# How Long Do Power Outages Last? EIA's 2025 Numbers

> EIA's 2025 data: the average U.S. customer lost 4.5 hours of power. State numbers for Missouri, Michigan and more, plus a 3-level plan for four people.

Canonical: https://www.backyardfreedomlab.com/blog/what-power-outage-data-actually-says-about-how-long-youll-be/
Published: 2026-11-05

---

EIA published its 2025 power reliability data on Oct. 21, and Missouri is in the middle of Winter Weather Preparedness Week (Nov. 2 to 6), with Indiana and Michigan starting theirs on Nov. 8 and Illinois on Nov. 9. The federal table gives you the actual yearly average, a state-by-state breakdown going back to 2015, and a clean split between ordinary days and storm days. That split decides how many hours you should plan for.

> **The short answer:** The average U.S. customer lost 269.2 minutes of power (about 4.5 hours) in 2025, spread over 1.34 interruptions of roughly 201 minutes each. Remove major event days and the figure falls to 125.3 minutes, about 2.1 hours. The year before, storms pushed the all-events number to 662.6 minutes, about 11 hours ([EIA Electric Power Annual, Table 11.1](https://www.eia.gov/electricity/annual/html/epa_11_01.html), release date Oct. 21, 2026, read Nov. 3, 2026). Your own state's worst year in the table is a better planning number than the national average.

Disclosure: some links on this page are affiliate links, so we may earn a commission if you buy through them, at no extra cost to you. This is a research-based guide. We read the federal, state and manufacturer pages listed under Sources on Nov. 3, 2026, and we have not bought or tested any product on this page.

## What EIA Actually Measures

EIA collects reliability data from utilities on Form EIA-861. Three terms matter, defined under Table 11.1:

- **SAIDI:** the minutes of non-momentary interruption the average customer had in a year.
- **SAIFI:** how many times the average customer lost power in a year.
- **CAIDI:** the average minutes to restore each interruption (SAIDI divided by SAIFI).

For 2025 the national numbers on the IEEE basis, which is the first column set in the table, were a SAIDI of 269.2 minutes, a SAIFI of 1.34 and a CAIDI of 201 minutes. That reads as: the average customer lost power a little more than once, for a little over three hours each time. The average counts every customer, including the many who lost nothing, so it says little about a single house at the end of a long rural line.

Each measure also comes in two versions. "All events" includes major event days, when a storm pushes a utility past a daily SAIDI threshold calculated from the past five years. "Without major event days" strips them out. Utilities that do not use the IEEE method decide for themselves what counts.

## The National Numbers, 2015 to 2025

| Year | All events | Without major event days |
|---|---|---|
| 2025 | 4.5 h (269.2 min) | 2.1 h (125.3 min) |
| 2024 | 11.0 h (662.6 min) | 2.2 h (131.6 min) |
| 2023 | 6.1 h (366.6 min) | 2.1 h (123.9 min) |
| 2022 | 5.5 h (333 min) | 2.2 h (131.1 min) |
| 2021 | 7.9 h (475.8 min) | 2.1 h (125.7 min) |
| 2020 | 7.6 h (456.1 min) | 1.9 h (116 min) |
| 2019 | 4.9 h (295.5 min) | 2.0 h (122.2 min) |
| 2018 | 5.8 h (349.2 min) | 2.0 h (121.4 min) |
| 2017 | 8.4 h (505.9 min) | 1.9 h (117 min) |
| 2016 | 4.5 h (268.4 min) | 2.0 h (119.8 min) |
| 2015 | 3.5 h (209 min) | 1.9 h (117 min) |

Source: EIA Table 11.1, IEEE columns. Hours are our conversion of EIA's minutes.

Our arithmetic on those eleven years: the average all-events figure is about 381 minutes (6.4 hours), and the average without major event days is about 123 minutes (2.0 hours). The second column barely moves. The first swings from 3.5 hours to 11. Almost everything that varies is storms.

