Overview
- AWS said Project Rainier is fully operational less than a year after it was announced, spanning multiple datacenters.
- The cluster currently runs on nearly half a million Trainium2 chips, with one Indiana site planned to reach 30 buildings and seven reported as active.
- Anthropic is already training and serving Claude workloads on the system, and AWS says it will scale past one million Trainium2 chips by year-end.
- AWS frames proprietary silicon and vertical control of the stack as a competitive edge, and reporting describes the setup as the largest non‑Nvidia AI cluster.
- Local officials and the utility expect peak demand to roughly double as build-out continues, fueling questions about environmental and community impacts.