U.S. data-center power demand by state & region

AI's electricity load isn't spread evenly — it concentrates onto a handful of grids. Here's where, and why each market matters.

A Gridlas analysis · public EIA / LBNL / ERCOT / CBRE data · current as of late 2025 · Updated Jul 2026

4.4% → 6.7–12%

DC share of U.S. power, 2023 → 2028

~226 GW

ERCOT large-load queue (Nov 2025)

~4.5 yrs

median interconnection wait

Data centers consumed 4.4% of U.S. electricity in 2023, a share LBNL projects will reach 6.7%–12% by 2028. But the national average hides the real story: the load lands on specific grids, around the clock, faster than utilities have ever had to respond. These five markets are where that collision is sharpest.

🧮 Try the free estimator — size any facility in any of these markets and see its power, grid wait and water footprint. →

National map of U.S. data-center clusters over the power grid — Major U.S. data-center clusters over the power grid. Source: public EIA / CBRE / JLL (late 2025); clusters approximate.

Regional deep-dives

More markets: Silicon Valley & Santa Clara — the mature, power-capped market · Chicago & Illinois — the nuclear-powered Midwest hub · New York Tri-State — the financial-latency hub · Hillsboro & Oregon — the Pacific NW hydro & fiber hub · Memphis & xAI Colossus — the grid bottleneck, in the open · Iowa — the wind-powered heartland cluster · Salt Lake City & Utah — the state that legislated the workaround · Reno & Northern Nevada — the speed-to-power frontier · Nashville & Middle Tennessee — TVA's gas-fed power corridor · Kansas City — the Midwest breakout, Evergy's gas pivot →

The problem isn't total U.S. capacity — it's location and timing. Compute scales in months; power scales in years.

Want the full picture? The Gridlas report has all five regional deep-dives in depth, the ranked metro tables, high-res maps, a 12-month outlook, and the underlying dataset (CSV/GeoJSON) — built entirely from public EIA, LBNL & ERCOT data.

Get the report →

Prefer the free 3-page summary first? The 8 findings and the chart that explains the AI power crunch — straight to your inbox.

No spam. Unsubscribe anytime.

Sources: U.S. EIA (public domain); LBNL 2024 Data Center Energy Usage Report & "Queued Up" 2025; ERCOT (via Latitude Media / Utility Dive); CBRE / JLL H2 2025. See the full methodology & sources.

Gridlas · independent & unaffiliated · built from public data. · Read the full analysis → · The grid bottleneck → · Powering around the queue → · The water problem →