OpenAI's models are coming to Google Cloud, billed against the commitment you already have

Google Cloud and OpenAI announced on Wednesday that OpenAI's current models will be sold through Google Cloud. The list has three models: GPT-6 Astra for the hardest coding, computer use and research work, GPT-6.1 Sol, which Google describes as close to Astra at a lower price, and GPT-6 Luna, a fast and cheap model for everyday volume. OpenAI's Codex coding agent is part of the deal too.

Google now says it is the only large cloud provider that offers all three leading closed models on one platform. That is Google's claim, but the practical point stands: a company that already runs on Google Cloud can now test all three without a new vendor contract.

What is actually available today

Less than the headline suggests. Google says the models are rolling out in private preview to selected enterprise customers, and everyone else fills in a sign-up form. Google gives no date for general availability. Developers who use Codex on the command line or in the desktop app can point it at Google's platform for inference. OpenAI models in Google's Gemini app for business and in the Antigravity coding tool are marked as coming soon, without a date.

The money and the data

The billing detail matters most for finance teams. OpenAI usage runs through Google Cloud Marketplace, lands on the same Google Cloud invoice and draws down existing cloud commitments. A company that has already promised Google a yearly spend can use part of it on OpenAI models instead of opening a separate budget line.

On data, Google says prompts, context and outputs stay inside the customer's own data perimeter and are not used to train models. Access runs under named identities, VPC Service Controls and a shared audit log. Two protections are still only promised: zero data retention and Confidential Compute are both listed as coming soon. The announcement does not say which regions the models run in or what they cost. Earlier this week we wrote that Claude in Gemini Enterprise now runs in the EU region. For OpenAI on Google Cloud, nobody has said that yet.

What to do on Monday

• ✳If you have a Google Cloud commitment, ask finance how much of it will go unused this year. That number tells you whether testing OpenAI models through Google costs you anything extra
• ✳If your developers already use Codex or the OpenAI API, ask your Google account manager for preview access instead of signing a second contract with OpenAI right now
• ✳Do not send personal data or client documents to these models until Google publishes the regions and states that processing stays in the EU. Write that rule down for the team before the preview opens
• ✳Pick three real tasks from your own work and run them on Astra, Sol, Luna, Gemini and Claude. The best model for a contract summary and the best model for code are rarely the same one, and now you can compare them on one bill
• ✳Decide who in the company may switch these models on. When a new model arrives in the console, someone will try it the same day

The bigger change is in procurement. Choosing an AI vendor used to mean choosing a cloud. For a Google Cloud customer that link is now weaker, so the question to answer is which model does which job best for you, and that is something you can only learn by testing on your own tasks.