Amazon Web Services and Nvidia announced plans to deploy an additional two million Nvidia GPUs into AWS’s global infrastructure by 2028, as part of a multi-year expansion of AI computing capacity.
The collaboration includes the integration of Nvidia’s next-generation Blackwell Ultra, Rubin, and Rubin Ultra GPUs, alongside Nvidia RTX PRO 4500 Blackwell Server Edition GPUs for Amazon EC2 G7 instances. AWS will be the first major cloud provider to offer RTX PRO 4500-accelerated compute instances, according to the companies.
As part of the initiative, AWS and Nvidia will build AI factories for the U.S. government, housing 100,000 GPUs within AWS’s secure infrastructure to support classified federal and national security workloads at Impact Level 6 and above.
The expansion also includes the integration of Vera CPU-based infrastructure and the rollout of Nvidia NVLink Fusion with custom high-bandwidth memory technology. This enhancement is designed to improve memory access speeds and energy efficiency for AWS’s Trainium chips.
AWS highlighted performance gains from GPU acceleration, noting that Amazon EC2 G7 instances using Nvidia GPUs deliver data processing speeds up to 3.7 times faster and a 30% better cost-performance ratio compared with CPU-based configurations. Additionally, GPU-accelerated vector indexing on Amazon OpenSearch Service is reported to be up to nine times faster at one-quarter of the cost.
Beyond cloud computing, Amazon Robotics will integrate Nvidia’s physical AI platform to support next-generation robot development. Nvidia’s Nemotron open models will be made available on Amazon Bedrock and Amazon SageMaker, expanding the accessibility of AI tools for enterprise and government applications.
Matt Garman, CEO of AWS, stated that customers seek "the freedom to choose the best tools for their AI workloads." Jensen Huang, founder and CEO of Nvidia, emphasized that the partnership is expanding "across the entire stack — GPUs, CPUs, networking, open models, and software."
