OpenAI Comes to AWS Bedrock With GPT-5.5 and Codex

OpenAI is moving deeper into AWS, and the important part is not just another model listing. On April 28, 2026, OpenAI and Amazon Web Services announced that OpenAI models, Codex and Amazon Bedrock Managed Agents powered by OpenAI are coming to Amazon Bedrock in limited preview.

For enterprise AI buyers, this is a distribution story. OpenAI’s frontier models are no longer tied as tightly to a single commercial cloud path, while AWS gets to offer OpenAI capabilities inside the identity, billing, security and governance structures many large companies already use. The move does not erase Microsoft from the picture, but it does make OpenAI easier to buy and operate for AWS-heavy organizations.

What AWS and OpenAI Announced

The rollout has three parts. First, OpenAI models, including GPT-5.5, are being made available through Amazon Bedrock. Second, Codex can run with OpenAI models served through Bedrock, starting with the Codex CLI, desktop app and Visual Studio Code extension. Third, AWS is launching Bedrock Managed Agents powered by OpenAI for production-oriented agent workflows.

AWS describes all three as limited preview offerings. In its What’s New announcement, AWS said OpenAI models on Bedrock inherit controls such as IAM, AWS PrivateLink, guardrails, encryption and CloudTrail logging. That is the enterprise pitch in one sentence: use OpenAI capabilities without building a separate security and procurement model from scratch.

Why Bedrock Is the Distribution Layer

Amazon Bedrock has become AWS’s model access and orchestration layer for generative AI. Adding OpenAI gives customers another top-tier option alongside the existing Bedrock model marketplace, and it gives AWS a stronger answer to companies that want model choice without spreading governance across a patchwork of direct vendor accounts.

The AWS News Blog says the OpenAI preview includes GPT-5.5 and GPT-5.4 through Bedrock APIs, with unified security, governance and cost controls. That is especially relevant for enterprises that already use AWS commitments, centralized logging and private connectivity as part of their cloud operating model.

Codex Becomes an Enterprise Cloud Product

Codex is the more interesting part for software teams. OpenAI says more than 4 million people now use Codex weekly, across work such as writing code, explaining systems, refactoring applications, generating tests and modernizing legacy codebases. Bringing Codex to Bedrock turns it from a standalone coding agent into something enterprises can route through AWS credentials and controls.

That matters because coding agents touch sensitive material. They see proprietary code, build scripts, infrastructure configuration, tickets and internal documentation. OpenAI says customers can configure Codex to use Bedrock as the provider and that customer data is processed by Amazon Bedrock. For security teams, that still leaves work to do: define tool permissions, review generated changes, log agent actions and decide which repositories or environments are appropriate for agent access.

Managed Agents Point to the Next Platform Fight

The Managed Agents announcement shows where the cloud vendors are heading. AWS says Bedrock Managed Agents combines OpenAI frontier models with AWS infrastructure, giving each agent its own identity, logging actions and running inference through Amazon Bedrock. OpenAI describes the product as a way to deploy agents that maintain context, execute multi-step workflows, use tools and take action across business processes.

This is not just chatbot packaging. The enterprise market is moving toward agents that can carry out bounded tasks across systems: triage support cases, draft reports, update records, analyze code changes or coordinate routine operations. The winners will not be chosen only by benchmark scores. They will be chosen by how well the agent stack handles identity, auditability, permissions, reliability and recovery when something goes wrong.

The Microsoft Context

The timing is important. One day before the AWS announcement, OpenAI and Microsoft announced an amended partnership. Microsoft remains OpenAI’s primary cloud partner, and OpenAI products are set to ship first on Azure unless Microsoft cannot and chooses not to support the necessary capabilities. But Microsoft’s license to OpenAI models and products is now non-exclusive, and OpenAI can serve its products to customers across any cloud provider.

So this is not an Azure divorce. It is a widening of OpenAI’s enterprise routes to market. Microsoft keeps a privileged position and a long-running OpenAI relationship, while AWS gains a credible OpenAI story for customers whose production workloads, data controls and procurement already sit inside Amazon’s cloud.

What Enterprise Buyers Should Do Next

The first step is to treat the Bedrock launch as a preview, not a finished migration plan. Buyers should confirm availability, regions, pricing, model versions, latency, logging behavior, retention rules and whether usage can count toward existing AWS commitments. They should also compare the Bedrock route against direct OpenAI access and Azure-based OpenAI deployments rather than assuming one channel is automatically superior.

The second step is to evaluate workflows, not just models. GPT-5.5 may be the headline model, and OpenAI positions it as strong in agentic coding and computer-based work. But enterprise value will come from controlled deployment: good prompts, clean tool boundaries, retrieval design, human review, incident handling and measurable outcomes against real internal tasks.

The Bottom Line

OpenAI on AWS changes the buying surface for enterprise AI. It gives AWS customers a more direct path to GPT-5.5, Codex and OpenAI-powered agents inside Bedrock, while giving OpenAI broader distribution into the cloud estate where many production applications already live.

The careful read is that this is a strategic opening, not a finished market reset. Limited preview means details still matter. But the direction is clear: frontier AI is being packaged less as a single destination and more as a set of capabilities distributed through the clouds, controls and procurement systems enterprises already trust.