OpenAI brings its GPT-6 models to Google Cloud, the only cloud that now sells GPT, Claude and Gemini

On 21 October 2026, OpenAI and Google Cloud announced a deal that would have sounded unlikely only a year ago. OpenAI's newest frontier models are coming to Google Cloud's Model Garden, the catalogue where Google sells its own Gemini models next to more than 200 first party, third party and open models. Google Cloud customers will be able to buy GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna through the Google Cloud Marketplace, pay for them on their normal Google invoice and count the spending toward contracts they have already signed.

With this step, OpenAI's models are now offered on all three of the largest US cloud platforms: Microsoft Azure, Amazon Web Services and Google Cloud. And Google can now say something none of its rivals can say: it is the only hyperscale cloud provider that offers all three leading closed model families, OpenAI's GPT, Anthropic's Claude and its own Gemini, on a single platform.

This article explains what exactly was announced, what is available today and what is only promised, how the deal grew out of the end of Microsoft's exclusive license in April, why both companies want it, and what it means if you build software with AI.

The short version

• What: OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are coming to Google Cloud's Model Garden, and Codex workflows can run inference through Google's platform.
• Status: a private preview for selected enterprise customers started on 21 October. General availability is planned "in the coming weeks", according to OpenAI.
• Billing: purchases run through the Google Cloud Marketplace, appear on one Google Cloud invoice and draw down existing Google Cloud commitments.
• Later: OpenAI models will also appear in Google Antigravity, Google's agentic coding tool, and in Gemini for business (the Gemini Agent).
• Microsoft: OpenAI stresses that Microsoft remains its primary cloud partner.
• Context: Amazon started offering OpenAI models in April, one day after OpenAI and Microsoft changed their contract.

What exactly was announced

The details come from two posts published at the same time: one on the OpenAI blog and a longer one on the Google Cloud blog by Michael Gerstenhaber, vice president of product management for Gemini at Google Cloud.

Google lists three OpenAI models for the launch. GPT-6 Astra is described as the model for demanding work across coding, computer use, professional tasks and scientific research. GPT-6.1 Sol offers "near Astra intelligence at lower cost" for agentic coding, computer use and professional work. GPT-6 Luna is the fast and cost efficient option for everyday work at scale. In other words, Google is not getting a cut down or older line up. It gets the same three tiers that OpenAI sells directly.

The models are rolling out now in private preview to selected enterprise customers, and interested companies can request access through a form. Starting on launch day, customers in the preview can build with the models through Google's APIs and connect their local Codex CLI and Codex Desktop workflows to Google's Gemini platform for inference. That detail matters for developers: a team that already uses Codex as a coding agent can keep its tools and send the model calls through its Google Cloud account instead of a separate OpenAI contract.

Two more integrations are announced but not live yet. OpenAI models will come to Google Antigravity, Google's agent based coding environment, and to Gemini for business, where the Gemini Agent works for employees. Ars Technica points out that the Antigravity integration may be the more interesting part for individual developers, because it puts GPT models inside a Google coding product that today mostly offers Gemini and Claude.

How OpenAI got here: the end of Microsoft's exclusive license

For most of OpenAI's commercial history, if a company wanted OpenAI's models through a big cloud provider, there was exactly one option: Microsoft Azure. Microsoft held an exclusive license to OpenAI's technology as part of the partnership that started in 2019 and grew into a multibillion dollar investment.

That changed on 27 April 2026. In a post titled "The next phase of the Microsoft OpenAI partnership", OpenAI described an amended agreement with four key points. Microsoft keeps a license to OpenAI's models and products through 2032, but that license is now non exclusive. Microsoft remains the primary cloud partner, and OpenAI products ship first on Azure unless Microsoft cannot or chooses not to support them. OpenAI may now serve all of its products to customers on any cloud. And the money flows were reshaped: Microsoft no longer pays a revenue share to OpenAI, while OpenAI keeps paying Microsoft a revenue share through 2030, at the same percentage but with a total cap.

Amazon moved within a day. On 28 April, AWS announced that OpenAI's latest models, Codex and "Managed Agents powered by OpenAI" were available on Amazon Bedrock in limited preview. Customers could use their normal AWS identity, logging and encryption controls, and apply the spending toward AWS commitments. The Google deal follows almost exactly the same template, six months later.

The relationship with Google is not completely new either. Last year OpenAI started using Google's cloud capacity to run ChatGPT and its API, as CNBC recalls. Back then Google was a supplier of computing power. Now it also becomes a sales channel.

Google becomes the model supermarket

The most quoted line from Google's post is the claim that Google Cloud is "now the only hyperscale cloud provider offering access to all three leading closed source LLMs". Claude has been available in Google Cloud for a long time, and Gemini is Google's own model. With GPT added, a Google Cloud customer can in theory route each task to whichever of the three performs best, without signing new vendor contracts.

Google frames this as a question of matching the right model to each workload. Its post says customers can "dynamically select and swap the optimal model for any use case based on performance, latency, and cost efficiency without architectural lock in or separate vendor contracts". Because OpenAI models are sold through the Google Cloud Marketplace, the usage also counts toward Google's Flexible Savings Plans across model providers.

