Outage history

Power outages in the United States, 2014–2025

Twelve years of recorded outages, from the public archive kept by Oak Ridge National Laboratory, and the major incidents utilities reported to the Department of Energy.

Recorded outages

A customer-hour is one customer without power for one hour. Ten thousand customers out for three hours is 30,000 customer-hours.

The 10 worst recorded days

Dates are in UTC, so an evening event can appear on the next day. Several days in a row usually belong to one storm. The count is the highest total seen in two 15-minute records in a row, across every outage in the area at that moment.

† Some counties were reported with more customers out than Oak Ridge estimates they have, so this total is probably too high. It is shown as published.

Outage data: Brelsford, Christa; Tennille, Sarah; Myers, Aaron; Chinthavali, Supriya; Tansakul, Varisara; Denman, Matthew; et al. (2023). The Environment for Analysis of Geo-Located Energy Information’s Recorded Electricity Outages 2014-2025. figshare. Dataset. https://doi.org/10.6084/m9.figshare.24237376.v4 Used under the Creative Commons Attribution 4.0 license. Changes: Power Grid Watch aggregated the published 15-minute county snapshots of customers without power into daily county totals, then into daily, monthly and yearly state and national figures (peak simultaneous customers out, approximate customer-hours, counties reporting, share of 15-minute intervals with any record), with each peak taken as the highest total seen in two consecutive 15-minute snapshots so that one-snapshot glitches cannot set it, keeping one value per county per 15-minute interval where the files repeat a county; no values were imputed or zero-filled, and territories were left out.

By state

Pick a state for its own history, its worst days and its hardest-hit counties.

† Probably overstated in the source: some counties were reported with more customers out than Oak Ridge estimates they have.

Major incidents reported to the Department of Energy

A different record: the emergency reports utilities and grid operators must file with the Department of Energy. It goes back further, but it only covers events that cross a reporting threshold.

What the reports were about, all years

1. Severe weather1,962
2. Physical attack or vandalism1,518
3. System operations932
4. Transmission or generation equipment530
5. Fuel supply124
6. Cyber110
7. Other/unspecified16

How these figures were made

• Outage history comes from EAGLE-I recorded electricity outages 2014–2025 (Oak Ridge National Laboratory; Figshare version 4), used under the Creative Commons Attribution 4.0 license. Brelsford, Christa; Tennille, Sarah; Myers, Aaron; Chinthavali, Supriya; Tansakul, Varisara; Denman, Matthew; et al. (2023). The Environment for Analysis of Geo-Located Energy Information’s Recorded Electricity Outages 2014-2025. figshare. Dataset. https://doi.org/10.6084/m9.figshare.24237376.v4
• We changed the data by adding the 15-minute county records up to days, months, years and states. We did not fill gaps. The original files are unchanged in our archive.
• The counts are customer accounts or meters as utilities reported them, not people. The same customer is counted again in every 15-minute record while still out, which is why we show customer-hours and the highest count at one time, never a total number of customers affected. The highest count is the largest total seen in two 15-minute records in a row, so a one-record glitch cannot set it; it adds up every outage in the area at that moment.
• Incident reports come from the Department of Energy's Form DOE-417 annual summaries. A report that names several states appears under each of them but once in the national count.

Limits of the outage record

• Counts are electric customer accounts (meters) reported without power at 15-minute snapshots, not people, households or unique customers; customer-hours are an approximation.
• EAGLE-I's coverage grew a lot. 2014 covers only 1 November-31 December; Hawaii and Utah have no records before 2017 and Alaska and Nevada none in 2014; Montana, Wyoming, Alaska, Alabama, Oregon and South Dakota are thinly covered before 2017. ORNL estimates national customer coverage at about 92% only by 2022. A rise across years can reflect wider reporting, not more outages; compare years from 2018 on, and treat 2014-2016 as partial.
• A missing record is not a zero. Up to 2021 counties with no outage are usually absent from the files, so a quiet period and a reporting gap look alike; some 2017-2021 and 2025 rows have blank counts, treated as unknown. Explicit zero rows appear only from 2022.
• The 2014-2021 files stop at 00:00 UTC on 31 December, so the last day of those years is missing; there are also some multi-hour national gaps (listed per year).
• Peaks are the largest simultaneous total held for at least two consecutive 15-minute snapshots, because the source contains one-snapshot glitches far above neighbouring values.
• Days are UTC calendar days (evening outages in the western US fall on the next date).
• Some large-event totals are likely overstated by the source: counties sometimes report more customers out than ORNL models them to have (e.g. Ventura County, California, in January 2025; North Carolina during Helene in 2024). Such days are marked with a dagger (†) and are shown as published, not corrected.

More about sources and coverage