NVIDIA and AWS are launching new EC2 G7 instances with Blackwell GPUs and integrating NVIDIA's cuVS vector search librar

NVIDIA and AWS are addressing the core challenges of large-scale AI deployment through coordinated infrastructure expansion. EC2 G7 instances, powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, bring new compute capabilities to AWS for AI inference, graphics, spatial computing and GPU-accelerated data analytics. These instances deliver up to 4.6x AI inference performance and up to 2.1x graphics performance compared with G6 instances, while significantly accelerating GPU-accelerated data analytics on Amazon EMR using the NVIDIA cuDF library for Apache Spark workloads.

The G7 architecture supports configurations ranging from single-GPU to eight-GPU setups, with 256GB of total GPU memory, 700 Gbps of EFA-enabled networking and up to 7.6TB of local NVMe SSD storage across one-, two-, four- and eight-GPU configurations plus bare metal. This flexibility allows customers to right-size infrastructure for their specific workloads rather than over-provisioning. The instances are accessible through AWS Deep Learning AMIs, Amazon Deep Learning Containers, Amazon EMR, Amazon EKS, Amazon ECS and graphics AMIs, with availability coming soon to Amazon SageMaker AI.

The second pillar of the expansion involves GPU-accelerated vector indexing in Amazon OpenSearch Serverless, now powered by NVIDIA cuVS as the default compute choice for all vector collections. This shift has direct implications for teams building retrieval-augmented generation, semantic search, recommendation systems and agentic AI applications. Vector indexing is now up to 10x faster at a quarter of the cost compared with CPU-only builds, making billion-scale vector databases practical to construct in under an hour. By integrating NVIDIA cuVS, AWS customers get serverless scaling that reduces operational overhead when workloads are idle while providing production-ready AI retrieval infrastructure.

AWS has also achieved NVIDIA Exemplar Cloud status for the NVIDIA GB300, meeting NVIDIA's rigorous performance thresholds for large-scale training workloads. This certification, the result of deep co-engineering between AWS and NVIDIA teams, allows developers and AI leaders to be confident in consistent, high-performance infrastructure for training, helping teams improve total cost of ownership and transition from planning to production more efficiently.