GPT-5.5‑Cyber and Codex variants now available on AWS Bedrock — OpenAI expands cloud rollout

Cloud distribution is the story here — not just another model launch. OpenAI put GPT‑5.5, Codex, and Bedrock Managed Agents onto AWS on April 28, 2026, which means enterprises can now buy and run flagship OpenAI tools inside Amazon’s stack instead of going through Azure alone. That sounds like plumbing. But in enterprise AI, plumbing is the product. The big shift is that OpenAI’s cloud route just widened in a very public way. (openai.com) ### What actually launched on AWS? Three things launched together in limited preview: OpenAI models on Amazon Bedrock, Codex on AWS, and Amazon Bedrock Managed Agents powered by OpenAI. OpenAI explicitly named GPT‑5.5 as part of the Bedrock rollout, and AWS folded the launch into its “What’s Next with AWS” event the same day. This is not just API access in the abstract — it is O(openai.com)s managed enterprise AI platform. (openai.com) ### Why does Bedrock matter so much? Because Bedrock is where a lot of big companies already do the boring but essential stuff — identity, billing, compliance, security controls, procurement, and production deployment. OpenAI’s own pitch is basically: if your company already lives in AWS, you can now use OpenAI without ripping up the rest of your stack. AWS is making the same ar(openai.com)marketplace with governance and guardrails built in. That lowers the friction from “interesting demo” to “approved enterprise purchase order.” (openai.com) ### Why is Codex part of the headline? Because coding agents are becoming one of the clearest enterprise AI budgets right now. OpenAI says more than 4 million people use Codex every week, and the AWS setup lets companies point Codex at Bedrock-served OpenAI models while keeping AWS billing, availability, and security in the loop. There’s also a very practical procurement angle —(openai.com)ng AWS cloud commitments. Basically, AWS is turning OpenAI from a separate vendor conversation into something that can fit inside an existing infrastructure contract. (openai.com) ### So did Microsoft really lose exclusivity? Yes — in the way that matters for distribution. On April 27, 2026, Microsoft and OpenAI amended their partnership so Microsoft’s license became non-exclusive, while Microsoft remained OpenAI’s “primary cloud partner.” OpenAI can now serve its products across any cloud provider. Microsoft still keeps a license to OpenAI IP through 203(openai.com)tinue through 2030, subject to a cap. So this is not a breakup. It is a loosening. (openai.com) ### What does “primary cloud partner” mean now? It means Azure still gets privileged positioning, but not total channel control. Microsoft says OpenAI products will ship first on Azure unless Microsoft cannot or chooses not to support the needed capabilities. That is a very different arrangement from straight exclusivity. Think less “sole distributor” and m(openai.com)Codex one day after the revised deal tells you how real that change is. (blogs.microsoft.com) ### Why should enterprises care? Because this shifts leverage. A company that wants OpenAI models can now compare clouds on price, governance, latency, existing commitments, and where its data already lives. If you are deep in AWS, Bedrock is now a cleaner path. If you are standardized on Azure, Azure still has the(blogs.microsoft.com)ure competition again — which is exactly what hyperscalers want. (openai.com) ### What’s the catch? The launch is still in limited preview, and “available on a cloud” does not mean every feature, region, or enterprise control is instantly identical everywhere. Microsoft also still has meaningful structural advantages — primary-partner status, long-dated IP rights, and ongoing economics. So the exclusivity era is ending, but the Microsoft relationship is still deeply embedded in OpenAI’s business. (openai.com) ### Bottom line? This is OpenAI turning cloud access into a multi-platform business. AWS did not just add another model. It broke the old assumption that using top-tier OpenAI systems meant defaulting to Azure. (openai.com)