Skip to content
NewsProduct launch

OpenAI agents now run natively on AWS through new Bedrock Managed Agents

· by Pondero Newsdesk

The short version

OpenAI and AWS launched Bedrock Managed Agents in preview on September 29, 2026, letting OpenAI's agent stack run entirely inside AWS instead of Microsoft Azure or OpenAI's own servers.

OpenAI agents now run natively on AWS through new Bedrock Managed Agents

For seven years, running an OpenAI model in production has meant routing through Microsoft's cloud. That changed on September 29, 2026, when OpenAI and Amazon shipped Bedrock Managed Agents in preview, a service that runs OpenAI's own agent stack entirely inside AWS, under AWS identities and permissions rather than OpenAI's or Microsoft's.

What shipped

OpenAI unveiled Bedrock Managed Agents during its DevDay 2026 keynote alongside more than 20 other announcements, per OpenAI's DevDay 2026 recap. The service combines OpenAI's models with the Codex harness and Amazon Bedrock AgentCore, according to AWS's product page. It carries over the customization of the Agents API OpenAI introduced earlier in the same keynote, but the agent runtime and model inference now stay inside AWS rather than OpenAI's own infrastructure or Azure, per OpenAI. AWS says each agent gets its own AWS identity and defined permissions, so teams can control exactly which resources an agent can touch and which actions it can take. The launch is a preview; AWS has not published general-availability pricing, a GA date, or supported regions.

What AWS-committed buyers get that Azure did not offer

OpenAI's infrastructure has run almost exclusively on Microsoft Azure since Microsoft's initial 2019 investment, and Azure has remained the default enterprise on-ramp for OpenAI's models. Bedrock Managed Agents gives AWS-committed enterprises, the ones that already run identity, security, and compliance tooling on AWS, a way to deploy OpenAI agents without also standing up Azure infrastructure alongside it. That is the concrete choice this creates: a team evaluating where to run an OpenAI-powered agent no longer has to weigh "build on Azure" against "build the agent orchestration ourselves on AWS." It can now ask AWS to run the whole stack.

Salesforce is an early named partner. Per AWS's page, Salesforce President of Enterprise & AI Technology Joe Inzerillo said the company's Headless 360 product will pair OpenAI's models and AWS infrastructure with Salesforce's own data, workflows, and controls, adding that in that setup "data never leaves AWS." That is Salesforce and AWS characterizing the integration, not an independently audited claim, and neither company has published a customer using the integration in production yet.

What to watch next

Watch whether Microsoft comments publicly on OpenAI's AWS expansion, given the exclusivity that has defined the two companies' cloud relationship since 2019. Also watch whether Google Cloud or another hyperscaler ships a comparable first-party OpenAI-agent integration, and whether AWS names a GA date or pricing for Bedrock Managed Agents once the preview period ends.

Sources