OpenAI Models, Codex, and Managed Agents Launch on Amazon Bedrock in Limited Preview

OpenAI and Amazon Web Services announced an expanded partnership on April 28 that brings OpenAI’s frontier models, Codex coding agent, and a new co-built Managed Agents service to Amazon Bedrock. The launch came exactly one day after OpenAI revised its partnership with Microsoft to end its exclusive cloud arrangement, clearing the path for multi-cloud distribution.

Three Products, One Launch

The partnership delivers three capabilities, all in limited preview with general availability expected in coming weeks, according to CNBC.

First, OpenAI’s latest frontier models including GPT-5.5 are now accessible through Amazon Bedrock. AWS customers can build with OpenAI models using their existing IAM, PrivateLink, guardrails, encryption, and CloudTrail logging. Usage applies toward existing AWS cloud commitments.

Second, Codex on Bedrock brings OpenAI’s coding agent into AWS environments. More than 4 million people now use Codex weekly, according to OpenAI, across code generation, system explanation, legacy codebase modernization, and professional workflows extending beyond coding. The integration supports Codex CLI, the desktop app, and the VS Code extension.

Third, Amazon Bedrock Managed Agents, powered by OpenAI, is a jointly built service for deploying production-ready agents. Each agent gets its own identity, logs every action, and runs within the customer’s environment with all inference on Bedrock. According to AWS, the service handles orchestration, tool use, and governance through integration with Amazon’s security and compliance controls, with Bedrock AgentCore providing the default compute environment.

The Microsoft Factor

“This is what our customers have been asking us for for a really long time,” AWS CEO Matt Garman said at the San Francisco launch event, according to CNBC.

OpenAI CEO Sam Altman appeared via recorded message, as he is currently in court in Oakland for his case against Elon Musk. The timing underscored how quickly OpenAI moved after the Microsoft exclusivity revision: the partnership terms changed on Monday, and the AWS products launched on Tuesday.

OpenAI’s revenue chief Denise Dresser had previewed the strategic logic in a memo to employees earlier in April, writing that the Microsoft relationship “has also limited our ability to meet enterprises where they are, for many that’s Bedrock.”

Building Blocks Already in Place

The AWS launch did not come from nowhere. In November 2025, OpenAI announced a $38 billion commitment with AWS, according to CNBC. Three months later, Amazon invested $50 billion in OpenAI, with OpenAI committing to use two gigawatts of AWS’s custom Trainium chips for model training. Until now, only open-weight models from OpenAI were available on AWS.

The Agent Distribution Race

The Managed Agents product is the most strategically significant of the three. Anthropic launched its own Managed Agents on Claude for enterprise customers in early April. Google has Vertex AI Agent Builder. Now OpenAI’s agents run natively on AWS infrastructure with inherited governance controls.

For enterprises already running workloads on AWS, the friction to deploy OpenAI-powered agents just dropped to near zero: same IAM, same billing, same compliance framework. The question for the next quarter is whether Managed Agents on Bedrock can close the gap between prototype and production faster than Anthropic’s offering on Google Cloud or standalone agent platforms that require separate governance layers.