← November 21, 2026 briefing

Amazon SageMaker HyperPod adds Global Inference Gateway: one endpoint routes inference across clusters and regions with automatic failover

Announced November 20, 2026

Verified (fact)

AWS introduced Global Inference Gateway, a managed single endpoint for SageMaker HyperPod. It sends inference requests to HyperPod clusters within one region or across multiple regions, selects the most suitable cluster, and automatically fails over to another cluster when one goes down. All requests are authenticated with SigV4 and IAM over AWS PrivateLink.

Why it matters (interpretation)

Teams serving models on their own clusters no longer need to build cross-region routing and failover themselves, easing the burden of securing GPU capacity and managing availability.

Sources

More about AWS