A global map of data centers and frontier AI clusters, laid over each country's AI-adoption intensity. The gap between the two — heavy use with little domestic compute — is where strategic dependency hides.
⚡ Globally, data centres use ~1.5% of electricity (~460 TWh, IEA 2022), heading toward ~1,000 TWh by 2026 — country-level energy-vs-grid analysis is in the GAIA report.
Country shading = share of global Claude.ai usage (GAIA adoption). Dots = data centers; hollow = announced/under construction.
Click a country on the map to see its compute snapshot.
The map is the free view. The decision-grade layer is available on request:
Methodology & honesty note. The faint dots are a real coverage layer of 4,300+ facilities ingested live from OpenStreetMap (locations real; OSM carries no capacity). The larger colored markers are a curated capacity layer of notable hubs & frontier AI clusters where MW/type/status are compiled from public announcements — approximate, and with operational kept separate from announced on purpose (the market routinely inflates "planned capacity"). The verified, provenance-tagged facility registry and all cross-analysis (dependency index, energy, fiscal) are available in the full GAIA dataset and Country Reports (contact for details). Layers: OpenStreetMap (coverage), Epoch AI-style tracking (frontier clusters), IEA (energy benchmarks). Country shading: GAIA adoption (share of global Claude.ai usage).