Amazon brings new OpenAI products to AWS after OpenAI reworks its exclusive Microsoft deal

Amazon moved almost immediately to bring new OpenAI products to AWS. Amazon Bedrock now offers the latest OpenAI models, Codex and Bedrock Managed Agents, a service for running agentic systems. The launch came one day after OpenAI renegotiated its agreement with Microsoft and dropped the cloud exclusivity that had long kept its products inside Azure.

What arrived on AWS

AWS opened a limited preview covering three areas at once. Amazon Bedrock is getting OpenAI's latest models, including GPT-5.4, and the company promises to add GPT-5.5 soon. Codex is coming too: OpenAI's coding agent can run through a CLI, a desktop app and a Visual Studio Code extension. The third piece is Amazon Bedrock Managed Agents, a new service for building and running OpenAI agents directly on AWS infrastructure.

• Access to OpenAI models through the familiar Amazon Bedrock APIs
• Codex inside the AWS environment, authenticated with AWS credentials
• Managed Agents for production-ready agentic workloads
• Unified enterprise controls: IAM, encryption, PrivateLink and CloudTrail

For enterprise teams, this detail matters: they don't need to stand up separate infrastructure for OpenAI or rebuild a security layer from scratch. Amazon is betting that companies will want models, a coding agent and an agent runtime in a single perimeter where access, logging, governance and budget limits are already configured. On top of that, spending on OpenAI services can count toward existing cloud commitments with AWS. That makes adoption much easier for procurement and compliance.

Why it happened now

Timing is everything here. On April 27, 2026, Microsoft and OpenAI announced a revised partnership: Microsoft remains the primary cloud partner, but its license becomes non-exclusive, and OpenAI gains the right to serve all its products through any cloud provider. By April 28, AWS had seized the opening and brought OpenAI to Bedrock with almost no delay. For the market, that is practical confirmation that the new arrangement took effect immediately.

"We have a lot of developers and companies who want to run services on

OpenAI models in AWS," — Andy Jassy.

The groundwork was laid in advance. Back in February, Amazon and OpenAI announced a strategic partnership that included up to $50 billion in investment from Amazon, joint development of a Stateful Runtime Environment, and plans to build out an agent platform on Bedrock. At the time, Microsoft's existing exclusivity remained the legal bottleneck. Once it was relaxed, the February agreements began turning into concrete products rather than just press-release promises.

What businesses get

The most practical part of this announcement isn't the new product names but the delivery model. OpenAI arrives on AWS in a form that suits large companies: access management, logging of every agent action, execution inside the customer's environment, and no need to train teams on a new security model. This matters most for Codex, because the agent works right next to real codebases, IDEs and internal development services. Security and platform teams will have fewer new perimeters to sign off on.

Managed Agents are also aimed at long-running workloads rather than demos. AWS specifically highlights that each agent gets its own identity, logs its actions and uses the OpenAI agent harness for more precise control over long tasks. It's a sign that cloud competition is shifting from simple model hosting to a more lucrative layer: who can best package agents, memory, orchestration and enterprise security. And that layer usually decides whether a pilot makes it to production.

What it means

For the market, this is another sign that the era of "one model provider, one cloud channel" is ending. OpenAI is widening its distribution, AWS is strengthening Bedrock as the enterprise entry point for agentic AI, and Microsoft is losing part of its exclusive advantage. Businesses get more choice: they can use OpenAI's strong models without leaving their AWS perimeter.