Research on infrastructure at AI scale

Energy Infrastructure for Hyperscale Data Centers

A plain-language guide to building a 1 GW gas plant for AI compute, overlaid with the data center it serves.

A layered reference covering the full physical infrastructure stack required to deliver power to hyperscale AI workloads: gas supply and firm transport, large power transformers and switchgear, turbines and OEM lead times, water at scale, labor and EPC capacity, capital markets, interconnection queues, and the cross-constraint compounding that makes single-input planning insufficient. ERCOT primary, PJM (Ohio and Pennsylvania) comparison. 29 deep dives plus methodology layer with confidence levels and sensitivities.