.svg){kind=spacer}
Details
- NVIDIA and AWS have expanded their strategic partnership at AWS re:Invent 2025, integrating NVIDIA’s advanced interconnect and software with AWS’s custom silicon and cloud infrastructure.
- AWS will incorporate NVIDIA NVLink Fusion into its next-generation Trainium4 chips for AI inference and agentic training, as well as Graviton CPUs and Nitro System virtualization, streamlining large-scale deployment and management.
- The new AWS AI Factories provide regionally controlled, sovereign AI infrastructure powered by NVIDIA Blackwell GPUs, helping organizations safeguard proprietary data and comply with local regulations.
- NVIDIA Nemotron open models are now available on Amazon Bedrock, empowering developers to create agentic AI applications for text, code, images, and video. Robotics companies are leveraging NVIDIA Cosmos world foundation models on Amazon EKS for real-time control and simulation, with early adopters including CrowdStrike and leading robotics firms.
- Notable software enhancements include GPU-accelerated, serverless vector search in Amazon OpenSearch using the NVIDIA cuVS library, achieving up to 10x faster indexing at a fraction of the previous cost. NVIDIA was recognized with the AWS Global GenAI Infrastructure and Data Partner of the Year award for its leadership in the ecosystem.
Impact
This expanded collaboration strengthens AWS's stance against cloud rivals like Google Cloud and Microsoft Azure, especially in enterprise generative AI and data sovereignty. By advancing agentic and robotics-related AI capabilities, the partnership underscores a growing convergence in enterprise and embodied AI. Rapid sector adoption is expected as NVLink Fusion sets a new benchmark for hyperscale AI deployment.
