About Datacenter Atlas

Why it exists

Data centers are becoming major consumers of electricity, water and land, yet basic facts, where they are, who runs them, what they draw, what communities say, are scattered, paywalled or unsourced. The atlas puts those facts in one place, links each one to its origin, and says plainly what is not known.

How the data is built

• Facilities are merged from open sources (OpenStreetMap, PeeringDB, Wikidata) and operator disclosures, de-duplicated, and given permanent ids. Capacity, PUE and other metadata are added only when a source states them.
• Collection is partly automated, including AI-assisted research agents. Their output is held to the same rule as everything else: a value is published only with a public source that was retrieved and checked against it; unverifiable candidates are dropped.
• What is checked, and how. Merging, de-duplication, country resolution and the correction rules run automatically and are logged; every reported correction is examined before it is applied. The atlas does not claim that each of its thousands of records was individually reviewed by a person, which is why every value carries its source.
• Derived indicators (water stress, climate, grid proximity, social friction) are computed from named open datasets with published formulas and validation, see the methodology and the indicator register.

Editorial rules

• Measured, not estimated. Missing means unknown, never zero. Announced figures are labelled as announced.
• Every fact has a source you can open. Research summaries are written neutrally, in our own words.
• Indices describe places, not conduct. A high water-stress or friction score says something about a location, not about how a company operates.
• No advertising. The site carries no ads and no sponsored content.

Corrections

Anyone, including operators, can report an error from any facility, operator or conflict page (“Report an error”) or via the contact form. Please include a public source. Accepted corrections are applied to the data and listed in the release notes; known data issues are documented rather than hidden.

Reuse & credit

Maps, figures, texts and the free sample can be reused with credit: “Source: Datacenter Atlas, https://www.datacenter-atlas.com (data version 26.08-c2, accessed <date>)”. Data is under ODbL 1.0, content under CC BY 4.0, see licensing. The full inventory is not distributed; the free sample is.