NVIDIA and AWS Strengthen Collaboration for Large-Scale AI Production Deployment

NVIDIA and AWS are strengthening their infrastructure collaboration to help enterprises build large-scale AI systems while minimizing operational complexity. This collaboration focuses on delivering low-latency inference, fast vector search, and powerful GPU performance.

The newly launched Amazon EC2 G7 instances are equipped with NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs to support AI inference, graphics, and data analytics workloads. Compared to G6 instances, AI inference performance improves by up to 4.6x and graphics performance by up to 2.1x, while the NVIDIA cuDF library significantly accelerates Apache Spark-based data analytics.

Amazon OpenSearch Serverless has adopted the NVIDIA cuVS library as its default compute option to support GPU-accelerated vector indexing. This enables vector indexing speeds up to 10x faster and costs reduced to a quarter compared to CPU-based approaches when building RAG (Retrieval-Augmented Generation) and agentic AI systems.

In addition, AWS has obtained NVIDIA Exemplar Cloud certification for training workloads based on NVIDIA GB300. This means AWS meets the rigorous performance standards aligned with NVIDIA's reference architecture, allowing customers to confidently use optimized, high-performance infrastructure when training large-scale AI models.