Power Outage Risk by State: SAIDI, SAIFI and Minutes Without Power

Two numbers govern how much backup power a house actually needs, and most people shopping for a generator have never seen either. SAIDI counts the minutes a year the average customer in a state spends without power. SAIFI counts how many separate times the lights go out. Every investor-owned utility, cooperative and municipal system files both with the federal government, state by state, and the gap between the worst state and the best runs 96 to one.

This page is the whole filing, ranked. It uses the final 2024 data year from Form EIA-861, released by the Energy Information Administration on 3 December 2025 — the most recent complete year covering all 50 states and the District of Columbia.

What SAIDI and SAIFI Actually Measure

SAIDI is the System Average Interruption Duration Index: total customer-minutes of interruption in a year, divided by the number of customers served. If every customer on a system lost power once for an hour, SAIDI would be 60. It is an average, so a state where a tenth of customers lost power for ten hours produces the same figure.

SAIFI is the System Average Interruption Frequency Index: the number of sustained interruptions per customer per year. The Form EIA-861 instructions set the threshold at longer than five minutes, so the flicker that resets your clocks is excluded.

Divide one by the other and you get CAIDI, the average length of a single interruption. That is the number that decides whether a portable power station covers you or whether you need a generator, because it describes the typical event rather than the annual total.

The second pair of columns is the one worth studying. Utilities that identify Major Event Days do it under IEEE Standard 1366, in the 2003 or 2012 edition, and file the year both ways. The all-events column describes the year you might have to live through. The excluding-major-events column describes the ordinary background reliability you live with every other week.

Every State Ranked, 2024

Sorted by total minutes without power, worst first. Any column heading re-sorts; a state name opens its page.

Customer-weighted from Form EIA-861, Reliability_2024, State Totals sheet. Figures use the broadest reporting set, covering every utility that filed under any reporting standard. Download the full table as CSV — it carries the same columns plus the narrower IEEE 1366 subset for every state.

The Same Data as a Map

Each square is one state or the District of Columbia, laid out in roughly its geographic position and shaded by hours without power in 2024. The number inside each square is that state’s figure.

The pattern is a hurricane season, drawn. South Carolina, North Carolina, Georgia, Florida and Texas form one band; Maine, New Hampshire and Vermont form another, driven by winter and spring storms in forested terrain. Arizona, South Dakota, North Dakota and the District of Columbia sit at the calm end, all under two hours for the year.

The grid places Hawaii and Alaska for legibility rather than geography, and every square hides the spread inside a state: a dense urban feeder and a rural cooperative line thirty miles up a valley sit in the same number.

Which Days the Minutes Fall On

Split each of the twelve worst states into the minutes that fell on ordinary days and the minutes that fell on major event days, and the shape of the risk changes completely.

Florida is the clearest case. It ranks fourth in the country for total outage minutes at 22.0 hours, and seventh best for ordinary-day reliability at 66 minutes a year. A Florida utility runs a tidy system for eleven and a half months and then loses a fortnight to a hurricane. Major event days carried 95% of its outage minutes that year.

West Virginia is the mirror image. Its 486 minutes of ordinary-day SAIDI is the highest baseline in the country — more than eight hours a year of routine outages, before a single storm is counted. Vermont, Maine and Mississippi follow. In those states the problem is not one bad week; it is steady attrition on long rural distribution lines through timber.

That distinction changes what to buy. Against a Florida risk profile, capacity matters: enough stored energy or fuel to carry days, used once every few years. Against a West Virginia risk profile, availability matters: something that comes on by itself several times a year without anyone thinking about it. The backup power calculator sizes both cases from the same load list.

The Events Behind the 2024 Numbers

Beryl, Helene and Milton account for 80% of the hours without electricity the country recorded in 2024, on the Energy Information Administration’s reading of the file (Today in Energy, 1 December 2025). The peak customer counts below are its figures; the dates are the storm lifespans in the National Hurricane Center’s final reports for the 2024 Atlantic season.

Storm	Dates (NHC report)	Hit hardest	Customers out at peak
Hurricane Beryl	28 June – 9 July 2024	Texas	2.6 million
Hurricane Helene	24–27 September 2024	Florida, Georgia, the Carolinas and inland Appalachia — 10 states	5.9 million
Hurricane Milton	5–10 October 2024	Central and west-coast Florida	3.4 million

Helene is the reason South Carolina tops the ranking at 52.3 hours, the highest state figure in the 2024 release. Major event days carried 96% of its outage minutes. Its ordinary-day SAIDI, 118 minutes, is better than the national average.

The northeastern states near the top of the table were hit by a different family of events: winter and early-spring storms that drop wet snow and ice onto trees. They draw far less national coverage than a hurricane, and they cost more. Maine finished 2024 second in the country at 29.1 hours, ahead of Florida at 22.0 and Texas at 21.2.

