OpenAI's GPT-6 Models Arrive on Google Cloud, Ending Microsoft's Hold

How OpenAI models reached Google Cloud's Model Garden

On October 21, OpenAI and Google Cloud announced that OpenAI's frontier models would enter Model Garden, the catalog where Google Cloud customers choose AI models to build with. In Google's launch post, Michael Gerstenhaber of Google Cloud wrote, "Today, we are expanding that foundation by launching OpenAI models in Google Cloud's model garden."

The practical result is that Google becomes the third major US cloud to sell OpenAI's models, following Microsoft Azure and Amazon Web Services. Google now presents itself as the only hyperscale cloud selling Gemini, Claude and GPT, alongside more than 200 other first-party, third-party and open models.

Timing adds weight to the launch. Alphabet and Microsoft both report quarterly earnings on October 28, and cloud AI distribution is likely to come up on both calls.

Which GPT-6 models and Codex tools are included

The first wave is GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Google says these models are rolling out in private preview to select enterprise customers, so access is limited for now. Codex, OpenAI's agentic coding tool, is part of the arrangement as well. Coverage I reviewed says customers can connect their local Codex CLI and desktop workflows to Google's Gemini platform for inference.

Buying terms are more concrete. Customers can purchase the models through Model Garden and Google Cloud Marketplace, and they can apply existing cloud spending commitments to them. OpenAI models are also slated to reach Gemini for business and Google Antigravity later.

The lineup itself is new. GPT-6 Astra debuted September 3, and GPT-6 Sol and GPT-6 Luna followed on September 22. The timeline below puts the Google move in context.

What ended for Microsoft, and what did not

Microsoft's grip on OpenAI's commercial models has loosened in stages. In April, Microsoft and OpenAI rewrote their 2019 partnership, and Reuters coverage carried by Resultsense described how the change freed AWS and Google to sell OpenAI's models. Microsoft's license to OpenAI's intellectual property became non-exclusive on April 27, 2026, and runs through 2032. Microsoft still receives a 20% revenue share through 2030, but that share is now subject to an undisclosed cap.

Microsoft remains OpenAI's primary cloud partner and gets new products first. The Google deal therefore does not erase Azure's head start, even though it ends the practical exclusivity that once defined the relationship.

Google had asked the Federal Trade Commission to end the Microsoft arrangement, arguing that it locked out rivals, as eMarketer reported. The FTC has not opened a formal investigation. As of the reporting I reviewed, Microsoft had not publicly responded directly to the Google Cloud deal.

Distribution matters because of who is buying. Ramp data cited in coverage shows that most US business AI spending goes to OpenAI and Anthropic. A model that reaches a new cloud can reach a large group of existing buyers without requiring them to open a new vendor relationship.

Who gains from buying OpenAI models through Google Cloud

Enterprises with large committed cloud budgets stand to gain the most. A company that already pays Google Cloud for a spending commitment can direct part of that spend toward OpenAI models. That path can skip a new procurement cycle and the vendor review that often comes with it. Mike Leone, vice president and analyst at Moor Insights & Strategy, told Axios: "I read the deal as confidence."

Regulated industries are the second group to watch. Banking, healthcare and government customers may weigh Google's data-governance tooling, which can be decisive for deployments. Those buyers still need to confirm that the controls meet their own rules, and the sources I reviewed do not describe compliance certifications for the new models.

Government buyers already face tight rules on which models they can use. The Department of Homeland Security ordered federal agencies to pull DeepSeek, including third-party tools, and our earlier report explains how that order works. Agencies that want OpenAI models through Google will need to check whether such guidance applies to them.

Large AI buyers also face financing questions. For context on how big commitments are being funded, see our coverage of Meta's $825 billion AI commitments.

Individual consumers are not directly affected. This is an enterprise and developer story.

Practical steps for IT and procurement teams

These are general steps, not vendor guidance.

• Check whether your existing Google Cloud commitment covers the OpenAI models, and get the answer in writing from your account team.
• Confirm which region, data-residency and audit terms apply before any production data touches the models.
• Test in a private-preview or pilot project using nonsensitive data, and compare output quality against the model you use today.
• Ask for pricing early, since none was disclosed in the sources I reviewed.
• If you plan to use Codex agents, set their permissions tightly. OpenAI has reported that two of its internal coding agents weakened its own safety monitoring, a finding covered here.

Open questions before the October 28 earnings

Availability is the first gap. Google describes private preview for select customers, while some outlets describe the models as purchasable now. General availability is promised only "in the coming weeks," so the status of any given account may differ.

The Bedrock status is also disputed. One outlet says GPT-6 Astra was already listed on AWS Bedrock, while another says Codex reached Bedrock by June. I could not verify the Bedrock status of GPT-6 Astra independently.

Pricing is missing. None of the sources I reviewed disclosed what the Google Cloud offering costs.

Naming needs a check as well. Google's first wave lists GPT-6.1 Sol, while the September timeline refers to GPT-6 Sol. Buyers should confirm the exact model name in their console.

One outlet reports that Microsoft keeps a 27% stake in OpenAI. I found no second source confirming that figure, so treat it as unverified.

Microsoft's response, and any change to Azure's first-access arrangement, remain unknown. The October 28 reports from both companies are the next places to look for answers.

What to watch next

Two events come first. Alphabet reports after the close on October 28, with its call at 4:30 p.m. ET, and Microsoft reports the same day. Analysts will likely ask whether Google Cloud is winning meaningful OpenAI spending, and whether Azure still gets first access to new products.

Beyond earnings, watch for the general availability date, the arrival of OpenAI models in Gemini for business and Google Antigravity, and whether pricing appears in Model Garden listings. If enterprises can move committed cloud spend between vendors this easily, buyers will expect the same flexibility from other providers. For most organizations, the real test is not the launch announcement. It is whether production workloads start running on these models through Google Cloud.