EC2 G7e Instances Expand to EU and Asia, Boosting AI Inference
NVIDIA RTX PRO 6000-powered EC2 G7e instances are now available in Frankfurt, Stockholm, and Mumbai, offering up to 2.3x inference performance.
Editorial summary and commentary based on the original from AWS What's New. Read the original
EC2 G7e instances are now available in Frankfurt, Stockholm, and Mumbai, bringing NVIDIA RTX PRO 6000 Blackwell GPUs to more regions for AI and spatial computing.
What changed
- Amazon EC2 G7e instances are now available in AWS Europe (Frankfurt, Stockholm) and Asia Pacific (Mumbai) Regions.
- These instances feature NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, offering up to 2.3x inference performance compared to G6e instances.
- G7e instances support up to 8 GPUs with 96 GB memory per GPU, 192 vCPUs, and 1600 Gbps networking bandwidth.
Why it matters
This expansion brings high-performance GPU compute, specifically targeting large language models (LLMs) and generative AI workloads, to three new AWS regions. The inclusion of NVIDIA RTX PRO 6000 Blackwell GPUs, with their substantial memory (96 GB per GPU) and networking capabilities (1600 Gbps), directly addresses the growing demand for accelerated inference and training of complex AI models. The honest version: This is a move to capture more AI workload spend in these specific regions, offering a tangible performance uplift for compute-bound AI tasks. It should be considered for any workload that previously pushed the limits of G6e or similar instances.
The catch
While the announcement touts a 2.3x performance increase, this figure is likely for specific inference benchmarks and may not translate directly to all generative AI or spatial computing workloads. Watch out: The specific number of GPUs per instance type is not detailed, only the maximum. Furthermore, the announcement does not specify pricing differences between the new and existing regions, which can often vary significantly. Pairs with: Amazon SageMaker for managed ML deployments.
Ship it
If you are running LLM inference, agentic AI, or multimodal generative AI workloads and are currently in or planning to deploy to Frankfurt, Stockholm, or Mumbai, evaluate G7e instances. Compare their performance and cost against your existing G6e or other GPU instances in these regions before migrating. On-Demand, Spot, and Savings Plans are available options.
Bottom line: EC2 G7e instances with NVIDIA RTX PRO 6000 Blackwell GPUs are now in more EU and APAC regions, a significant boost for AI inference workloads.
— Filed to /aws
Source (AWS What's New): Amazon EC2 G7e instances now available in additional regions