AI Rival Becomes Customer: Google Cloud Starts Selling Three OpenAI GPT Models, Secures Exclusive Channel for All Three Closed-Source Models
Google Cloud and OpenAI announced on October 21 that OpenAI's newest AI models will be available on Google Cloud's Model Garden platform. This will let business customers choose GPT models directly from the cloud services they already use and pay for them with money they have already committed to spending on Google Cloud. The deal brings the two companies closer together in selling AI products, even though they compete directly in the AI market. It also makes Google Cloud the only large cloud provider that currently offers all three major groups of closed-source AI models: Google's Gemini, Anthropic's Claude, and OpenAI's GPT.
The three models being added are GPT-6 Astra, GPT-6.1 Sol, and GPT-6.0 Luna. Each is designed for different needs. Astra is built for difficult tasks such as coding, operating computers, professional work, and scientific research. Sol aims to provide reasoning skills close to Astra's but at a lower cost, making it suitable for AI agents that write code and other professional uses. Luna focuses on fast responses and low costs, making it suitable for everyday tasks at large scale. Google Cloud said the three models are currently available in a limited preview for selected business customers. OpenAI expects them to become generally available within the next few weeks.
The biggest immediate benefit for businesses is simpler purchasing and billing. They can buy OpenAI models through Google Cloud Marketplace, see the costs on one combined cloud bill, and use funds they have already committed to spending on Google Cloud to pay for them. These purchases also count toward flexible savings programs that cover models from different providers. This means businesses no longer need separate contracts with each model developer. They can also switch between models based on the needs of each task, including performance, response time, and cost.
Google Cloud said that when businesses use OpenAI's models or other third-party models, their prompts, background information, and generated results stay within set data boundaries and are not used to train foundation models. A zero-data-retention option, which would mean keeping no user data after processing, is planned for the future. OpenAI's models run on dedicated Google Cloud infrastructure. Requests must pass through a specially secured gateway that helps prevent third-party model providers from accessing business cloud environments without permission. Support for confidential computing is also planned. Every interaction between an AI agent and a model uses a verified identity. IT administrators can monitor activity from start to finish through VPC Service Controls and shared audit logs.
OpenAI co-founder and President Greg Brockman said Google Cloud is a leading platform where developers and businesses build AI agents, create software, and develop AI-powered business applications. He said the partnership is an important step toward letting customers use OpenAI's most capable models and Codex directly in the development environments they already use. Google Cloud CEO Thomas Kurian said the company wants to give customers a wide range of choices, including OpenAI models, so they can pick the model that best fits their business while balancing quality and cost.
The deal comes after OpenAI's exclusive distribution agreement with Microsoft ended in April 2026. Since then, OpenAI has worked to make its models available through more cloud providers. Amazon Web Services later began offering OpenAI models, and Google Cloud's addition means OpenAI now has sales channels through all three major U.S. cloud infrastructure providers. OpenAI also stressed in its announcement that Microsoft remains its main cloud partner.
The relationship goes beyond one company selling another company's products. OpenAI already uses Google Cloud's infrastructure to run ChatGPT and its API services. It will now also use Google Cloud as another way to sell its models. The two companies compete strongly in AI model development and search, but they also depend on each other because developing and running large AI models costs so much. For Google Cloud, offering a major competitor's leading models strengthens its position as a platform that supports different products. It may also attract businesses that do not want to depend on just one technology ecosystem.
Across the industry, cloud platforms are quickly becoming the main way to sell advanced AI models. Cloud providers usually earn a share of the money from sales made through their platforms. For AI developers, these platforms offer an established way to reach business customers and let those customers pay from budgets they already use for cloud services. Anthropic, a major rival of OpenAI, provides an example of this dependence. Its filings show that 47% of its revenue in 2025 came through Amazon's and Google's cloud platforms.
Google Cloud's own rapid growth is another reason for this move. Market information shows that, as demand for AI infrastructure increased, Google Cloud's annualized revenue growth rate reached about 82% in the second quarter, up from 63% in the first quarter. Over the same period, Amazon's and Microsoft's cloud businesses grew by 37% and 42%, respectively. Adding a competitor's models while growth is speeding up could increase the amount of business done through Google Cloud. It could also make customers less likely to move to another cloud provider just because they prefer a particular AI model.
The agreement also covers tools that developers use to write software. OpenAI's coding agent Codex will connect with Google Cloud, allowing developers to send computing work for AI model responses to Google Cloud while using Codex CLI on their own computers or the Codex desktop app. Google's AI development environment, Antigravity, previously supported only Gemini and Claude. It is also expected to add OpenAI models in the future. Support for Gemini-related enterprise services is planned as well. However, Google has not said which models will be available through Antigravity.
For investors, the news shows that cloud companies may be changing how they compete. The market once tended to treat Gemini and GPT as rivals in a winner-takes-all contest. Google has now chosen to offer a competitor's product on its own platform. This suggests that its cloud strategy is moving away from keeping customers tied to its own models and toward offering a wider choice of models to keep customers using Google Cloud. Whether this approach leads to more cloud usage and a larger share of sales will be important to watch as Google Cloud continues to grow. After OpenAI makes its models available across all three major cloud providers, changes in the share of its business revenue will also help show how mature its business model has become.