Amazon opens $11 billion AI data center in rural Indiana as rivals race to break ground (www.cnbc.com)

Amazon has quietly brought Project Rainier — an $11 billion, 1,200‑acre AI campus in New Carlisle, Indiana — online, marking one of the largest operational AI supercomputing sites in the world. Built in roughly a year, seven buildings are already live (30 planned) and the site will eventually draw ~2.2 GW of power. Rainier is dedicated to training and serving Anthropic’s models and is notable for deploying AWS’s custom Trainium silicon at scale: roughly 500,000 chips are already running there (Amazon expects ~1 million by year‑end) and Trainium 3, co‑designed with Anthropic, is slated to arrive soon. AWS says this is the largest known deployment of non‑Nvidia compute and touts higher density and energy efficiency over GPU‑based designs. For the AI/ML community, Rainier is a concrete signal that hyperscalers are moving from plans to live, bespoke infrastructure optimized for frontier model workloads. Technical implications include tighter hardware–software co‑design (AWS + Anthropic) to reduce latency and energy per flop, a push toward multi‑chip/multi‑cloud strategies (Anthropic also tapped Google TPUs), and material impacts on cost, power demand, and deployment timelines. Rainier underscores how custom silicon and integrated stacks can reshape training economics and capacity — but it also raises concerns around local land use, grid stress, and the true scale of capital and energy commitments required to sustain next‑generation AI.