Amazon and OpenAI Form Major Long‑Term AI Alliance * Science and Technology Times

Amazon and OpenAI have announced a multi‑year partnership focused on building large‑scale AI infrastructure and new enterprise‑grade tools.
The agreement includes joint development of a stateful AI runtime, expanded cloud distribution, and customized models for Amazon’s own products.
Its scope reflects one of the largest collaborations yet between a cloud provider and an AI research company.

A New Stateful Runtime Environment for Developers

Amazon Web Services and OpenAI will co‑create a Stateful Runtime Environment powered by OpenAI models and delivered through Amazon Bedrock. The system is designed to let developers maintain context across tools, data sources, and compute resources, enabling AI agents to work on ongoing tasks rather than isolated prompts. These environments will be optimized for AWS infrastructure and integrated with Bedrock AgentCore to ensure smooth interaction with existing enterprise applications. The companies expect the new runtime to launch within the next few months.

Stateful environments represent a shift from traditional stateless model interactions, which require users to repeatedly re‑establish context. By contrast, the new approach allows AI systems to remember prior work, manage identity, and operate across multiple software tools. AWS says this will help organizations build more capable agents that can handle complex workflows. Developers interested in early access are encouraged to contact AWS for deployment guidance.

AWS Becomes Exclusive Cloud Distributor for OpenAI Frontier

The partnership also makes AWS the exclusive third‑party cloud distribution provider for OpenAI Frontier, the company’s most advanced enterprise platform. Frontier enables organizations to build, deploy, and manage teams of AI agents with shared context, governance controls, and enterprise‑grade security. These capabilities are intended to support companies moving from experimentation to production‑scale AI deployments. AWS customers will be able to integrate Frontier into existing systems without managing underlying infrastructure.

OpenAI will rely heavily on AWS hardware to support demand for Frontier and the new runtime. The companies expanded their existing $38 billion agreement by an additional $100 billion over eight years. As part of this expansion, OpenAI committed to consuming roughly 2 gigawatts of Trainium compute capacity. This long‑term arrangement secures dedicated resources for OpenAI while helping AWS scale its custom silicon ecosystem.

Trainium3 and Trainium4 to Power Future AI Workloads

The agreement spans both current Trainium3 chips and next‑generation Trainium4 hardware. Trainium4, expected to begin delivery in 2027, will offer higher FP4 compute performance, expanded memory bandwidth, and increased high‑bandwidth memory capacity. These improvements are designed to support increasingly capable AI systems at global scale. OpenAI will use this infrastructure to power workloads related to the Stateful Runtime Environment, Frontier, and other advanced applications.

Amazon and OpenAI will also collaborate on customized models for Amazon’s own customer‑facing applications. Amazon developers will be able to tailor OpenAI models for use across AI products and agents that interact directly with users. These models will complement Amazon’s existing Nova family, giving teams additional options for building large‑scale AI features. Both companies describe the collaboration as a way to bring practical, widely accessible AI tools to consumers and enterprises.