EIA's write-up of the 2024 figure says U.S. customers averaged 11 hours of interruptions in 2024, nearly twice the annual average of the decade before, and that Hurricanes Beryl, Helene and Milton accounted for 80% of the hours without electricity ([EIA Today in Energy](https://www.eia.gov/todayinenergy/detail.php?id=66744), page read Nov. 3, 2026). 2025 had no storm year like that, which is why the new number looks calm. It does not mean the risk went away.

## State by State: One Year Tells You Little

EIA's Table 11.4 gives the same measures for each state. We pulled the 2025 figure, the 2015 to 2025 simple average and the worst single year for a mix of states, including the four whose winter-preparedness weeks run this month. The average is our arithmetic, an unweighted mean of eleven annual values.

| State | 2025, all events | 2025, without major event days | 2015 to 2025 average, all events | Worst year in the series |
|---|---|---|---|---|
| **Missouri** | 7.8 h | 1.9 h | 4.1 h | 7.8 h (2025) |
| **Indiana** | 6.3 h | 2.0 h | 5.1 h | 7.8 h (2023) |
| **Michigan** | 10.1 h | 2.7 h | 9.9 h | 18.8 h (2023) |
| **Illinois** | 2.4 h | 1.0 h | 2.7 h | 5.6 h (2020) |
| **Ohio** | 3.4 h | 2.2 h | 4.8 h | 8.5 h (2024) |
| **Pennsylvania** | 6.9 h | 2.3 h | 4.7 h | 9.4 h (2018) |
| **West Virginia** | 16.9 h | 8.3 h | 14.2 h | 19.4 h (2024) |
| **North Carolina** | 3.2 h | 2.4 h | 10.1 h | 31.3 h (2018) |
| **Florida** | 1.2 h | 1.1 h | 10.0 h | 44.9 h (2017) |
| **Texas** | 3.5 h | 2.5 h | 9.0 h | 26.9 h (2024) |
| **Louisiana** | 4.5 h | 3.8 h | 21.5 h | 97.8 h (2021) |
| **Maine** | 6.1 h | 3.2 h | 17.9 h | 41.5 h (2017) |
| **South Dakota** | 4.7 h | 1.1 h | 2.9 h | 5.6 h (2022) |
| **Arizona** | 1.1 h | 1.0 h | 1.6 h | 2.3 h (2022) |

Source: EIA Table 11.4, IEEE columns, read Nov. 3, 2026. Hawaii is left out because the IEEE series has gaps there.

Three things stand out.

**The ranking changes year to year.** Florida's 2025 figure was 1.2 hours, the second-lowest in the list. Its 2015 to 2025 average is 10.0 hours because one year, 2017, reached 44.9. North Carolina went from 31.3 hours in 2018 to 3.2 in 2025. A state that looks quiet in a single year has usually just skipped a storm.

**Some states sit high even without storms.** West Virginia's "without major event days" figure was 8.3 hours in 2025, four times the national 2.1. That points at everyday grid conditions rather than a once-a-decade event. If you live in a state like that, the ordinary number is the one that matters.

**Missouri set its own eleven-year high in 2025.** Its 7.8-hour all-events figure is the highest in the series, and the same year's number without major event days was 1.9 hours. The state's own preparedness release (SEMA, Oct. 27, 2026) says plans should include several days' worth of supplies. The data fits that advice: the long hours come from a few big days, and nobody knows which days.

## Why the Long Outages Are Mostly Weather

The EIA tables do not break outages down by cause. Climate Central did, using the Department of Energy's OE-417 reports of major outages, which it defines as events that affect at least 50,000 customers or 300 megawatts. Of the major outages reported from 2000 to 2023, 80% (1,755) were weather-related. Within that group, 58% came from severe weather, 23% from winter storms and 14% from tropical cyclones including hurricanes. Climate Central also found about twice as many weather-related major outages in 2014 to 2023 as in 2000 to 2009, and the five states with the most were Texas (210), Michigan (157), California (145), North Carolina (111) and Ohio (88) ([Climate Central](https://www.climatecentral.org/climate-matters/weather-related-power-outages-rising), Apr. 24, 2024).

Two limits apply. The 50,000-customer threshold leaves out small outages, and those counts stop at 2023, so they miss 2024 and 2025.

