What we found
The supplied EAGLE-I records sum to 1.89B customer-hours in 2024 and 0.89B in 2025. That is a 52.8% decline. Each snapshot is assigned to its own UTC calendar year; no multi-year event is moved into its starting year.
Why this is surprising
An eleven-year upward-looking line can obscure a large reversal in the latest year. A claim that outages are accelerating needs to explain this reversal, rather than stopping at the 2024 maximum. The largest declines are geographically concentrated; a change in aggregate burden is not the same as a uniform improvement in every county.
Where the decline came from
| State / jurisdiction | 2024 | 2025 | Contribution to net decline |
|---|---|---|---|
| Texas | 349.59M | 63.38M | 28.7% |
| Florida | 275.11M | 22.96M | 25.2% |
| South Carolina | 141.93M | 8.53M | 13.4% |
| Georgia | 132.42M | 23.75M | 10.9% |
| North Carolina | 120.46M | 17.80M | 10.3% |
| Puerto Rico | 121.78M | 78.76M | 4.3% |
| Ohio | 50.76M | 22.51M | 2.8% |
| Maine | 31.36M | 5.17M | 2.6% |
| Louisiana | 38.30M | 14.91M | 2.3% |
| Washington | 42.06M | 24.22M | 1.8% |
Does the reversal survive alternative comparisons?
| Comparison | 2024 customer-hours | 2025 customer-hours | Change |
|---|---|---|---|
| All positive snapshots | 1,891.68M | 892.74M | -52.8% |
| Omit values above modeled customers | 1,664.91M | 859.60M | -48.4% |
| Omit values above 2× modeled customers | 1,801.79M | 878.85M | -51.2% |
| Counties with a positive record in every year | 1,606.70M | 687.29M | -57.2% |
The decline appears in all four comparisons, ranging from 48.4% to 57.2%. This range is a set of sensitivity results, not a statistical confidence interval. The denominator screens do not identify all reporting errors. A common-record comparison reduces changes in county membership but cannot control changing utilities or collection success within those counties.
Sources and reproducibility
WattSizer calculations from the local EAGLE-I annual CSVs. Read the full methods and quality review. Original sources: Figshare historical data, Brelsford et al., Scientific Data (2024; 2014–2022 descriptor), and ORNL’s 2025 release. The descriptor explains the collection system; it does not describe observations from 2023–2025. The source manifest contains local input SHA-256 hashes, file sizes and timestamp bounds.