OpenAI Models Land on Google Cloud: GPT-6 and Codex Go on Sale in a Rival's Cloud

OpenAI is making its GPT-6 AI models and Codex coding agent available to Google Cloud customers. The models will sit on the same shelf as Google's own Gemini family. The move shows OpenAI has effectively stepped away from the distribution model it built on Microsoft Azure over the years.

What happened?

According to statements from both companies, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna have entered private preview in Model Garden, Google Cloud's model catalog. Select enterprise customers are getting first access. OpenAI plans to move to general availability in the coming weeks. The Codex coding agent is also part of the package.

Why does it matter?

OpenAI's commercial rise was built on its partnership with Microsoft. Microsoft both funded the infrastructure and became the only major cloud offering OpenAI models to enterprise customers, through Azure OpenAI Service. The arrangement worked for both sides: OpenAI got compute, and Microsoft sold a product its rivals didn't have.

But the relationship has loosened step by step over the past two years. To avoid relying on a single supplier for compute, OpenAI signed infrastructure deals with Oracle, Google and Amazon. Its open-weight gpt-oss models had already been released on other clouds. The restructured partnership agreement between the two companies also limited the exclusivity terms. Even so, seeing the most powerful models enter a rival cloud's catalog marks a different threshold. This is no longer just about renting GPUs; it's direct product distribution.

The answer to why Google would make room for a competitor lies in cloud economics. For Google Cloud, the real prize is less the model license than keeping customers' data, storage and workloads on its own platform. If a company has to move to Azure just to access GPT models, Google risks losing that customer. Keeping Model Garden broad prevents that exodus. Having Gemini and GPT side by side in the same catalog also gives Google a direct point of comparison.

Who does it affect, and how?

The clearest winners are enterprise customers. A company whose data lives in BigQuery and whose applications run on Google Cloud will no longer need to manage a separate cloud account, separate billing and separate security audits just to use OpenAI models. At large companies, this kind of friction often determines model choice more than the quality of the model itself.

The picture is more complicated for Microsoft. Azure loses its advantage as the sole home of OpenAI access. On the other hand, Microsoft is pursuing a similar path of diversification. The company is developing its own models and has added Anthropic models to its products. In other words, the decoupling isn't one-sided: both companies are building an arrangement in which they depend less on each other.

Anthropic was the first to blaze this trail. Claude models were sold from the outset through both Amazon Bedrock and Google Cloud's Vertex AI. What's more, both cloud providers are among Anthropic's investors. OpenAI is coming from the opposite direction: starting with a single cloud and opening up to many.

This difference matters, because one reason for Anthropic's rapid growth in the enterprise market was its availability on every cloud. OpenAI's move can be read as an attempt to close that distribution gap.

What's next?

The first thing to watch is pricing. Will GPT models sell on Google Cloud at the same price as on Azure, or will competition emerge between the clouds? Second, whether new models will launch on both clouds at the same time. If Azure retains priority access, the special side of the partnership lives on.

The third question is AWS. OpenAI already has a compute deal with Amazon. Bringing its most powerful models to Bedrock would be the logical next step. If that happens, the relationship between major model developers and cloud providers will have shifted from exclusive partnerships to a marketplace where everyone sells to everyone.

This move is not a break between OpenAI and Microsoft. OpenAI now wants to sell everywhere, and it is strong enough to do so. In the short term, enterprise customers will be the winners. In the long term, model providers' bargaining power over the clouds is growing, while the clouds' power over the models is shrinking.