AWS and OpenAI Seal USD 38 Billion AI Partnership

In a landmark agreement valued at USD 38 billion, AWS and OpenAI have formed a strategic multi-year partnership to facilitate advanced artificial intelligence workloads on AWS's infrastructure. This collaboration aims to augment OpenAI's computing capacity by leveraging AWS's price-point, performance, scale, and security.

OpenAI will gain immediate access to AWS's substantial compute power, including hundreds of thousands of NVIDIA GPUs and the capability to escalate to tens of millions of CPUs. The deployment is anticipated to conclude by the end of 2026, with potential growth into 2027. AWS highlighted its unique capabilities in securely and reliably managing large-scale AI infrastructure, noting its clusters exceed 500,000 chips. The infrastructure dedicated to OpenAI will utilize NVIDIA GB200 and GB300 GPUs, integrated via AWS's EC2 UltraServers on a low-latency network, aimed at supporting both real-time inference and the training of emerging models.

OpenAI CEO Sam Altman emphasized the necessity for massive and reliable computing in scaling frontier AI, asserting that their partnership with AWS strengthens the comprehensive compute ecosystem essential for advancing AI technology for universal benefit. Concurrently, AWS CEO Matt Garman affirmed that AWS's prime infrastructure will underpin OpenAI's innovative AI pursuits. This alliance builds on a pre-existing collaboration, wherein OpenAI's foundation models became accessible on AWS's Amazon Bedrock service, allowing millions of AWS customers to utilize them.