AWS and OpenAI announce multi-year, $38B strategic partnership (www.aboutamazon.com)

AWS and OpenAI announced a multi-year strategic partnership worth $38 billion that gives OpenAI immediate and expanding access to AWS’s cloud infrastructure to run and scale its most intensive AI workloads. The deal delivers hundreds of thousands of state-of-the-art NVIDIA GPUs (including GB200s and GB300s) via Amazon EC2 UltraServers, with the ability to scale to tens of millions of CPUs and capacity targeted to be deployed by the end of 2026 (with further expansion possible into 2027 and beyond). AWS points to its experience operating very large clusters (hundreds of thousands of chips) and emphasizes price, performance, scale and security as core benefits for OpenAI’s training and inference needs. Technically, the deployment centers on low-latency, high-throughput GPU clustering on the same network fabric—enabling efficient multi-node training and agentic, real-time inference at scale. That architecture supports both serving ChatGPT-class workloads and training next-generation foundation models, while making OpenAI’s open-weight models available to AWS customers via Amazon Bedrock. For the AI/ML community this means accelerated frontier-model development and more accessible large-scale inference for enterprises, but also further concentration of leading-edge compute with a major cloud provider, which has implications for competition, pricing and ecosystem dynamics.