OpenAI brings GPT-5.5 and Codex to AWS Bedrock

For enterprise IT teams, the hardest part of adopting AI has never been picking a model. The real headache is getting it through compliance, procurement, and security when your entire company already lives inside AWS.

OpenAI and AWS just expanded their strategic partnership to tackle that exact problem, launching three major capabilities in limited preview on Amazon Bedrock.

Frontier models and Codex inside your existing stack

The marquee update brings OpenAI's frontier models, including GPT-5.5, directly onto Amazon Bedrock. That means enterprises can build using OpenAI within the security controls, identity systems, and procurement pipelines they already have in place on AWS.

Codex is also coming along for the ride. OpenAI says more than 4 million people now use Codex weekly, not just for writing code and modernizing legacy codebases, but increasingly for parsing documents and generating briefs.

Developers can now configure Codex to run directly on Amazon Bedrock. This works across the Codex CLI, the desktop app, and the VS Code extension. The killer feature for enterprise procurement is that eligible customers can count their Codex usage against their existing AWS cloud spending commitments, while keeping all data processing inside Bedrock.

Agents that skip the infrastructure headache

The third piece of the rollout is Amazon Bedrock Managed Agents powered by OpenAI.

Building reliable agents usually means spending weeks wrangling context management, tool calling, multi-step orchestration, and enterprise governance. Managed Agents handles that heavy lifting out of the box.

By taking care of the deployment plumbing and tying directly into native AWS security policies, it aims to give companies a much straighter shot from an internal hackathon demo to an agent running in actual production.

Why it matters

For enterprise developers and engineering leaders locked into AWS, this is a massive win. You no longer have to beg procurement for new vendor approvals or navigate complex multi-cloud setups just to get your hands on GPT-5.5 or the official Codex tools. Being able to burn down AWS spend commitments makes greenlighting these projects significantly easier.

For the broader cloud race, this marks a major shift. Microsoft Azure's long-standing positioning as the sole enterprise home for cutting-edge OpenAI tech is continuing to erode. AWS is aggressively turning Bedrock into the ultimate multi-model hub, putting immense pressure on rivals like Google Cloud Vertex AI.

We think this is a brilliant distribution play by OpenAI. Chasing enterprise budget means meeting customers where their data and billing already live, rather than forcing them into a different cloud ecosystem. The biggest question mark now is performance and rollout speed. With everything currently in limited preview, teams will need to see whether Bedrock can deliver the latency, quota reliability, and fast feature updates that mission-critical production workloads demand.