The AI data center buildout, checked at the source.

What a record holds

Every site is a real, named project: a powered campus becoming AI compute, from ERCOT and PJM to MISO and SPP. Each record is read the same way:

• Where it is: the city, the county, the state and the grid region.
• How big it is: the megawatts energized today and the total planned at full build, as far as they are disclosed.
• Where it stands: live, under construction or planned, with dates such as first power and full buildout.
• Who holds each layer: power and interconnection, site and build, compute, the cloud operator, the anchor tenant and the capital.
• The sources: every document the record rests on, linked from the site's page.

How a fact is verified

Primary sources come first: SEC filings and their exhibits, company releases, and utility and grid operator records. Press coverage can add a lead, and the record says so when it does. Each named party carries one of three tags.

Capacity follows the same rule. A record shows the number a source states. Where nothing is disclosed, it says undisclosed instead of guessing.

How it stays current

New SEC filings are checked every day, and a record changes when a filing changes it. Each page shows the month it was last checked.

Membership

The public view shows each site's basics and its public parties. Free membership, for professionals in AI infrastructure, opens the rest of every dossier: the anchor tenant, the capital behind the deal, and the Registry's read on the next milestone. You sign in with a link by email or with Google, and tell us your name, company and role once.

Corrections

See something wrong or out of date? Send us the source and we will check it. Records are corrected in place.