Outside the 50 states, Puerto Rico is in a category of its own. The Puerto Rico Electric Power Authority reported SAIDI of 4,415 minutes for 2024 — 73.6 hours — over 18.6 interruptions. Excluding major event days it still reported 1,811 minutes, about 30 hours. Across 2021 to 2024 the Energy Information Administration puts the island’s average at about 27 hours a year even with hurricanes taken out, against roughly two hours on the mainland.

Turning a Ranking Into a Battery Size

The annual total is the wrong number to size against. What sizes a system is the length of a single event, which is SAIDI divided by SAIFI.

Work South Carolina through it: 3,136 minutes over 2.37 interruptions is about 1,325 minutes — roughly 22 hours for a typical 2024 interruption. Run a refrigerator, a chest freezer, a well pump, a modem and a few lights at a rough 350 watt average and 22 hours costs about 7.7 kWh. That is beyond most portable power stations and inside the range of a mid-size home battery or a small inverter generator with fuel on hand.

Now the District of Columbia: 33 minutes over 0.27 interruptions is about 120 minutes per event. The same load list for two hours is 0.7 kWh — a single mid-size power station, comfortably.

Same load list, answers a factor of ten apart. Put your own appliance list and your own state’s duration into the backup power calculator instead of a national average. To see what stops working as the hours pass, the week-without-the-grid breakdown walks the failure order and the food safety clock puts a limit on the refrigerator.

Method, Coverage and Sources

All figures come from the State Totals sheet of Reliability_2024.xlsx, one of the files in the Form EIA-861 annual data release. The Energy Information Administration published the final 2024 file on 3 December 2025. Utilities file the underlying numbers themselves.

Two reporting sets exist in the file. The narrower covers utilities reporting under IEEE Standard 1366, representing 129.9 million customers; the broader adds utilities using other internal standards, representing 155.6 million. The tables here use the broader set: it covers 26 million more customers, and in some states the IEEE subset is a small minority — New York has 8.4 million customers in the broad set against 2.0 million in the IEEE subset. Both are in the CSV.

The choice moves the national headline slightly. On the IEEE subset the customer-weighted national average is 663 minutes, 11.0 hours, which is the figure the Energy Information Administration published in December 2025. On the broader set it is 611 minutes, 10.2 hours.

An early release of the 2025 data year appeared on 4 August 2026. It covers 90.0 million customers across 48 states — the District of Columbia, New Hampshire and New Jersey are absent — and the file itself carries a warning from the Energy Information Administration that the data is not fully edited and should not be aggregated to state or national totals. That rules it out for a ranking. This page stays on 2024 until the final 2025 file lands.

The causes behind the numbers are not in the EIA file. For the national cause split, Climate Central’s September 2022 analysis of major US outages found that about 83% of reported major outages between 2000 and 2021 were weather-related, of which 58% were severe weather, 22% winter weather, 15% tropical cyclones and about 5% heat and wildfire.

Verified September 2026. Figures checked against Form EIA-861 Reliability_2024.xlsx as published. Refresh: annual. The Energy Information Administration lists the final 2025 file for October 2026, and this page moves to it then. data.csv carries all 51 rows, both reporting sets and the national line. The underlying EIA data is a US government work in the public domain; this compilation is free to reuse with credit to Grid CEO.

Common Questions

Does a low SAIDI mean my own street is reliable?

No. SAIDI is a state-wide customer average, and the spread inside a state is usually wider than the spread between states. Treat it as the odds you are playing, and your own outage history over the last five years as the better estimate for your address.

Why do some states look far worse with major event days included?

Because one storm can outweigh the rest of the year. South Carolina, Florida and Texas all report ordinary-day figures at or better than the national average, then lose between 21 and 52 hours to one hurricane season. IEEE Standard 1366 separates the two so utilities can be compared on the days they control.

Which number should I size backup power against?

Neither total on its own. Divide SAIDI by SAIFI for the length of a typical interruption, then decide how far past that you want covering. Storm-driven states need a long tail on top; states with a high ordinary-day baseline need something that starts by itself often.

How current is this, and when does it change?

Data year 2024, from the final file published on 3 December 2025. The 2025 early release covers 48 states and the Energy Information Administration marks it as not for aggregation, so it cannot be ranked yet. This page refreshes once a year when the next final file appears.

Sources: U.S. Energy Information Administration, Form EIA-861 annual electric power industry report, Reliability_2024.xlsx, State Totals sheet (final data released 3 December 2025) and the 2025 early release (4 August 2026); EIA Today in Energy, “Hurricanes in 2024 led to the most hours without power in the United States in 10 years” (1 December 2025) and “Even without hurricanes, customers in Puerto Rico lose about 27 hours of power per year” (13 August 2025); Form EIA-861 instructions (2024), definitions of SAIDI, SAIFI and the sustained interruption threshold, and IEEE Standard 1366 for Major Event Days; National Hurricane Center tropical cyclone reports, 2024 Atlantic season, for storm dates; Climate Central, “Surging Weather-Related Power Outages” (14 September 2022).