Amazon and OpenAI have entered a multi-year strategic partnership that pairs OpenAI’s frontier models and enterprise agent platform with Amazon Web Services (AWS) infrastructure, custom silicon, and global distribution. As part of the deal, Amazon will invest $50 billion in OpenAI—an initial $15 billion followed by $35 billion subject to conditions—while OpenAI expands its existing AWS compute agreement by $100 billion over eight years.
Under the agreement, OpenAI and AWS will jointly develop a Stateful Runtime Environment powered by OpenAI models and delivered via Amazon Bedrock. The stateful architecture allows AI systems to retain context across sessions, access memory and identity, and orchestrate workflows across software tools and data sources. AWS will integrate the runtime with Bedrock AgentCore and its broader infrastructure stack to support production-scale generative AI applications and autonomous agents. The companies expect to launch the Stateful Runtime Environment in the coming months.
AWS will also become the exclusive third-party cloud distribution provider for OpenAI Frontier, OpenAI’s enterprise platform for building and managing teams of AI agents. Frontier supports shared context, governance, and enterprise-grade security without requiring customers to manage underlying infrastructure. To support growing demand, OpenAI has committed to consume approximately 2 gigawatts of AWS Trainium capacity, spanning Trainium3 and next-generation Trainium4 chips. Trainium4, slated for delivery beginning in 2027, will add higher FP4 compute performance, expanded memory bandwidth, and increased high-bandwidth memory capacity for large-scale AI workloads. The companies will also co-develop customized OpenAI models tailored for Amazon’s customer-facing applications, complementing Amazon’s Nova model family.
- Amazon will invest $50 billion in OpenAI ($15B initial, $35B contingent).
- OpenAI expands its AWS compute agreement by $100 billion over 8 years.
- OpenAI commits to consume ~2 gigawatts of AWS Trainium capacity.
- AWS becomes exclusive third-party cloud distribution provider for OpenAI Frontier.
- Joint development of a Stateful Runtime Environment delivered via Amazon Bedrock.
- Customized OpenAI models to power Amazon customer-facing applications.
- Trainium4, expected in 2027, targets higher FP4 performance and memory bandwidth.
“OpenAI and Amazon share a belief that AI should show up in ways that are practical and genuinely useful for people,” said Sam Altman, co-founder and CEO of OpenAI. “Combining OpenAI’s models with Amazon’s infrastructure and global reach helps us put powerful AI into the hands of businesses and users at real scale.”
🌐 Analysis: This agreement signals a deeper vertical alignment between model developers and hyperscale cloud providers, as competition intensifies among cloud platforms to secure differentiated AI workloads. By committing to Trainium capacity at multi-gigawatt scale, OpenAI strengthens AWS’s position against competing AI infrastructure ecosystems built around NVIDIA GPUs and custom silicon from other hyperscalers, while Amazon gains a flagship frontier-model partner for Bedrock and enterprise agent orchestration.