OpenAI arrives in Google Cloud, sign of a more open AI market

OpenAI will offer its models on Google Cloud, initially in a private preview for enterprise customers. This move confirms the end of exclusivities and the rise of a more agnostic AI cloud.

OpenAI will make its models available via Google Cloud, after years of rivalry between the two groups. The announcement, made on October 21, 2026, initially concerns enterprise customers in private preview. It matters for the market because it places, for the first time, OpenAI in the three major American clouds, while the exclusivity with Microsoft ended in April.

A partnership that mainly says something about the market

The rapprochement is not a love match, but it says a lot about the state of the sector. Google Cloud wants to sell an open platform, capable of hosting several families of models, while OpenAI is expanding its commercial distribution beyond its first historic ally. As a result, companies can now purchase OpenAI models via Google Cloud Marketplace and have them billed through their existing cloud commitments.

In detail, Google Cloud says it is adding GPT-Astra, GPT-6.1 Sol, and GPT-6 Luna to its model garden, a catalog of AI models accessible to technical teams. The first targets the most demanding uses, from coding to scientific work; the second seeks a balance between cost and capability; the third presents itself as a fast and economical option for large-scale routine tasks.

Why companies see immediate value

For CIOs and product teams, the benefit is simple: less contractual lock-in and more flexibility in model choice. Google Cloud explains that customers can switch models based on performance, latency, or cost, without having to rebuild their architecture at each switch. In a context where AI budgets are tightening, this flexibility weighs heavily.

The announcement also emphasizes governance guarantees. Prompts, context, and responses remain within the client’s data perimeter and are not used to train foundation models. Google Cloud also mentions dedicated access, isolated gateways, unified audit logs, and authenticated identity for each interaction. Zero Data Retention and Confidential Compute are announced as coming soon.

The signal sent to Microsoft, Amazon, and the rest of the sector

This openness confirms a shift already underway since the end of the exclusivity agreement between OpenAI and Microsoft in April. Since then, OpenAI has also expanded its presence to Amazon Web Services, placing it now on the three major American hyperscalers. For Google Cloud, the challenge is twofold: to show it can host models competing with Gemini and to attract companies wanting to diversify their providers.

OpenAI, for its part, gains broader distribution at a time when the group seeks to accelerate its monetization. The company is reportedly targeting a valuation of 1.4 trillion dollars [to be verified] in anticipation of an IPO. No small ambition, then. And a clear message sent to the market: in enterprise AI, the model war also runs through the cloud pipelines.

Key points

  • OpenAI arrives on Google Cloud Marketplace on October 21, 2026.
  • Three models are in private preview for enterprise customers.
  • Google Cloud claims over 200 models on its platform.
  • OpenAI is now present on the three major American clouds.
  • Distribution via cloud follows the end of Microsoft exclusivity in April.

By the numbers

  • 3 — major American clouds now hosting OpenAI.
  • 200+ — models available on the Google Cloud platform, according to the October 21, 2026 announcement.
  • 47% — share of Anthropic’s revenue through cloud marketplaces in 2025, according to its IPO documents.
  • 82% — annualized revenue growth of Google Cloud in Q2 2026, according to the group.
  • 1.4 trillion dollars — valuation targeted by OpenAI, according to the cited source [to be verified].

To read