weeks; Google's own post says private preview for select customers. For a bank or a hospital, the gap between a promise on a launch blog and a contractually backed control is the whole ballgame, and I would not move sensitive workloads until those pieces are live.

The strategic picture is stranger. Google sells Gemini, and it is now inviting customers to pick a competitor's model, soon even inside Antigravity and its agent products. That tells us the cloud business has decided the platform and the billing relationship matter more than whose model wins any given task. OpenAI, for its part, is widening distribution well beyond Microsoft while stating plainly that Microsoft remains its primary cloud partner. Both firms are hedging, and models are starting to look like interchangeable line items in a cloud catalog.

If that scales, the downside is concentration of a different kind. The model labs compete, but the toll booth consolidates at the hyperscaler that controls identity, data and invoices. Smaller clouds and independent model hosts have a harder time competing when the frontier labs all sell through the same few marketplaces. We are pro-choice here, and this delivers more choice for customers. Whether it delivers more competition in the market is an open question.

Microsoft has been OpenAI's primary cloud partner, and OpenAI repeats that in this announcement. Google Cloud's Model Garden already hosts more than 200 first-party, third-party and open models, by Google's count, and Google also says developers can connect local Codex CLI and Desktop workflows to its Gemini platform for inference starting today.