Quick answer: A single interruption averages just over 2 hours in normal operation and 7.2 hours when a storm is counted. Across a whole year the average US customer lost 11 hours in 2024. Where you live moves that by a factor of 96.
How long do power outages last is two questions wearing one coat, and almost every answer online merges them. One is how long a single interruption lasts. The other is how many hours you lose across a year. They are different numbers, the federal data reports them separately, and mixing them is why you will see everything from “one to two hours” to “six hours” quoted as the same fact. Our power outage checklist uses the annual figure to set expectations for the first hour; this page takes the numbers apart.
Three indices, three different questions
Utilities report reliability to the EIA on Form EIA-861 using three standard indices. Each answers something different, and the one that matches the question in the title is the one nobody quotes.
| Index | What it answers | 2024, normal operation | 2024, with major events |
|---|---|---|---|
| CAIDI | How long one interruption lasts | 123.6 min | 432.7 min |
| SAIFI | How many interruptions you get in a year | 1.065 | 1.531 |
| SAIDI | How many minutes you lose across the year | 131.6 min | 662.6 min |
CAIDI is the answer to the question. In normal operation a single interruption runs about 123.6 minutes, so a shade over two hours. Once major event days are counted the same figure becomes 432.7 minutes, or 7.2 hours, because those days are all storm restorations.
SAIDI is the number that gets quoted as if it were an outage length, and it is not: 662.6 minutes is the whole year added up. With SAIFI at 1.531 interruptions a year, the typical household experience is roughly one or two cuts, not a continuous eleven-hour event.
What a major event day actually is
It is not a label a utility applies when it feels overwhelmed. Under IEEE Standard 1366 a major event day is a day where the daily SAIDI exceeds a statistical threshold calculated from that utility’s own previous five years of data. The point of the separation is to let normal operation be studied without a single hurricane swamping the trend.
That matters here because it means the two columns above are not optimistic and pessimistic versions of one estimate. They are two genuinely different regimes, and which one you live in is mostly geography.
Where you live moves it by a factor of 96
| State | Hours lost in 2024, all events | Same year, normal operation only |
|---|---|---|
| South Carolina | 52.3 h | 2.0 h |
| Maine | 29.1 h | 4.6 h |
| North Carolina | 24.0 h | 2.4 h |
| Florida | 22.0 h | 1.1 h |
| Texas | 21.2 h | 2.2 h |
| West Virginia | 19.4 h | 8.1 h |
| US average | 10.2 h | 2.1 h |
| Arizona | 1.5 h | 1.2 h |
| South Dakota | 1.4 h | 1.0 h |
| Washington DC | 0.5 h | 0.4 h |
The finding that should change how you read this
Look at South Carolina in both columns. It is the worst state in the country for total hours lost, at 52.3. Strip the major event days out and it drops to 2.0 hours — better than the national average. Florida is more extreme still: fourth worst on total hours, and at 1.1 hours one of the best-run grids in the country the rest of the time.
Hurricane Helene is most of South Carolina’s 2024 figure. Outages there peaked around 386,000 customers, and across much of the western half of the state restoration took a week or more, while in Florida more than half of those affected were back within 24 hours. Same storm, same month, very different restoration curve.
West Virginia is the opposite case and the one worth studying if you live there. Its 19.4 hours is not dominated by storms: 8.1 of them happen in normal operation, the worst ordinary-year figure in the country, roughly four times the national average. That is a different problem with a different answer, because it is not something one bad week creates.
So the useful question is not how long outages last on average. It is which of those two profiles describes your address: a grid that is fine until a storm arrives, or one that drips outages all year.
What each profile should actually buy
| Your profile | What the data says | Where to start |
|---|---|---|
| Storm belt, quiet in between | Rare but very long. A week is possible | Fuel and refills, not just capacity: generator sizing |
| Frequent short cuts | One or two a year, about two hours each | A battery covers this without noise or fuel: power station sizing |
| Rural, high ordinary-year hours | Both patterns at once | Size for the loads that cannot wait, starting with the refrigerator |
The two-hour interruption is the one almost everyone actually meets, and it is comfortably inside the margin a closed chest freezer gives you for free. It is the multi-day tail that costs money to cover, and the cost there is mostly fuel per hour or a way to refill a battery, not the machine itself.
How we read this
Every figure here comes from the EIA’s Electric Power Annual 2024, which compiles what utilities themselves report on Form EIA-861. The national table uses the IEEE 1366 method; the state table uses the any-method series, which is why the US totals differ slightly between the two (662.6 against 611.3 minutes). We have not mixed them inside a single comparison, and converting minutes to hours is the only arithmetic of ours on this page.
Two limits worth stating. These are averages across every customer in a state, so a rural feeder and a downtown network are in the same number. And 2024 was an unusually bad year nationally, driven by Beryl, Helene and Milton, so treat it as a bad-year benchmark rather than a typical one.
Frequently asked questions
How long do power outages last on average?
About two hours in normal operation, from the CAIDI index. Counting storm days the average rises to 7.2 hours. Both are averages of very different things, which is why a single number never fits.
How many outages will I get in a year?
Roughly one to one and a half. SAIFI was 1.065 in normal operation and 1.531 including major events, so most households meet one or two cuts a year, not a constant drip.
Which state has the worst power reliability?
On 2024 totals, South Carolina at 52.3 hours. But that is a hurricane, not a chronic fault: in normal operation it is better than the national average. For chronic problems the honest answer is West Virginia, at 8.1 hours of ordinary-year outage.
Why do sites quote six hours, or one to two?
Because they are quoting different indices without saying so, and often older years. Six hours is roughly a past year’s national SAIDI; one to two hours is close to CAIDI. Neither is wrong on its own, but they answer different questions.
Is my utility allowed to exclude bad weather?
It reports both, which is why both columns exist. The exclusion follows IEEE 1366: a major event day is one where daily SAIDI passes a threshold derived from that utility’s own five previous years, not a judgement call made after the storm.
If you take one thing from the numbers, make it this: how long do power outages last depends far more on your state than on any national average. Plan for the profile, not for the average. A two-hour cut once or twice a year needs nothing you do not already own. A week without power after a storm is a different purchase entirely, and the federal data tells you, state by state, which of the two you should be planning for.