## What This Winter's Outlook Does and Does Not Say

NOAA's U.S. Winter Outlook (Oct. 15, 2026) expects a strong El Nino to shape December through February. It favors warmer-than-average temperatures across the northern U.S. and the Midwest, and wetter-than-average conditions across the southern tier and along the East Coast. NWS director Ken Graham said a very strong El Nino "tilts the odds toward stormier weather across the South," and that too much precipitation on dry soil can fuel flash floods and mudslides (NOAA, Oct. 15, 2026).

That is a seasonal tilt, not a forecast of outages. The outlook does not say where an ice storm or a windstorm will land, and we found no official source that predicts outage counts for this winter. Missouri's release warns of snow, sleet and ice and lists loss of power and heat among the hazards. Use the outlook to decide which kind of weather to prepare for, and the EIA table to decide how long.

## What This Changes for Your Household

The official advice stays the same. Ready.gov's kit page (updated July 1, 2026) still lists water at one gallon per person per day for several days, at least a several-day supply of non-perishable food, a flashlight, a battery-powered or hand-crank radio and a NOAA Weather Radio with tone alert, and a phone with chargers and a backup battery. What the data adds is a way to choose a number of hours to plan around.

Two fixed points from the government pages:

- **A closed refrigerator keeps food cold for about four hours, and a full freezer for about 48 hours.** Ready.gov (June 4, 2026) says to throw out food at 40°F or higher. FoodSafety.gov (Aug. 8, 2024) puts a half-full freezer at about 24 hours and says to discard perishable food that sat at 40°F or above for more than two hours.
- **The national average customer lost about 4.5 hours in 2025.** That is longer than the four-hour refrigerator window, so even a calm year can run past it for some households.

Then look up your state's worst year in the table above and read it as your planning number, not a prediction. A Missouri household plans for about eight hours. A Michigan household plans for about 19. A Louisiana household that wants to cover its worst year in the series is planning for roughly four days, and that case is a reason to talk about generators and fuel, not a typical need. Averages hide the outages above and below them, so treat the number as a floor.

## A Three-Level Plan for a Household of Four

| Level | What to do | Quantity for 4 people |
|---|---|---|
| **Tonight (covers about 4 hours)** | Keep the refrigerator and freezer closed. Put a thermometer in each. Set one flashlight where each person can find it. Charge phones and a backup battery. Find the radio. | 4 flashlights, 1 radio, 1 backup battery |
| **This month (covers 48 hours and more)** | Keep the freezer full, since a full one holds about twice as long as a half-full one. Store water at one gallon per person per day. Put a carbon monoxide alarm with battery backup on every level. Write down your state's worst-year hours from the table. Save your utility's outage phone number and map link. | 4 gallons per day, so 12 gallons for 3 days, plus coolers and ice |
| **Optional (covers your worst-year hours)** | List what must keep running, add up watts times hours, and size a battery or a generator to that list. Medical devices come first. | Your own list: see the sizing section below |

Our [Blackout Readiness Score](/tools/blackout-readiness/) takes six questions about power, water, food and light and returns your three biggest gaps. Our guide to building a [blackout kit room by room](/blog/build-blackout-kit-room-by-room/) turns each row into a shopping and storage list.

<NewsletterForm source="article:what-power-outage-data-actually-says-about-how-long-youll-be" />

## Sizing Backup Power From Your Hours

If you do want backup power, the data gives you the duration and the arithmetic gives you the size. The formula is watts times hours equals watt-hours, plus some margin because the inverter wastes part of the energy.

