OpenAI’s new partnership with Amazon Web Services is a much bigger deal than just officially marking the end of its exclusive Microsoft pact.

As the world’s largest cloud platform, AWS is trusted by millions of organizations from startups to enterprises to government agencies. They often store their data there, and they already have their corporate security policies and governance in place across AWS systems.

That’s one of the biggest reasons Amazon Bedrock has rapidly become a preferred platform for piloting and launching AI projects. But at the What's Next with AWS event on Tuesday in San Francisco, AWS CEO Matt Garman tacitly admitted that Amazon Bedrock has been missing something: customers have been asking for OpenAI models.

That’s why both companies were eager to pound their chests about the new partnership. On Tuesday, the two teams revealed that they’ve been collaborating for eight weeks to launch a set of solutions for the enterprise.

"We are co-developing an agent platform from the ground up,” said OpenAI CEO Sam Altman. “This is day one. We're on a journey to give customers transformative agentic solutions."

For now, they announced three programs in limited preview:

• OpenAI models on Amazon Bedrock: One of the biggest draws of Bedrock is its wide selection of open-source and proprietary models from Anthropic, Meta, Nvidia, Mistral, DeepSeek, Cohere, and others, making it simple to deploy and test different models. They can access all of them through the Bedrock APIs and use the existing controls to manage security and costs. Adding OpenAI’s models fills the product’s biggest gap.
• Codex on Amazon Bedrock: After just eclipsing 4 million weekly users, Codex will now be available to an even broader set of business customers. Enterprise software teams can now use Codex to automate workflows from within their existing AWS environment, using their AWS credentials and infrastructure.
• Amazon Bedrock Managed Agents from OpenAI: Because agents can get so complex and dangerous, there’s a growing appeal to run them in a cloud like AWS, where identity management, auditing, persistent memory, and production-ready enterprise standards are already in place. That makes this the most turnkey agentic announcement from the two companies, and it matches Anthropic’s recent Claude Managed Agents offering.

Clearly, AWS customers don’t just get access to the latest OpenAI models in this deal, but also get access to its agent harness, which has quickly gained momentum as a rival to Claude Code. A harness is simply the software that makes LLMs agentic. “The harness is the playbook for the model,” said Anthony Liguori, AWS distinguished engineer, at the event.

Our Deeper View

It’s only been in the last couple of months that OpenAI has made a hard pivot toward the enterprise, but it continues to make deliberate moves that show a much stronger commitment to becoming a business partner. That's a welcome development after OpenAI drew broad criticism for focusing on too many bets in late 2025. And as rival Anthropic continues its mission to become the preferred model provider for enterprises—and positions itself for a soaring valuation because of it—any strategy that can give OpenAI a leg up among businesses is a plus. And enterprises will always welcome the competition and additional options.

Disclosure: Jason Hiner's travel to What's Next With AWS was paid by Amazon. The Deep View's coverage is editorially independent from the companies we cover.