Methodology

How this map is compiled, verified, and kept current, and where its figures come from. The dataset covers 4,800 US data center facilities across 52 states and territories, refreshed as of Jul 2026. It is built entirely from public sources; no proprietary or paywalled datasets are used.

What This Dataset Covers

The map tracks physical data center facilities in the United States: all 50 states plus Washington D.C. and Puerto Rico. It does not cover facilities outside the US.

Facilities are classified by type (colocation, hyperscale, enterprise, and neocloud, the AI-focused operators) and by lifecycle status:

  • Operational: in service.
  • Under construction: actively being built.
  • Planned / approved: announced or permitted but not yet built.
  • Cancelled: publicly abandoned projects, retained as part of the buildout record.

Counting announced and planned projects alongside operational ones is deliberate: it is what makes the map useful for tracking where new capacity is heading. It is also part of why the total differs from operational-only counts elsewhere. See why counts differ between sources.

How Facilities Are Identified and Verified

Facility locations, operators, and attributes are compiled from public sources: company announcements and operator facility pages, government filings and permits, utility records, news reports, and public industry records. Records are cross-referenced across more than one source where possible, then deduplicated, geocoded, and classified by facility type based on operator, published capacity, and tenant model.

The dataset is refreshed on a rolling basis, roughly monthly, as new announcements and status changes are incorporated and existing records are re-checked. The current snapshot date is shown throughout the site as "data as of Jul 2026". Because facilities move through their lifecycle and new ones are announced constantly, any snapshot lags real-world activity to some degree; for the very latest position on a specific project, cross-reference the original source.

How Capacity Figures Work

Power capacity is reported in megawatts (MW) and represents nameplate or planned capacity, not necessarily current operational load. It is the most incomplete field in the dataset: capacity is published for only about one-third of facilities. Many colocation providers do not disclose facility-level power figures, and older facilities often predate public reporting. Aggregate MW totals therefore represent known capacity only, and the real total is higher. Facilities with no published capacity should not be assumed to be small.

Where a value is present, a capacity_estimated flag records whether it was directly reported or approximated:

  • Reported values are sourced from operator disclosures, permit filings, or verified industry reports.
  • Estimated values are approximated, for example from metro-level averages, figures extracted from unstructured descriptions, or inferred from reported building size using standard power-density ratios. Treat these as order-of-magnitude approximations, not precise measurements.

Aggregate totals can be skewed upward by a handful of very large planned hyperscale campuses (1,000+ MW each), so the median facility is much smaller than the mean.

Environmental and Energy Data Sources

Each facility is geocoded, then environmental and energy attributes are joined by geographic overlap. These come from public government and research sources.

Climate risk: FEMA National Risk Index (2025 release)

Composite natural-hazard risk scores from the 2025 release of FEMA's National Risk Index, evaluated at the county level across 18 hazard types (hurricanes, flooding, wildfire, earthquake, extreme heat, and others). The composite combines expected annual loss, social vulnerability, and community resilience. Each facility inherits the score of the county it sits in.

Water stress: WRI Aqueduct methodology

Water stress estimates follow the World Resources Institute's Aqueduct methodology, which measures the ratio of water demand to available renewable supply. Levels range from Low to Extremely High, assigned by the basin/county containing each facility.

Grid carbon & renewables: EPA eGRID (subregion level)

Grid carbon intensity (lbs CO2/MWh) and renewable generation share come from the EPA's Emissions & Generation Resource Integrated Database (eGRID2023, Revision 2, the latest official release), mapped at the eGRID subregion level.

Electricity prices: US EIA (industrial rates)

Industrial electricity rates (cents/kWh) are state-level figures from the US Energy Information Administration (EIA-861, 2024 final, the latest final annual release).

Grid and energy figures are updated annually by their respective agencies (EIA, EPA eGRID) and reflect the most recent official release available as of the Jul 2026 snapshot (checked July 2026). FEMA NRI scores use NRI v1.20 (December 2025), the current release.

Known Limitations

  • Geocoding precision varies. Most facilities are geocoded to a precise address, but roughly one in ten falls back to a city-centroid approximation. On a zoomed-out map this is invisible; zoomed into a city, several facilities may appear to stack at the center. Approximate-location facilities should not be used for precise distance or flood-zone analysis without verifying the actual address.
  • Capacity coverage is partial. Roughly two-thirds of facilities have no published capacity figure, so MW totals understate the true installed and planned base.
  • Planned projects can change. Pipeline status may lag real-world progress, and announced projects can be delayed, resized, or cancelled without public notice.
  • Environmental scores are regional. Climate risk and water stress are county/basin-level averages; actual risk at a specific site depends on its elevation, infrastructure, and mitigation measures.
  • Operator names are not fully normalized. The company field may reflect a tenant or parent entity, and acquisitions may leave a facility listed under a former name.

Nothing on this site constitutes investment, financial, or legal advice. It is intended for research and reference. For a deeper field-by-field breakdown, see the site's data quality documentation.

Corrections

Spotted a facility that is missing, misclassified, or out of date? Corrections are welcome and help. Email [email protected] with the facility name, location, and a public source, and it will be reviewed for the next refresh.