Are Power Outages Getting Worse in the U.S.?

The clearest measure of the change is SAIDI, the System Average Interruption Duration Index. It measures how many minutes of sustained power interruptions the average customer experiences over a year.

Using the U.S. Energy Information Administration’s Any Method national reliability series, average annual SAIDI increased from 298.5 minutes during 2014–2018 to 444.1 minutes during 2020–2024.

That is a 48.8% increase.

Individual years are volatile because hurricanes, winter storms and other major disruptions can push one year’s outage total far above another. Five-year averages help smooth some of that volatility while still showing a substantial deterioration.

EIA publishes outage duration both with and without major events. Subtracting the latter from total SAIDI gives a useful analytical estimate of the outage time associated with the major-event component.

One distinction is essential: major event does not mean weather event. For utilities using IEEE methodology, a Major Event Day is determined through a statistical threshold based on unusually high daily outage duration. Utilities using other reporting methods may determine major events differently.

Many of the country’s worst outage years clearly coincide with hurricanes, severe storms, extreme cold and other weather hazards. But the entire calculated major-event component should not be labeled as weather-related.

If the everyday electric grid were simply failing more often, we would expect the reliability measures excluding major events to rise sharply as well.

They haven’t.

Average routine outage frequency was 1.032 sustained interruptions per customer per year in 2014–2018 and 1.034 in 2020–2024. That is effectively no change.

Routine outage duration moved somewhat more, from 114.3 to 122.1 minutes per year, but the increase was still only about 7%.

The interruptions associated with bad years are also lasting longer

CAIDI measures the average duration of a sustained interruption. Across all events, average interruption duration increased from about 228 minutes to 310 minutes, a rise of roughly 36%.

When major events are excluded, CAIDI increased by only about 7%, from roughly 111 to 118 minutes.

The growing reliability problem is therefore much more about the amount of time severe disruptions can leave customers without service than an explosion in ordinary outage frequency.

The most recent finalized year shows how large the major-event component can become.

EIA’s broader public summary described 2024 as about 11 hours of electricity interruptions per customer, with major events accounting for about 80% of outage hours. Hurricanes Beryl, Helene and Milton were among the year’s largest contributors.

Hurricane Helene caused outages affecting roughly 5.9 million customers across 10 states. South Carolina customers averaged nearly 53 hours of interruption during the year.

Those figures make 2024 useful for understanding the scale of a severe modern outage year, but it should not be used by itself to prove the broader trend. Our five-year comparison is stronger because the deterioration remains after individual disaster years are averaged together.

To look past the national average, we calculated each state’s average annual SAIDI during 2014–2018 and compared it with 2020–2024. We then repeated the calculation using outage duration excluding major events.

The largest increases were striking.

These results should not be treated as a permanent ranking of which states have the least reliable grids. One hurricane, winter storm or other rare event can heavily influence a five-year average.

The table answers a narrower question: where did average outage duration change most between the two periods? That differs from our Backup Power Need Index, which looks at where household backup pressure is highest rather than where reliability changed most.

Louisiana is in a category of its own

Louisiana’s five-year average increased from 5.77 hours to 33.50 hours per customer per year, an increase of almost 28 hours.

Routine outage duration increased by only about half an hour. By our calculation, almost 98% of Louisiana’s additional outage time between the periods was associated with the major-event component. The recent period includes extraordinary hurricane impacts, particularly in 2020 and 2021.

Connecticut’s routine reliability actually improved

Connecticut’s total outage duration increased from 4.37 to 10.88 hours per year, yet its routine outage duration declined from about 79 to 69 minutes.

Its recent deterioration therefore did not come from a steadily worsening ordinary baseline. Major-event years added far more downtime on top of routine performance that had actually improved.

California shows a different pattern

California’s total outage duration increased from 2.58 to 4.72 hours per year, while routine outage duration also rose from about 98 to 142 minutes.

That is roughly a 45% increase in the non-major-event measure. Major events still explain most of California’s total deterioration, but the routine baseline worsened materially as well.

Some states improved

Delaware, Florida, Idaho, Maryland, South Dakota and Wyoming had lower recent five-year averages for both total and routine outage duration.

Florida is a useful reminder of how strongly event timing can shape these comparisons. Its total average fell from roughly 11.2 to 9.7 hours even though hurricanes remain a major risk. The earlier period included the extraordinary 2017 Hurricane Irma outage year.

The strong role of major events naturally raises another question: how much of the worsening outage trend is caused by weather?

There is strong evidence that extreme weather contributes to U.S. outage risk. But our EIA analysis cannot tell us that every additional major-event outage minute was caused by weather or climate change.

The Fifth National Climate Assessment provides useful independent evidence using a different definition of a major outage. It examines large outages affecting more than 50,000 customers and reports that the average number was roughly 64% higher during 2011–2021 than during 2000–2010.

Among weather-related outages in that dataset, severe weather accounted for 58%, extreme cold for 22% and tropical cyclones for 15%.

Those statistics should not be mathematically combined with our EIA Major Event analysis because the datasets use different definitions. They are valuable as independent evidence that large weather-related disruptions have become more common over a longer historical period.

Whether a weather hazard becomes a long outage also depends on trees and vegetation, overhead versus underground lines, utility maintenance, grid design, development patterns, restoration resources and infrastructure hardening.

The outage data could give the impression that the electric system is simply deteriorating while utilities stand still. The investment data shows something more complicated.

Berkeley Lab’s 2026 distribution-cost study found that most of the recent increase has come from capital spending.

Utilities are replacing equipment and investing in reliability, safety and resilience. Depending on the system, that can include stronger distribution equipment, vegetation management, automation, undergrounding, substation work and other hardening measures.

The same decade that produced more severe outage years also produced a sharp increase in distribution-grid investment.

That is why the historical trend should not simply be projected forward indefinitely. Future reliability will partly depend on whether infrastructure upgrades can keep pace with the hazards and loads stressing the system.

“The grid” also contains two different reliability problems

Most of the EIA data in this analysis describes distribution reliability: whether electricity can physically reach a home after local lines, poles, transformers or substations are damaged.

A different issue is bulk-system adequacy: whether the larger generation and transmission system has enough capacity to meet demand.

NERC’s 2025 Long-Term Reliability Assessment forecasts more than 224 GW of additional summer peak demand and roughly 246 GW of additional winter peak demand across its assessment areas over the coming decade.

That growth creates a separate future resilience challenge. It should not be interpreted as the cause of the historical SAIDI increase in this article.

A tree bringing down a neighborhood feeder and a regional shortage of generation capacity are fundamentally different failure modes. Both can leave a home without electricity.

If Americans were simply experiencing many more short, ordinary outages, the changing need could often be addressed with relatively modest battery capacity.

That is not where most of the national deterioration appears. The much larger increase is in major-event outage duration.

For longer disruptions, stored battery capacity is only part of the equation. Households also need to consider which loads must remain powered, how much energy those loads consume, whether storage can be expanded and how the system can be replenished if the outage lasts for days instead of hours.

• Critical-load prioritization
• Usable battery capacity
• Expandable storage
• Solar replenishment
• Generator integration
• Household electrical integration

That does not mean every household needs a whole-home battery. It does mean the part of the reliability problem that has changed most is the part that makes runtime, replenishment and energy management increasingly important.

For a broader look at where outage exposure and household energy factors create the greatest need, see our Backup Power Need Index. Our Changing Economics of Backup Power examines the cost side, while State of Home Backup Power in America 2026 connects outages with generators, solar, batteries and the changing backup-power market.