Amazon Web Services will deploy an additional two million NVIDIA GPUs across its global infrastructure by 2028 under an expanded partnership with NVIDIA, the companies announced on Wednesday.
The collaboration will integrate NVIDIA’s latest GPU architectures—Blackwell, Rubin, Rubin Ultra, and RTX PRO 4500 Blackwell Server Edition—into AWS services including Amazon EC2, Amazon Bedrock, Amazon SageMaker, and Amazon EMR. The integration aims to enhance AI workload performance and scalability for enterprise and government customers.
AWS will also allocate 100,000 NVIDIA GPUs to build AI factories within a secure AWS infrastructure dedicated to U.S. government workloads. These facilities will support classified workloads up to Impact Level 6 and above, meeting federal security standards for national security applications.
The expanded partnership includes broader adoption of NVIDIA’s Vera CPU and Nemotron open model family, available on Amazon Bedrock and Amazon SageMaker. AWS will also leverage NVIDIA’s physical AI platform for Amazon Robotics, incorporating Jetson modules, Omniverse libraries, and the Isaac development platform for robotics applications.
Matt Garman, CEO of AWS, stated that customers seek flexibility in selecting AI tools while ensuring seamless integration across platforms. Jensen Huang, founder and CEO of NVIDIA, noted that demand for AI infrastructure is outpacing forecasts, with the AWS-NVIDIA collaboration serving as a key growth driver in the AI era.
Deployment of the additional GPUs is scheduled between 2027 and 2028, aligning with AWS’s infrastructure expansion plans.
