OpenAI Lands on Amazon Bedrock the Day After Microsoft's 7-Year Exclusivity Ends
GPT-5.5, Codex, and Managed Agents are now in limited preview on Amazon Bedrock, giving AWS customers native access to OpenAI's frontier models for the first time. The move dismantles the Azure lock-in that defined the AI cloud market since 2019.
OpenAI’s models are on Amazon Web Services. On April 28 — exactly one day after Microsoft restructured its exclusivity arrangement to allow OpenAI to distribute products across all clouds — Amazon Bedrock went live with GPT-5.5, Codex, and Managed Agents in limited preview. Seven years of single-cloud lock-in ended in 24 hours.
What’s Available on Bedrock Now
Three distinct products launched simultaneously:
GPT-5.5 and GPT-5.4 on Bedrock. The latest OpenAI frontier models are accessible through Amazon Bedrock’s standard model APIs. AWS customers can fine-tune, evaluate, and orchestrate them through the same Bedrock toolchain they already use for Claude, Llama, and Amazon Titan — same SDK, same auth, same monitoring.
Codex on Bedrock. OpenAI’s coding agent runs on Bedrock infrastructure. AWS credentials handle authentication; Bedrock manages inference routing. For organizations already using Bedrock for AI workflows, adding Codex requires no new vendor relationship.
Amazon Bedrock Managed Agents. A new service combining OpenAI frontier models with AWS infrastructure for building production-ready AI agents. GeekWire noted that OpenAI gave AWS first access to Managed Agents — before Google Cloud — a signal of strategic prioritization that goes beyond a simple partnership announcement.
The Microsoft Context
Microsoft Azure remains OpenAI’s “primary” cloud partner under a licensing agreement that runs through 2032, and Azure will still receive priority access to new model releases. But the unilateral exclusivity that prevented any rival cloud from hosting OpenAI models is gone.
The original deal, struck in 2019 and extended with a reported $38B commitment, made Azure the default AI cloud for any enterprise wanting frontier OpenAI models. That structural advantage no longer exists. Enterprises that avoided OpenAI specifically because of Azure dependency now have a path through AWS or, eventually, Google Cloud.
Microsoft’s stock moved modestly lower on the news. The company doesn’t disclose Azure revenue separately, but Azure grew 40% year-over-year in Q1 2026 — a number that signals the OpenAI partnership wasn’t the only growth driver anyway.
What This Means for Developers
Teams running Amazon Bedrock as their primary AI gateway can now access OpenAI models without switching infrastructure. The more consequential shift: mixing OpenAI and Anthropic models within the same Bedrock orchestration layer is now possible natively. That combination — Claude for certain tasks, GPT-5.5 for others, unified routing through Bedrock — was previously impossible without multi-cloud plumbing.
AWS logged 28% revenue growth in Q1 2026 to $37.6B. The OpenAI partnership adds a forward-looking argument to that number: Bedrock is no longer positioned as the “not-Azure” option for teams that don’t want OpenAI’s models. It’s now a complete marketplace.
The limited preview is open to AWS customers today. General availability timing hasn’t been announced.
The Broader Shift
The exclusivity era shaped which cloud you used based on which AI lab you preferred. That’s over. OpenAI’s models will eventually be available on every major cloud, and the competition will shift to who builds the better orchestration, pricing, and compliance story around them.
For AWS, adding OpenAI solves the one criticism that Bedrock customers raised consistently: the absence of GPT models in an otherwise broad model catalog.
Related reading
- Cloud & Infrastructure AWS Launches Graviton5, Blackwell GPUs, and Bedrock AgentCore at New York Summit
- Cloud & Infrastructure AWS Ships Lambda S3 Files, Bedrock AgentCore CLI, and Aurora Serverless v4 in a Single Week
- Cloud & Infrastructure AWS Commits $200B in 2026, Deprecates 12+ Services, and Launches 'AI Factories' for On-Premises Deployments