Data status

19 sources are checked on a schedule; when one publishes new data, the site rebuilds and republishes. Last build: 2026-10-28 16:31 UTC. 7 sources brought new data in the past 7 days; 19 were added in the past 30 days. A source is marked late when it goes longer than expected without new data.

Analysis

The analysis reruns in full whenever any source brings new data: capacity cost per home, the regressions, the case studies, county and utility data center load, and neighborhood costs. Findings page figures and wording come from that run. Results change mainly when EIA publishes a new month of utility sales (about two months after the month ends). Three inputs are entered by hand when published, so they lag until then: PJM capacity auction results, the market monitor's data center share, and the curated utility tables. The Methods page describes the method; its example figures carry an as-of date.

Coverage of residential customers

EIA's monthly survey, the source of the site's utility prices and bills, does not cover every home. Homes are missing for two reasons: some buy supply from a competitive supplier or a town aggregation program; the utility still delivers their power, but EIA's monthly survey counts them only in a statewide total, not by utility; others are served by smaller utilities (mostly co-ops and municipal utilities) that report to EIA only once a year, so they are not in the monthly figures. States below 90% coverage in 2026-08 are listed; split by reason using EIA's annual survey (2025). The Your utility page and the map flag affected utilities, and estimate what homes on competitive supply pay where EIA reports the utility's delivery-only customers.

Findings checks

Each statement on the Findings and Methods pages is stored with the condition it rests on (for example, that a confidence interval excludes zero) and with plausibility ranges for its numbers. Every analysis run checks all 57 statements. A statement that no longer fits the data is shown as current figures without interpretation, marked under review, and opened as a GitHub issue until it is rewritten and reviewed. Last check: October 28, 2026.

All statements fit the current data.

Data center data quality

Pairs pending review are nearby records with a shared operator or similar name that the automatic rules could not settle, such as a campus total listed alongside its buildings, or an operating building next to a proposed expansion. They are kept, so some double counting may remain. Report a correction (no account needed), or open a GitHub issue.

Consistency check. Operating sites total about 59 GW of estimated capacity. At typical 50-70% utilization that is roughly 259-362 TWh a year, versus Lawrence Berkeley National Laboratory's estimate of 176 TWh of US data center use in 2023 (December 2024 report). The gap is expected: listed capacity is often full planned build-out, and use has grown since 2023. Totals are best read as upper bounds.

Hand-entered references (PJM capacity auction results, market monitor findings, curated utility tables) are updated when published; see Methods.