AWS and NVIDIA Launch G7 Instances, GPU Vector Search Default, and Syslog Ingestion

AWS and NVIDIA announced a new EC2 instance family, a default GPU vector search capability, and a managed syslog ingestion service on June 23, 2026. The updates span compute, retrieval, and observability infrastructure for AI and traditional workloads.

The new Amazon EC2 G7 instances are powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. Compared with G6 instances, they deliver up to 4.6x AI inference performance and up to 2.1x graphics performance, according to NVIDIA's blog post.

EC2 G7 instances target multiple workloads

Each G7 instance supports up to eight GPUs with 256GB of total GPU memory, 700 Gbps of EFA-enabled networking, and up to 7.6TB of local NVMe SSD storage. Configurations include one, two, four, and eight GPU options, with bare metal coming soon.

The instances are designed for:

• AI inference with lower latency
• Graphics, rendering, and video workflows
• Spatial computing and virtual desktop infrastructure
• GPU-accelerated data analytics on Amazon EMR using NVIDIA cuDF
• Vector database and retrieval-augmented generation workloads

Access is available through AWS Deep Learning AMIs, Deep Learning Containers, Amazon EMR, EKS, ECS, and graphics AMIs. Support for Amazon SageMaker AI is coming soon.

GPU vector search becomes default in OpenSearch Serverless

Amazon OpenSearch Serverless now uses GPU-accelerated vector indexing powered by NVIDIA cuVS as the default compute choice for all vector collections. The change eliminates the need for customers to manually configure GPU-accelerated vector search.

NVIDIA reported that vector indexing is up to 10x faster at a quarter of the cost compared with CPU-only builds. This makes billion-scale vector databases practical to build in under an hour. The new default applies to agentic AI, semantic search, recommendation systems, and RAG applications.

AWS also achieved NVIDIA Exemplar Cloud status for GB300 training performance, meaning AWS meets NVIDIA's reference architecture benchmarks for large-scale AI training workloads.

CloudWatch Logs adds managed syslog ingestion

Separately, AWS announced managed syslog ingestion for Amazon CloudWatch Logs. Customers can send syslog messages from firewalls, routers, switches, and Linux servers over TCP, TCP+TLS, or UDP to a VPC endpoint without installing or managing agents.

CloudWatch Logs supports RFC 5424, RFC 3164, and Cisco FTD/ASA syslog formats. The service automatically parses incoming syslog messages. This reduces operational overhead for network and security teams that previously relied on third-party collectors or custom scripts.

The combination of GPU infrastructure for AI and managed log ingestion for operations reflects AWS's strategy to reduce friction for both AI and traditional infrastructure teams.