OpenAI shifts GPT‑5.5 distribution

Cloud distribution sounds boring. But this is really a control story. OpenAI did not just launch another model last week — it changed who gets to sell and host that model for big companies. On April 27, Microsoft said its OpenAI license is now non-exclusive, and on April 28 AWS immediately put OpenAI models, Codex, and Managed Agents into Bedrock. That is the part that matters. ### What actually moved? Two things moved in 48 hours. First, OpenAI launched GPT‑5.5 on April 23, then added it to the API on April 24. Second, the long-standing Microsoft arrangement changed on April 27, and AWS launched OpenAI offerings on Bedrock the next day. So the story is not “GPT‑5.5 exists.” The story is “GPT‑5.5 stopped being effectively tied to one cloud path.” ### Why is AWS Bedrock the big deal? Because enterprises do not buy frontier models the way individual developers do. They buy through existing contracts, identity systems, logging rules, networking controls, and compliance workflows. AWS says OpenAI models on Bedrock use IAM, PrivateLink, encryption, guardrails, and CloudTrail logging, and that usage can count toward existing AWS commitments. Basically, organizing around Azure. ### Didn’t Microsoft keep something? Yes — but less than before. Microsoft says it remains OpenAI’s primary cloud partner, and OpenAI products will ship first on Azure unless Microsoft cannot or chooses not to support the needed capabilities. Microsoft also keeps a license to OpenAI IP through 2032. But the key sentence is the one that changed the market to non-exclusive. ### So is Azure exclusivity over? For practical enterprise buying, yes. The older setup still showed up publicly as recently as February 27, when Microsoft said Azure remained the exclusive cloud provider for stateless OpenAI APIs. The April 27 amendment replaced that with a non-exclusive structure. That is why AWS could move so fast on April 28 without dancing around the old terms. ### Why does GPT‑5.5 matter in this fight? Because OpenAI did not loosen distribution around a minor release. GPT‑5.5 is the company’s newest frontier model for coding, computer use, knowledge work, and research workflows. OpenAI says it matches GPT‑5.4 latency while improving benchmark performance, including 82.7% on Terminal‑Bench 2.0 versus 75.1% for GPT‑5.4. In other words — the model is strong enough that cloud access really changes vendor leverage. ### What about Codex? Codex makes the shift more concrete. AWS did not just add a raw model endpoint. It added Codex through Bedrock, with access through the Codex CLI, desktop app, and VS Code extension. OpenAI has also been pushing Codex hard into enterprises, saying it reached 4 million weekly active users and expanded deployment help through major systems integrators. That turns this from a model-hosting story into a workflow story. ### Who gains the most? Enterprise buyers gain first. AWS gains a lot too, because Bedrock becomes a place to buy OpenAI without leaving AWS governance and spending frameworks. OpenAI gains leverage because it is no longer commercially boxed into one hyperscaler. Microsoft still benefits from first-ship status on Azure and its continuing IP license, but the clean exclusivity advantage is gone. That is the real reset. ### Bottom line? This is not just a model launch. It is OpenAI turning distribution into a competitive weapon. GPT‑5.5 may be the headline object, but the deeper change is that OpenAI can now reach enterprise customers through more than one cloud gatekeeper — and that weakens Microsoft’s old chokehold without ending the partnership.