Amazon SageMaker AI Now Supports Capacity-Aware Inference with Automatic Instance Fallback

This article announces capacity-aware inference with automatic instance fallback for Amazon SageMaker AI endpoints.

• SageMaker AI automatically provisions from prioritized instance types when preferred capacity unavailable
• Supports Single Model Endpoints, InferenceComponent-based endpoints, and Asynchronous Inference endpoints
• Scales down by removing lowest-priority instances first, preserving preferred infrastructure
• Specify different optimized models per instance type or use SageMaker inference recommendations
• Per-instance-type CloudWatch metrics provide visibility into latency, throughput, and GPU utilization
• Available in 16 AWS regions globally

SageMaker AI now handles capacity constraints automatically, ensuring reliable endpoint creation and scaling without manual intervention.