AI is driving a data-center power boom. Here are some of the biggest US projects being built.
Executive Take
AI growth now runs into a hard physical limit: power supply, not chips, decides who wins. Leaders betting on AI capacity need energy contracts and local permits locked in years ahead, not just compute deals.
Executive Summary
Business Insider catalogs major US AI data-center projects underway, including Meta's Hyperion (Louisiana, 5GW, $50B+), Fermi America's Project Matador (Texas, pursuing 17GW), and Ohio's PORTS campus (up to 10GW, 8GW contracted to OpenAI). US data centers used 313 terawatt-hours in 2025, nearly 40% of global data-center electricity, and could consume 11.8% of US electricity by 2030.
Why It Matters
Technology and AI leaders should care because compute capacity is now gated by electricity and water access, not just GPU supply or funding. Local backlash and lawsuits over power and water use could delay or kill planned AI infrastructure that companies are counting on.
Bizquad Perspective
Most leaders are tracking GPU supply and model benchmarks, but the real bottleneck for AI scale is now grid interconnection queues and local political approval, which move far slower than chip cycles.