What the live data covers right now, which utilities we can’t see, and how that compares with other outage trackers.
As of the most recent poll (October 7, 2026): how much of the national total we can pin to a specific county versus what we only know at the utility or state level. We never spread an un-localized total across counties by estimate. The feeds behind it are on sources & health.
Customers out
24,409
National total, all sources
Localized to counties
23,634
Placed on the county map
Not yet localized
775
Utility/state totals only: in state figures, not the county map
Sources live
797/801
Reporting successfully this poll
Customers under active watch
97,951,649
A measured floor, not an estimate: it sums the served-customer counts only from sources that publish one (so their “% out” is real). Feeds without a denominator contribute 0 here, so the true number of customers we watch is higher than this figure, never lower.
Who we can see
Every US electric customer counted by the utility that serves them, residential, commercial and industrial customers from , split by how we see that utility live right now. This is coverage of utilities’ customers, not of outages.
EIA-861 2024
Live · county
0.0%
0 customers · Reported by county every poll; 0 out is a real all-clear
Live · total only
84%
115,613,726 customers · Utility total every poll, not placed on counties
Outages only
6.8%
9,412,755 customers · Appear only while they have outages
No live feed
9.3%
12,830,204 customers · No live feed; outages here are not in our numbers
Our live coverage, 125,026,481 customers with any live feed (91%), of which 0 county-level (0.0%), against each tracker’s own published coverage claim. The gap is computed against their published claim, not their data: customers they report tracking that we have no every-poll feed for.
We don’t use or republish any other tracker’s data; these are their own public statements, copied by hand on the date shown.
How this is counted
County-level + not-localized always equals customers out. A source that reports only a utility- or state-wide total counts toward the state figure but is never spread across counties by estimate.
The split is taken from the merged rollups, not by summing raw source output: where one source authoritatively aggregates a state, the others’ rows there are dropped so nothing is double-counted.
“Customers under active watch” is a measured floor from sources that publish a served count; it is not an estimate of every customer we observe.
Connector status is per utility and state, a utility live in one state says nothing about another: Live · county means a feed reports it by county every poll, zero-out counties included, so 0 out is a real all-clear; Live · total only means its utility total every poll; Outages only means it appears in a feed only while it has outages, so 0 out may just mean nothing was reported; No live feed means we are blind to it. Customer shares use each utility’s EIA-861 2024 customers (residential + commercial + industrial).
The denominator is every customer on the wires, counted once. A utility’s customers are its bundled customers (it sells them energy and delivers it) plus its delivery customers (a competitive supplier sells the energy; the utility still owns the poles and restores the outage). Retail-only suppliers hold no service territory and file their own rows for customers already counted on the wires utility, so they are excluded, as is EIA’s 99999 state adjustment. A utility with no EIA service territory is not in the denominator at all, which is the limit of this measure rather than a claim of coverage.
County-level reports come from utilities’ own public outage maps. Where a map hides a small outage as “under N customers”, we count it as a masked outage (a count, never added to customers out), rather than guessing a number.
When a source fails a poll we keep its last good numbers, marked Stalewith the real “last fetch” time, for up to ~40 minutes, never silently dropped to zero. Past that it’s marked Down and its regions show a gap rather than stale data.