OpenAI plans to use 2GW of Trainium capacity through Amazon Web Services infrastructure and expand its contract with the cloud provider.

Trainium is Amazon's custom-designed AI training accelerator, which has historically lagged rival Google's TPU family as well as market leader Nvidia's GPUs. The Trainium family are also used by Anthropic, following an Amazon investment.

This deal is also tied to Amazon's investment, with the company set to pump $50 billion into OpenAI – $15bn up front and $35bn in the coming months after certain conditions being met.

In November, OpenAI announced it would spend $38bn on AWS cloud services, at the time highlighting that it was to access Nvidia GPUs. Now, OpenAI plans to expand that contract by $100bn, including Trainium chips as well as presumably more GPUs.

OpenAI has committed to use both the current Trainium3 chip and the next-generation Trainium4, currently planned for 2027.

On the software side, the two companies plan to develop a Stateful Runtime Environment powered by OpenAI’s models, which will be available through Amazon Bedrock. A Stateful Runtime Environment allows developers to keep context, remember prior work, work across software tools and data sources, and access compute.

AWS will also be the exclusive third-party cloud distribution provider for OpenAI Frontier, a platform for allowing companies to build, deploy, and manage teams of AI agents that operate across real business systems with shared context.

Finally, the two companies will develop customized models for Amazon's own developers for its customer-facing applications.

"OpenAI and Amazon share a belief that AI should show up in ways that are practical and genuinely useful for people," OpenAI CEO Sam Altman said.

"Combining OpenAI’s intelligence with Amazon’s infrastructure and global reach helps us put powerful AI into the hands of businesses and users at real scale.”