Private equity poured $108B into data center deals in 2024 — triple the prior year — with global data center investment exceeding $61B in the first 11 months of 2025. Yet no technical due diligence framework exists for evaluating AI data center investments from an infrastructure engineering perspective.

Two emergent metrics have no authoritative champion: “tokens per watt” — the AI efficiency metric connecting inference economics to power infrastructure — and “resilience-adjusted TCO,” a concept nobody has formalized that factors downtime costs, failure recovery time, and checkpoint overhead into GPU cluster total cost of ownership. An estimated 30–65% of data center power capacity is stranded, creating a natural bridge between infrastructure engineering and investment thesis development.

URE covers infrastructure economics from the facility up: how data center design drives cost per token, how placement decisions shape inference latency, how sovereign AI economics differ from what the map suggests, and how the phases of technology adoption create both risk and opportunity for infrastructure operators.