SageMaker HyperPod adds EFA-only network interface option
Amazon SageMaker HyperPod now supports EFA-only network interfaces, allowing users to configure dedicated Elastic Fabric Adapters without requiring traditional Elastic Network Adapters for IP networking. This change is crucial for large-scale AI/ML model development, enabling further scaling of training clusters by preventing IP address exhaustion within VPCs. The feature is available in all AWS Regions where SageMaker HyperPod is supported and is configured via the CreateCluster/UpdateCluster API.
- →EFA-only network interfaces for SageMaker HyperPod
- →Scale AI/ML clusters without IP exhaustion
- →Configuration via API
Features (1) ›
- EFA-only network interfaces for SageMaker HyperPod
Amazon SageMaker HyperPod now supports an EFA-only network interface option for cluster instance groups. This allows dedicated Elastic Fabric Adapters without traditional Elastic Network Adapters for IP networking, optimizing for inter-node communication bandwidth in large-scale AI/ML training.
Enhancements (1) ›
- Scale AI/ML clusters without IP exhaustion
The EFA-only configuration prevents IP address consumption on interfaces not requiring IP networking. This enables larger SageMaker HyperPod clusters within existing VPC subnets, mitigating the risk of IP address exhaustion.
Notes (1) ›
- Configuration via API
To enable EFA-only interfaces, specify 'efa-only' within the ClusterNetworkInterface configuration when creating or updating a SageMaker HyperPod cluster using the CreateCluster/UpdateCluster API.
https://aws.amazon.com/about-aws/whats-new/2026/06/amazon-sagemaker-hyperpod-efa-only/
Related releases
- AWS Glue zero-ETL Integrations Gain Table Property Ownership and Conflict Detection AWS What's New ·
- SageMaker JumpStart adds NVIDIA Qwen3.6-35B-A3B-NVFP4 and Alibaba Wan2.1-T2V-1.3B models AWS What's New ·
- Amazon SageMaker JumpStart Adds Granite-Speech, Kanana-2, and OpenFold3 Models AWS What's New ·
- Mistral AI's Ministral-3 Models Now on Amazon SageMaker JumpStart AWS What's New ·
- Gemma-4-31B LLMs from Google DeepMind and NVIDIA now on SageMaker JumpStart AWS What's New ·
- SageMaker HyperPod adds model caching for faster inference autoscaling AWS What's New ·