OpenAI and Amazon have signed a $38B strategic partnership agreement

OpenAI and Amazon Web Services (AWS) have signed a multi-year agreement that will allow OpenAI to launch and scale its AI systems on AWS infrastructure. The collaboration begins today and will last for at least seven years.

This is stated in a statement from OpenAI.

Under the agreement, OpenAI will have access to hundreds of thousands of NVIDIA graphics processing units (GPUs) hosted on Amazon EC2 UltraServers, as well as the ability to scale computing to tens of millions of central processing units (CPUs). This will enable the company to develop its generative models faster, including those underlying ChatGPT.

The total value of the partnership is estimated at $38 billion. All planned computing power is expected to be launched by the end of 2026, with the possibility of further expansion in 2027.

The infrastructure that AWS is building for OpenAI has an optimized architecture for handling large volumes of AI tasks.

Clustering NVIDIA GB200 and GB300 graphics processors ensures minimal latency in data processing and enables efficient execution of both ChatGPT inference and new model training.

The partnership continues the companies’ previous collaboration: this year, OpenAI models became available on the Amazon Bedrock platform, where they are already being used by thousands of customers, including Peloton, Thomson Reuters, Comscore, Verana Health, and others.

Recall that at the end of October 2025, Reuters, citing three sources familiar with the situation, wrote that OpenAI was preparing the ground for an initial public offering (IPO) that could value it at up to $1 trillion.

If this happens, the IPO could be one of the largest in world history.

Read more: Microsoft and OpenAI have signed a new agreement: the company is valued at $500 billion