February 28, 2026 – OpenAI and Amazon have just unveiled a multi – year strategic partnership that is set to reshape the AI landscape.
In a major financial move, Amazon is investing a staggering 50billioninOpenAIovermultipleyears.Theinitialinvestmentis15 billion, with an additional $35 billion to be injected in the coming months, subject to certain conditions being met.
A key aspect of the collaboration is the joint development of a “Stateful Runtime Environment” based on OpenAI’s models. This cutting – edge environment will be made available through Amazon Bedrock. The stateful developer environment represents the next generation of front – line model applications. It enables models to seamlessly access elements like computing power, memory, and identity. Developers can retain context, remember previous work, collaborate across software tools and data sources, and call on computing resources. It is specifically designed for handling ongoing projects and processes.
These stateful developer environments will be trained to run optimally on AWS infrastructure. They will be deeply integrated with Amazon Bedrock AgentCore and other infrastructure services. This integration ensures that customers’ AI applications and intelligent agents can work in harmony with other infrastructure applications running on AWS. The official launch of this environment is expected in the next few months.
AWS is also set to become the exclusive third – party cloud distribution provider for OpenAI’s Frontier platform. As the demand for AI deployment across various industries accelerates, this move will expand access to OpenAI’s most advanced enterprise platform. The Frontier platform allows organizations to build, deploy, and manage teams of AI intelligent agents. These agents can operate in real – world business systems, share context, have built – in governance and enterprise – level security features, and do not require the management of underlying infrastructure.
Building on their existing multi – year agreement worth 38billion,OpenAIandAWSarefurtherexpandingtheircooperation.Overthenexteightyears,anadditional10 billion will be invested. This expansion includes OpenAI’s commitment to consuming about 2 gigawatts of Trainium capacity through AWS infrastructure. This is to support the needs of the stateful runtime environment, Frontier, and other advanced workloads. The agreement is expected to reduce the cost and improve the efficiency of large – scale intelligent production.
Under this framework, OpenAI will have long – term computing power guarantees. It will also work with AWS to deploy dedicated chips within its overall computing ecosystem. This allows enterprises to access intelligence on demand without having to manage the underlying infrastructure. The commitment covers Trainium3 and the upcoming Trainium4 chips.
Amazon has announced that Trainium4, expected to be delivered in 2027, will bring another significant performance boost. It will feature a substantial increase in FP4 computing power, larger memory bandwidth, and higher – bandwidth memory capacity.
Finally, OpenAI will collaborate with Amazon to develop custom models. These models will be used by Amazon developers to support Amazon’s consumer – facing applications. Amazon’s teams will be able to customize OpenAI’s models for AI products and intelligent agents that directly serve customers.
