U.S. power outages by state

In EIA's 2024 Form 861 state totals, the customer-weighted average across about 155,585,529 reporting customers was roughly 611 minutes of sustained interruption per customer for the year when major event days stay in (~10.2 hours), versus about 126 minutes when those days are stripped out (~2.1 hours). Frequency landed near 1.53 interruptions per customer with major events included. South Carolina, Maine, North Carolina, Florida, Texas, and Georgia sit at the top of the with-major-events list — and most of that gap is storm-year math, not everyday feeder trouble.

Updated September 10, 2026

This is a supporting data page for planning a dark night, not a product review. The solar-generators hub maps the search term. The pencil for fridge, CPAP, lamps, and a sump is the load calculator. Appliance forks: fridge runtime, sump pump runtime, CPAP runtime.

What the year looked like

Two numbers matter before the table. SAIDI is average minutes of sustained outage per customer in the year. SAIFI is how many sustained interruptions the average customer saw. EIA also publishes each metric with major event days (MED) left in, and again with those days removed. The with-MED column is the storm-and-disaster year people remember. The without-MED column is closer to routine feeder reliability.

On the customer-weighted read of EIA's published state totals (Any Standard columns when present), 2024 looked like this:

• ~611 minutes SAIDI with major event days (~10.2 hours).
• ~126 minutes SAIDI without major event days (~2.1 hours).
• ~1.53 SAIFI with major event days.

That gap — roughly 486 minutes — is the year's major events talking. Florida is the clean example in the table: about 1,322 minutes with MED (~22 hours) versus about 66 minutes without. Same state, same customers, very different story once the storm days leave the average.

Highest with-MED SAIDI in this snapshot: South Carolina (~3,136 minutes / ~52 hours), Maine (~1,749 / ~29 hours), North Carolina (~1,441 / ~24 hours), Florida (~1,322 / ~22 hours), Texas (~1,271 / ~21 hours), Georgia (~1,229 / ~20 hours). Lowest: District of Columbia (~33 minutes), Arizona (~84), South Dakota (~86), North Dakota (~103), Massachusetts (~104).

Methodology

Source file: Reliability_2024.xlsx, sheet State Totals, from the U.S. Energy Information Administration's Form EIA-861 detailed data files (eia.gov/electricity/data/eia861/). Retrieved September 10, 2026. Data year on the rows is 2024.

For each state (and D.C.), we took EIA's published state totals. When the Any Standard customer count and SAIDI were present, we used those columns — broader utility coverage than IEEE-only reporters. Otherwise we fell back to the IEEE Standard columns. We did not rebuild state averages from the utility-row sheet; EIA already publishes the state totals we show.

Customer-weighted U.S. figures on this page are a desk aggregate: each state's SAIDI/SAIFI multiplied by that state's reported customer count, then divided by the sum of those customers (~155,585,529 in this pull). That is a transparent roll-up of the published state totals, not an official EIA national headline.

What we did not measure

• No live outage map, no PowerOutage.us scrape, no utility API ping on publish day.
• No county, ZIP, or feeder-level numbers — EIA's public state totals stop at the state.
• No NOAA storm-event re-attribution of which hurricane or ice storm drove each MED minute.
• No claim that your street matches the state average. Rural co-ops and urban feeders diverge.
• No Puerto Rico / Guam territory rows in the main table (territory reporting sits on a different sheet shape; leaving them out is honest, not a ranking slight).
• No invented "typical outage lasts X hours in your county" marketing line.

Sustained interruptions in this survey are the non-momentary kind utilities report on Form 861. Momentary blinks are a different column and are not what the SAIDI/SAIFI figures here summarize.

State table (2024)

Sort is alphabetical. Minutes and hours are the same SAIDI reading; hours are minutes ÷ 60 for readability. Customer counts are the reporting base EIA attached to the Any Standard (or IEEE fallback) totals.

Raw extract used to build this table: local desk file growth/data/eia861-2024-state-reliability.csv (not a public download URL). Re-pull from EIA when a newer Reliability year ships.

What this means for a backup plan

A state average is not a promise about your next Tuesday. It is a way to see whether the year you are planning against was a storm year or a quiet feeder year. If with-MED and without-MED are far apart where you live, sizing for "the week after the hurricane" and sizing for "a tree on the line in March" are different shopping lists.

For a ~2 kWh 120-volt portable station, the useful next step is still appliance math: closed fridge, sleep machine, sump start spike, lamps — then nights on the tank. That pencil lives here:

Open the load calculator — running watts × duty → Wh/day → nights on a 2,048 Wh tank at 90% usable. Arithmetic only.

Longer appliance essays: fridge · CPAP · sump. The night itself, cords and all: what a station really runs.

Hub: solar generators · pencil: load calculator · shopping night: solar generator for home backup · fridge: fridge runtime · sump: sump pump · sleep: CPAP runtime.