OpenAI brings GPT-5.5 and Codex to AWS Bedrock in general availability

OpenAI has announced that its flagship (frontier) models, including GPT-5.5, along with the Codex development tool, are now generally available on Amazon Bedrock. Enterprise AWS customers can now build solutions on OpenAI models using the same security, compliance and infrastructure-management mechanisms they already rely on across Amazon Web Services.

What changes for AWS customers

Amazon Bedrock is a managed AWS service that gives companies access to models from several developers through a single API, without having to deploy infrastructure themselves. Before OpenAI's models arrived on the platform, many enterprise AWS customers could already work with models from other providers, but to use OpenAI's models they had to step outside their own cloud perimeter, for example to Microsoft Azure, where OpenAI has historically had its closest partnership. GPT-5.5 and Codex in Bedrock remove that dependency: enterprise customers can now keep the entire stack, from data storage to model calls, within a single AWS infrastructure, with the same IAM, auditing and network isolation settings as for their other services.

The practical effect will be felt first by teams that have spent recent years running a multi-cloud architecture solely to reach a specific model, for example keeping part of their infrastructure in Azure only for GPT calls and the rest in AWS. That split is no longer necessary: the whole stack can be consolidated in one cloud without losing access to OpenAI's leading models.

Before OpenAI's models arrived, Amazon Bedrock already offered access to models from other major developers, including Anthropic's Claude family and Amazon's own models. Adding OpenAI effectively completes a picture in which nearly all the world's leading labs are available to AWS enterprise customers from a single entry point. For companies that have already chosen AWS as their main cloud platform, this removes the need to maintain separate contracts, billing and security procedures for each new model provider.

• Models — OpenAI's flagship (frontier) models, including GPT-5.5.
• Tool — Codex, an assistant for development and code writing.
• Platform — Amazon Bedrock.
• Status — general availability, not a limited preview.
• Value for customers — use of AWS's existing security, compliance and governance mechanisms.

What's next for OpenAI on cloud platforms

In its announcement, OpenAI explicitly calls the launch on AWS Bedrock only "the beginning of a broader expansion", that is, part of a strategy to be present on all major cloud platforms rather than tied exclusively to one cloud partner. This is a significant signal for the market: competition for access to frontier models increasingly takes place not so much between the model developers themselves as between cloud providers, each wanting to offer enterprise customers the widest possible choice of models within a single managed platform.

For AWS, the partnership strengthens Bedrock's positioning as a neutral access point to models from different developers, a competitive advantage over clouds that bet on a single model ecosystem. For customers it means fewer architectural compromises: they can test and combine models from different providers within one security perimeter, without rebuilding their infrastructure for each new vendor. The step also strengthens Codex's position as an enterprise development tool: integration with AWS governance tools makes it easier to adopt at companies with strict requirements for code auditing and access control, above all in regulated sectors such as finance, healthcare and government, where the move to cloud AI tools has long been held back precisely by compliance questions.

OpenAI and Microsoft have historically shared one of the industry's most prominent partnerships, involving large investments and tight integration of the models with the Azure cloud. The launch on AWS Bedrock does not cancel that partnership, but it clearly shows that OpenAI no longer treats Azure as the only distribution channel for enterprise customers. The company is going where enterprise cloud infrastructure budgets are already concentrated, regardless of historical agreements with any single cloud provider.