This service enables locking in GPU instance clusters from one to 64 units, reservable up to eight weeks ahead for durations up to six months. It supports P6-B200 and P6-B300 instance types, delivering low-latency, high-throughput connectivity through UltraClusters for ML tasks such as pre-training or inference surges.
Capacity Blocks can be shared across accounts using AWS Resource Access Manager, optimizing reserved capacity usage across organizational workloads. This offering addresses compliance-heavy sectors needing reliable GPU resource availability in GovCloud.
Source & further reading
Amazon EC2 Capacity Blocks for ML is now available in AWS GovCloud (US) Regions ↗Source published 2026-06-12. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.