NVIDIA adds new AWS GPU infrastructure for AI workloads

NVIDIA has expanded its work with Amazon Web Services across Amazon EC2 and Amazon OpenSearch, adding GPU-backed cloud instances, default GPU-accelerated vector indexing and benchmarked infrastructure for large training jobs. The changes target compute for AI and graphics workloads, retrieval infrastructure for vector databases and cloud environments used for model training.

Amazon EC2 G7 instances now use NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs for AI inference, graphics, spatial computing and GPU-accelerated data analytics. The companies said G7 instances deliver up to 4.6 times the AI inference performance of G6 instances and up to 2.1 times the graphics performance. Hardware options support up to eight GPUs, 256GB of total GPU memory, 700 Gbps of EFA-enabled networking and up to 7.6TB of local NVMe SSD storage.

Amazon OpenSearch Serverless now uses GPU-accelerated vector indexing with NVIDIA cuVS as the default compute choice for all vector collections, targeting retrieval-augmented generation, semantic search, recommendation systems and agentic AI applications. NVIDIA said vector indexing is up to 10 times faster and costs a quarter as much as CPU-only builds. AWS has also achieved NVIDIA Exemplar Cloud status for NVIDIA GB300 training workloads, signaling that it met NVIDIA-defined performance standards for GB300-based training environments.