Until May 2026, any enterprise on AWS that wanted to use GPT-5.5 needed a separate OpenAI account, another vendor relationship, and a willingness to move data outside their AWS region. As of June 1, they don’t. OpenAI’s three flagship models — GPT-5.5, GPT-5.4, and Codex — are now available on Amazon Bedrock, the same storefront where AWS offers its preferred partner Claude. The enterprise AI power dynamic just realigned.
For Anthropic, the impact is immediate: it loses the privileged position built on its AWS integration. For enterprises, flexibility wins — they can choose without switching platforms. For AWS and OpenAI, two direct competitors that until now operated in separate spheres, it’s the start of an uncomfortable but strategically necessary relationship.
What’s on Bedrock
The announcement covers three products:
- GPT-5.5: OpenAI’s most advanced frontier model, with a 272K token context window, available through the Responses API on Bedrock. Optimized for complex reasoning, coding, and agentic workflows.
- GPT-5.4: The best price-performance option for workloads that don’t require the 5.5’s maximum capability.
- Codex: OpenAI’s coding agent, used by over 4 million developers weekly, now with its inference routed through Bedrock.
Pricing is direct — no seat licenses or per-developer commitments — and enterprises pay per token as they would directly with OpenAI. But the advantage for companies already on AWS is significant: they don’t need to manage another vendor relationship, all processing stays within the Bedrock region they choose (critical for data residency requirements), and integration with the AWS ecosystem is native.
Why it’s a seismic move
AWS and OpenAI have been uncomfortable competitors for years. AWS invested $4 billion in Anthropic and positioned Claude as its preferred frontier model on Bedrock. That OpenAI — the undisputed market leader — now puts its best models on the platform of its partner’s competitor signals a shift in industry power dynamics.
For AWS, the calculation is clear: Bedrock will only be the dominant enterprise platform if it offers all important models. Excluding OpenAI — or limiting itself to Anthropic — would have given enterprise customers a reason to seek multi-cloud alternatives. With OpenAI on Bedrock, AWS becomes the undisputed center of enterprise AI.
For OpenAI, the move is equally strategic. Despite having its own platform, the real volume of enterprise spending is on AWS. Companies spending tens of millions on AWS prefer not to add another vendor. By being on Bedrock, OpenAI accesses that spending without friction.
What this means for Anthropic
Anthropic is the most visibly affected player. Dario Amodei’s company has built much of its enterprise strategy on its preferred integration with AWS. That OpenAI now shares the same storefront — with a comparable (and in some areas superior) frontier model — erodes one of Anthropic’s competitive advantages.
That said, Anthropic has two defenses: first, Claude Fable 5, which on coding and reasoning benchmarks surpasses GPT-5.5 on several key metrics (80.3% vs 58.6% on SWE-bench Pro). Second, the depth of its AWS integration goes beyond simply being on Bedrock: Anthropic has worked with AWS on inference optimizations, security, and specific use cases that OpenAI is only beginning to build.
A new phase in competition
The OpenAI-AWS alliance marks the end of the phase where models and infrastructure were separate domains. We are now in a stage where:
- Hyperscalers compete to be model aggregators — not exclusive providers of just one
- Model creators compete on multiple fronts — model quality, price, and enterprise accessibility
- Enterprises win — more options, simpler integration, more competitive pricing
OpenAI on Bedrock is ultimately good news for the enterprise market. The open question is whether Anthropic can maintain its differentiation when it shares shelf space with its main competitor.