The “exclusive” era of the Microsoft-OpenAI alliance has officially come to an end as OpenAI’s models are now available via AWS.
Just yesterday, Microsoft and OpenAI amended their long-standing partnership that allowed OpenAI to serve its models and products outside of the Azure cloud. Today, AWS announced that it is bringing OpenAI’s frontier models and agentic AI capabilities to Amazon Bedrock in limited preview.
The timing of the announcement indicates that OpenAI has been working with AWS in the background for quite some time, even before Microsoft officially signed off on the amended partnership.
With this new announcement, millions of enterprises will be able to use OpenAI models and agents within AWS environments by following existing security, governance, and operational controls. Today, AWS announced the following three new offerings:
- OpenAI models on Amazon Bedrock
- Codex on Amazon Bedrock
- Amazon Bedrock Managed Agents, powered by OpenAI
With OpenAI models on Amazon Bedrock, AWS customers will be able to access OpenAI frontier models, including the latest GPT-5.5, through the same Bedrock APIs and controls they already use for other model providers. Like other AI models, OpenAI models on Bedrock will support enterprise controls such as IAM-based access management, AWS PrivateLink connectivity, guardrails, encryption at rest and in transit, AWS CloudTrail logging, and integration with existing compliance frameworks. Customers can also apply OpenAI model usage toward their existing AWS cloud commitments.
AWS is also bringing OpenAI’s Codex coding agent to Amazon Bedrock, allowing enterprise customers to use Codex inside AWS environments, authenticate with AWS credentials, and process inference through Bedrock infrastructure. AWS says Codex on Bedrock will be available through the Bedrock API, starting with the Codex CLI, Codex desktop app, and Visual Studio Code extension.
Amazon Bedrock Managed Agents, powered by OpenAI, will help enterprises build and deploy production-ready AI agents. These are built with the OpenAI agent harness and include features such as persistent memory, skills for encoding procedures, identity and permissions, logging, and compute options. AWS highlighted that each agent runs with its own identity, logs actions for auditability, and keeps model inference on Amazon Bedrock.
AWS concluded the official announcement blog post by describing this launch as the beginning of a deeper collaboration with OpenAI, as it will continue to bring OpenAI’s latest reasoning and agentic AI advances to Amazon Bedrock as they become available.
