AWS Project Rainier
New Carlisle, Indiana, US · operating · operated for Amazon (AWS)
Amazon's purpose-built training campus for Anthropic in northern Indiana — 30 planned buildings designed around AWS's own Trainium2 chips rather than GPUs, brought from ground-break to a half-million-chip cluster in about a year.
Compute Atlas, the sister dataset on supercomputer and AI-cluster supply chains, has the component-by-component record for this campus: Project Rainier (system).
What's documented
- An $11 billion campus announced April 25, 2024 on roughly 870 acres near New Carlisle; 30 data center buildings of about 200,000 square feet each are planned, designed for 2.2 GW. [source]
- Came online October 2025 with nearly 500,000 AWS Trainium2 chips dedicated to Anthropic; Anthropic's fleet across AWS was expected to exceed 1 million Trainium2 chips by the end of 2025. [source]
- Uses Trainium2 UltraServers with petabit-scale EFA networking; from October to March the buildings use no water for cooling, and only a few hours of water-assisted cooling per day in summer. [source]
- Amazon subsequently announced a further $15 billion for new northern Indiana campuses adding up to 3 GW of capacity. [source]
Suppliers reported inside
| Company | What | Source |
|---|---|---|
| Amazon (AWS) | Trainium2 chips and UltraServers (Annapurna in-house silicon) | source |
| Anthropic | anchor tenant / training workloads | source |
Placements come from the cited reporting, not operator disclosures, unless a source says otherwise.
The generic bill of materials
Whatever the vendor list, a facility like this contains the same part categories:rack-scale systems, direct-to-chip cold plates, coolant distribution units, large power transformers (345–765 kV) & substations, UPS (static, modular, rotary/DRUPS) & UPS power semis, backup generation: gensets, fuel systems, HVO, chillers & compressors — the full set is in the scales matrix.