For companies with large existing cloud commitments, this is the practical heart of the deal. Many enterprises have signed multi year agreements to spend a fixed amount with one cloud provider. Until now, using GPT models meant a separate budget line with OpenAI or Microsoft. Now that money can come out of a commitment the company already has to spend anyway. As Reuters notes, cloud providers usually take a cut of sales through their platforms, so both sides earn from the arrangement.

Data governance: what Google promises

The second big selling point is security and compliance, aimed at banks, insurers, healthcare and the public sector. Google's post makes several specific promises.

Customer prompts, context and outputs stay inside the customer's designated data perimeter and are never used to train foundation models. Zero data retention options are listed as "coming soon". The OpenAI models run on dedicated Google Cloud infrastructure, and Confidential Compute is also "coming soon". Queries pass through isolated gateways so that data stays inside Google Cloud, and Google says third party model providers have no unauthorized access to the customer's cloud environment. Every interaction runs under an authenticated identity, with VPC Service Controls and unified audit logging.

It is worth reading the small words carefully. Two of the strongest protections, zero data retention and Confidential Compute, are not available yet. Companies with strict requirements should check what applies on the day they start, not what the launch post promises for later.

Box is the first named customer. Yashodha Bhavnani, Box's vice president of AI products, said using OpenAI models through Google Cloud will let Box give its customers access to OpenAI's frontier models together with Google Cloud's security and compliance features.

Why OpenAI wants this

OpenAI started as a consumer company with ChatGPT, but the bigger money in AI is in business use. According to data from corporate card company Ramp cited by CNBC, most business spending on AI already goes to OpenAI and Anthropic. To grow that share, OpenAI needs to be wherever enterprise developers already work, and a large part of them work in Google Cloud.

Greg Brockman, OpenAI's president and co founder, said the partnership is "an important step in making our most capable models and Codex available where customers are already building". There is also a financial backdrop. CNBC reports that OpenAI is courting investors at a valuation of about $1.4 trillion and is targeting an initial public offering. A broader distribution network is a good story to tell before an IPO.

Anthropic shows how valuable this channel can be. According to Reuters, which reviewed Anthropic's IPO filing, 47 percent of Anthropic's 2025 revenue came through the cloud platforms of Amazon and Google. OpenAI had largely left that path to Anthropic outside of Azure. Now it is closing the gap.

Why Google wants this

At first glance it looks strange that Google would sell the models of its biggest AI rival right next to Gemini. The logic becomes clearer when you look at what Google actually earns. Mike Leone, an analyst at Moor Insights and Strategy, told Axios that customers will still pay Google for its cloud platform, data services and security, regardless of which model they choose. He reads the deal as a sign of confidence, "because Gemini now has to win those customers with OpenAI sitting right next to it."

Google Cloud is also on a strong run. CNBC reports that growth in Google's cloud segment accelerated to about 82 percent in the second quarter from 63 percent in the first, helped by demand for AI models and custom chips. Microsoft and Amazon reported second quarter cloud growth of 42 and 37 percent. All three report third quarter results next week.

Thomas Kurian, chief executive of Google Cloud, said the company wants to give customers "a broad selection of models, including those from OpenAI, so they can choose the right model for any given job". A customer who might have moved workloads to Azure or AWS just to use GPT can now stay with Google.

There is a cost, though. The Decoder points to a recurring tension inside Google: computing capacity used to serve cloud customers, including those running rival models, is capacity the Gemini team cannot use for its own work.

What it means for companies and developers

For a team that already runs on Google Cloud, the change is mostly practical. You can test GPT models next to Gemini and Claude without a new contract, compare quality, speed and price on your own tasks, and move a workload between models when one becomes better or cheaper. If you use Codex, you can point it at Google's platform for inference once you are in the preview.

For everyone else, the bigger signal is that model access is becoming a commodity of the cloud layer. Leone put it simply to Axios: "Once every cloud has the top models, I think the competitive advantage goes to whoever does the best job with a company's data, governance, and costs." The models still compete with each other, but the place where a company buys them matters less and less.

A few cautions remain. The launch is a private preview, so availability, pricing and regional coverage are not public yet. Google has not published per token prices for the OpenAI models on its platform. And because Azure still gets new OpenAI products first, Google and AWS customers may wait for some features.

Our take

This deal completes a shift that started in April. OpenAI is no longer tied to one cloud, and the three hyperscalers are turning into neutral shops for frontier AI. For customers that is good news: more choice, fewer contracts and real pressure on price.

The interesting question is what happens to Gemini. Google is betting that it wins either way, because the customer stays on Google Cloud whatever model they pick. But putting GPT-6 Astra on the same shelf also makes it very easy for customers to compare. If Gemini falls behind on a task, switching will now take a single setting instead of a new vendor relationship.

For OpenAI, the move is less about one more partner and more about reach before a possible IPO. It now sells through every major cloud, just like Anthropic. The next fight will be about who offers the best agents, prices and governance on top of the same shelf, and that race starts with general availability in the coming weeks.

Sources: OpenAI, Google Cloud blog, CNBC, Reuters via Investing.com, Axios, Ars Technica, The Decoder, OpenAI on the Microsoft partnership, AWS.