Dashboard · United States · data to 12 October 2026
US states: data centres, power and AI
Data centres per million people
Virginia: 40.3 per million, 1st of 51. That is the highest.
States, ranked
Reference lines: US 4.7 per million · Europe 2.7 per million. US: all 51 units. Europe = EU-27 + UK, Norway, Switzerland, Iceland, computed by country_metrics with the same rule.
How to read this. Sites mapped in OpenStreetMap as data centres (telecom=data_center, or a building tagged data_center), duplicates removed, per million residents. Counts what volunteers have mapped: colocation halls are well covered, some hyperscale campuses are mapped only as unnamed buildings. Neither higher nor lower is better in itself.
Why it matters. Where the existing data-centre industry clusters once population is taken out.
Source: OpenStreetMap contributors (ODbL), via Overpass; placed on us-atlas state outlines, data for 2026-10-11, pulled 2026-10-11 · measured · 51 of 51 states with a figure · source · Compiled by StackPath
Checked against a second source: 11 of 12 agree
Third-party figures are only as good as their method, so each is set against another source. Disagreements are reported, not smoothed over.
| What is checked | Source A | Source B | Gap | Result |
|---|---|---|---|---|
| Top 10 data-centre states, United States 6 of the Census top 10 are in ours (CA, TX, VA, OH, NJ, GA). NAICS 518210 counts business establishments in computing infrastructure, data processing and web hosting, which follow population and offices more than data-centre buildings. Agree = at least 7 of 10. | 6 of 10 states | 10 of 10 states | -40% within 30%? | ✗ differ |
| Data centres mapped (OpenStreetMap), United States Same OpenStreetMap run and duplicate rule; a gap would mean US sites falling outside every state outline (none did). | 1,612 sites | 1,612 sites | 0% within 1%? | ✓ agree |
| Frontier AI data-centre power operating now, United States Same Epoch pull and summing rule; small gaps come from per-site rounding; every US site has a state. | 15,179 MW | 15,179 MW | 0% within 1%? | ✓ agree |
| Frontier AI data-centre power planned, United States 67 sites placed from Epoch's address field, 3 from the site name, 4 from the sources Epoch lists, 3 from a source Epoch cites (checked by hand, research/datasets/us_states.json). | 43,412 MW | 43,412 MW | 0% within 1%? | ✓ agree |
| Population, United States An exact match means the World Bank uses the Census estimate for the same year: a check that both are the same vintage, not two independent counts. Census excludes Puerto Rico and other territories. | 341,784,857 people | 341,784,857 people | 0% within 1%? | ✓ agree |
| Top 10 data-centre states, United States 8 of EPRI's top 10 are in our top 10 (VA, TX, CA, OR, AZ, IA, GA, WA). EPRI ranks electricity use in 2023, we rank mapped sites in 2026: a state with a few huge campuses ranks higher on power than on site count, and OpenStreetMap misses some hyperscale sites. Agree = at least 7 of 10. | 8 of 10 states | 10 of 10 states | -20% within 30%? | ✓ agree |
| Industrial electricity price, United States Both read the same EIA table; a gap would mean the caches are from different releases. | 9.03 US¢/kWh | 9.03 US¢/kWh | 0% within 1%? | ✓ agree |
| Industrial electricity price, United States Two EIA series for different periods (this year to date against last full year); prices rose in between. | 9.03 US¢/kWh | 8.51 US¢/kWh | 6% within 15%? | ✓ agree |
| Electricity generated, 2025, United States Ember's release (2026-08-19) came before EIA's final 2025 data (September 2026). Solar explains 89 of the 95 TWh gap (Ember 389 TWh against EIA's utility-scale 300 TWh), consistent with Ember adding small-scale (rooftop) solar, which this EIA table leaves out (our reading). | 4,519.8 TWh | 4,425 TWh | 2% within 3%? | ✓ agree |
| Generating capacity under construction (2026-08), United States Same EIA file and status rule; units outside the 50 states and DC are in the US total but not in a state. | 98,731 MW | 98,731 MW | 0% within 1%? | ✓ agree |
| Data centres' share of electricity, Virginia Different denominators: EPRI estimates all data-centre use in the state; Dominion reports one utility's sales. Dominion's later 10-Ks give 26% for 2024 and 28% for 2025. | 25.6 % | 24 % | 7% within 25%? | ✓ agree |
| US data-centre electricity use, 2023, United States Two independent estimates with different methods. EPRI's Feb 2026 update puts 2024 at 184 TWh. | 152.1 TWh | 176 TWh | -14% within 25%? | ✓ agree |
More dashboards
Europe
31 countries: data centres, power, AI chips and AI use
DashboardIreland by council
31 councils: grid draw, campuses, planning decisions, on-site power and rates
DashboardWorld
Every country: data centres, AI compute, power and its price
DashboardOur coverage
How deep our data goes in each country, and what to get next
Every view keeps its measure and layout in the address bar: copy the link and the reader opens the same view. Data build 2026-10-10. How we know · corrections.