Example loads are illustrative, so replace them with the wattage printed on your own devices. A router and a laptop that together draw 40 watts, running 8 hours, need 40 × 8 = 320 watt-hours. A 280-watt-hour battery falls short of that on paper. A 1,152-watt-hour battery covers it with room to spare (1,152 ÷ 40 = 28.8 hours before conversion losses). A refrigerator, a sump pump or a medical device changes the math, and the label on the device is the only number worth using.

| Option | Covers | Does not cover | Cost note |
|---|---|---|---|
| **Phone power bank** | Phones and a radio charger for a day or more. Ready.gov lists a backup battery in the kit. | Anything that plugs into a wall | Lowest cost of the four |
| **300W-class power station (about 280 Wh)** | Phones, lights, a router, a laptop for part of a night | A refrigerator, a heater, a well pump | Price not checked; see below |
| **1,000 Wh-class power station** | The same loads for a day or more, and a small appliance for hours | An electric furnace, a heat pump, a whole house | Price not checked; see below |
| **Portable gas generator** | Larger loads, if you store fuel and run it correctly | Indoor use, ever | Fuel and maintenance cost; keep it outdoors and away from windows, doors and attached garages (Ready.gov) |

Our guide to [300W versus 1,000W portable solar generators](/blog/300w-vs-1000w-portable-solar-generator-how-much-power-does/) walks through the runtime math for each class, and the [generator placement and clearance rules](/blog/carbon-monoxide-and-generator-placement-the-clearance-rules/) cover the one hard safety line for anything that burns fuel.

### Two stations to compare

<ProductCard slug="amz-300w-portable-solar-generator" name="300W Portable Solar Generator" ctaLabel="Check Price on Amazon" note="Sized for phones, lights, a router and a laptop. It does not run a refrigerator, a heater or a well pump. Sources disagree on the bundled solar panel size, so read the listing you receive.">ZeroKor's own page lists a 300W station with a 280 Wh lithium battery and a foldable 60W solar panel. Our earlier guide on this class, citing the Amazon listing, says 40W for the panel, so the two sources disagree. At 280 Wh, the 40-watt example above runs about 7 hours on paper (280 ÷ 40), less after conversion losses. It fits a household whose state shows 2 to 5 hours in a typical year and wants quiet, fume-free light and phone power. It does not fit anyone who needs a refrigerator or medical equipment kept running. We could not confirm today's price because Amazon blocks our checks.</ProductCard>

<ProductCard slug="amz-bluetti-ac180-portable-power" name="Bluetti AC180 (1,152 Wh, 1,800W)" ctaLabel="Check Price on Amazon" note="A larger unit for a household whose worst-year hours run to a day or more. It is not a whole-house system, and it is not a substitute for a medical-equipment plan.">Bluetti's product page lists 1,800W continuous output, a 1,152 Wh battery, a charge from 0 to 80% in 45 minutes with up to 1,440W input, and a 2,700W boost mode for some loads. That gives the 40-watt example about 29 hours on paper. It fits a household in a state like Michigan, West Virginia or Louisiana, where the average year runs 10 hours or more, and who wants to keep small loads and possibly a refrigerator going. It does not fit a household that only needs a few hours of light, where a power bank costs far less. We could not confirm today's price on Amazon, and we found no sourced buyer-complaint data we trusted enough to cite.</ProductCard>

## What We Could Not Confirm

- **Month-by-month outage data.** EIA's tables are annual. Nothing we read splits outage hours by November versus January, so we do not claim November is higher or lower risk.
- **Any official forecast of outages this winter.** NOAA's outlook covers temperature and precipitation. NERC's winter reliability assessment is reported by a secondary source as due Nov. 18, 2026, and we have not seen it on NERC's site.
- **Current prices.** Amazon blocks our price checks, so no price appears for either power station. We read only the manufacturers' product pages.
- **Customer complaints about either station.** We did not find review data we could cite with a date, so this page makes no claim about reliability or support.

We will update this page when NERC publishes its winter assessment, when NOAA issues its next outlook update (a secondary source says Nov. 19, which we have not confirmed on NOAA's site), and when EIA adds anything newer to the 2025 reliability tables.

## The Takeaway

The typical household lost about 4.5 hours of power in 2025 and about 2 hours in an ordinary year, but one bad storm season can double the national figure. Look up your state in the table, pick its worst year as your planning number, and build the three levels in order: closed doors and flashlights tonight, 12 gallons of water and a full freezer this month, and a battery sized from watts times hours only if your hours call for one.

The free outage power priorities worksheet walks through the must-run list and the math.
