OpenAI jumps out of Microsoft's bed, into Amazon's Bedrock
The Register ● Covered by 3 sources
OpenAI's models are now live on Amazon's Bedrock platform, not just Azure. Microsoft loosened its exclusivity grip, so OpenAI can now cut deals with any cloud it wants.
Based on reporting by The Register — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI has officially set up shop on Amazon Web Services, with its GPT-5.4 model now running through Bedrock, AWS's managed inference and agent-building platform. The move, unveiled Tuesday at an AWS event in San Francisco, gives enterprises a way to tap OpenAI's models without routing sensitive data through OpenAI's own APIs, addressing the security and sovereignty worries that Amazon says have been holding customers back.
The arrangement goes beyond simple inference access. OpenAI's models will also plug into Bedrock's Managed Agents and AgentCore tools, letting businesses build agents tied directly to their own data and systems. AWS used the same event to roll out a broader agentic push, including Quick, a cross-vendor assistant that plays in the same space as Microsoft Copilot, and expanded versions of Connect, Amazon's CRM platform, now stretching into HR, healthcare, and supply chain automation. Developers also get a notable perk: OpenAI's Codex coding agent can now run against models hosted in AWS datacenters, which should ease fears about proprietary code ending up as training fodder for some future GPT release.
Right now, everything is limited preview. GPT-5.4 is available today, and AWS CEO Matt Garman said GPT-5.5 should land within a couple of weeks. This is OpenAI following through on a commitment made back in February, when it agreed to bring its models to AWS in return for as much as $35 billion in financing, contingent on OpenAI eventually deploying two gigawatts worth of Amazon's Trainium chips.
What makes this possible, though, is a quieter shift in OpenAI's relationship with Microsoft. Microsoft agreed to loosen its exclusivity claim over OpenAI's models in exchange for walking away from its revenue-sharing arrangement. Microsoft still holds the title of primary cloud provider and keeps its technical access, but OpenAI is now free to strike deals wherever it likes. Amazon just happened to be first through the door, and given the size of the numbers involved, probably won't be the last.
My take — AI-written commentary, not fact-checked reporting
Everyone's framing this as OpenAI ditching Microsoft, but it's really Microsoft cashing out of an exclusivity deal that was starting to look like a liability once OpenAI's compute needs outgrew any single cloud. The real story is that OpenAI just became infrastructure-agnostic at exactly the moment every hyperscaler is desperate to say they have frontier models on their platform. Expect Google Cloud next, and expect this to become the new normal rather than a one-time favor to Amazon.
Read more about this at: The Register