AWS and NVIDIA Expand Partnership to Deploy 3 Million GPUs

Amazon Web Services (AWS) and NVIDIA have expanded their strategic collaboration to significantly increase AI compute capacity and integrate next-generation infrastructure. AWS will deploy 2 million additional NVIDIA Blackwell Ultra, Rubin, and Rubin Ultra GPUs across its global infrastructure between 2027 and 2028, adding to a previous plan to deploy 1 million GPUs starting in 2026.

The partnership extends to the development of secure AI factories for the U.S. government, providing 100,000 GPUs for national security workloads classified at Impact Level 6 and above. The companies are also integrating NVIDIA Vera CPU-based infrastructure into AWS and incorporating NVIDIA's physical AI platform into Amazon Robotics to advance warehouse automation.

Technical upgrades include extending NVIDIA NVLink Fusion with custom high-bandwidth memory for Amazon's Trainium chips and accelerating data processing on Amazon EMR and vector indexing on Amazon OpenSearch using NVIDIA CUDA-X libraries. Jensen Huang stated the expansion aims to make agentic and physical AI real at an unprecedented pace and scale.