Where we disagree with other trackers

Nobody audits data center trackers. The next best thing is to check them against each other in public, so a reader can see where two sourced datasets diverge and judge for themselves.

This page is that check, run on Sep 22, 2026 against Compute Atlas, whose dataset is published under CC BY 4.0. A disagreement is not proof that either side is wrong — trackers update at different times, and a project can move between the day they looked and the day we did. Where we differ, our evidence is in the last column and the full source trail is on the project page.

How the comparison works

Records are matched on a shared distinctive word in the project or developer name and geographic proximity in the same state. Proximity alone is never enough: two unrelated data centers can sit a few hundred metres apart in one industrial park, and a mix-up there would produce a disagreement that is really our own error. 13 Census records found no confident match and are left out rather than guessed at.

Statuses are compared through a crosswalk: proposed ↔ Proposed, approved ↔ Permitted, under construction ↔ Under construction, operating ↔ Operational, cancelled ↔ Cancelled. The 84 agreements are not listed individually; the whole comparison, including them, is in reference.json.

Compute Atlas data retrieved Sep 22, 2026 and used under CC BY 4.0. Their dataset is their work, not ours; any error in the matching above is ours, not theirs.