Amazon Web Services (AWS) and NVIDIA announced a major expansion of their long-running partnership to address surging demand for AI infrastructure. According to both AWS and NVIDIA, the companies plan to deploy 2 million additional NVIDIA GPUs across AWS's global infrastructure in 2027-2028, building on an earlier commitment of more than 1 million GPUs announced at NVIDIA GTC 2026. Wccftech reported that this expansion effectively brings AWS's total NVIDIA GPU commitment to as many as 3 million, tripling the original 1 million GPU deal.
Beyond raw GPU capacity, the expanded collaboration includes bringing NVIDIA's Vera CPU-based infrastructure to AWS, intended to support agentic AI workloads that need strong CPU performance alongside accelerators. The companies are also working to extend NVIDIA NVLink Fusion interconnect technology to support custom NVIDIA high-bandwidth memory (NVHBM) for Amazon's Trainium chips. Wccftech noted that NVIDIA has claimed NVHBM offers 30% more bandwidth and 15% higher power efficiency compared to HBM4E memory.
A significant portion of the expanded deal is aimed at government use: AWS and NVIDIA plan to build AI factories for the U.S. government that will include 100,000 GPUs running on AWS's secure infrastructure. Both sources say these systems are meant to support workloads classified at Impact Level 6 (IL6) and above for federal and national-security applications.
The partnership also covers deeper technical integration, including combining NVIDIA's platform with the AWS Nitro System and Elastic Fabric Adapter (EFA) for security and reliability, and continued support for NVIDIA's Nemotron open models on Amazon Bedrock and Amazon SageMaker. AWS and NVIDIA say they are accelerating data processing and vector search through NVIDIA's cuDF and cuVS libraries, with claims of up to 3.7x faster processing and 30% better price-performance on Amazon EMR, and up to 9x faster vector indexing at a quarter of the cost on Amazon OpenSearch Service. AWS also plans to expand Blackwell-based capacity, including RTX PRO 4500 GPUs for EC2 G7 instances, which AWS says deliver 4.6x AI inference performance and 2.1x graphics performance over the previous G6 generation.
On robotics, Amazon Robotics is adopting NVIDIA's physical AI platform—including Jetson, Omniverse, and Isaac—to advance warehouse automation and next-generation robot development, covering simulation, synthetic data generation, and real-world validation. AWS CEO Matt Garman said the expanded work aims to make AWS the best place to run NVIDIA AI technologies, while NVIDIA CEO Jensen Huang said demand for the partnership is running ahead of every forecast and that the companies are expanding across GPUs, CPUs, networking, open models and software.