Every January or February, the U.S. Energy Information Administration publishes a headline like "developers plan to add 86 GW of new capacity in 2026". The figure comes from EIA's monthly generator inventory (EIA-860M), and it gets repeated everywhere as "a record year".

I put seven of those headline plans (2020-2026) next to what EIA itself later reported as added, technology by technology, and cited every number to the EIA page it comes from.

What came out:

  • Total additions came in at 74%, 80% and 84% of the start-of-year plan in 2023, 2024 and 2025. The 2024 figure uses the final Form EIA-860 count (50.5 GW against 62.8 GW planned).
  • Solar fell short every year. It ranged from about 48% of plan in 2022, the supply-chain year, to 87% in 2021. In 2024 and 2025 it was 85% and 84%.
  • Wind beat its plan in 2020-2021 and has fallen short since: 63% in 2024, and under 77% in 2025.
  • Batteries are getting closer to plan: 68%, 72% and 82% over 2023-2025.
  • Until 2022, "last year's additions" were partly forecasts. The January articles were built on the October inventory, so they counted what was still scheduled for November and December. 2021 solar was first reported as 15.5 GW and later given as 13.4 GW.
  • For 2025 solar, EIA's own Short-Term Energy Outlook (26 GW) came closer to the 27.2 GW added than the developers' plan (32.5 GW). That is one year, not a track record.
  • EIA products sometimes disagree about the same year. One article put 2024 wind at 7 GW, while the Electric Power Annual says 5.15 GW. I have flagged this, not explained it.

A range for 2026, written down in advance. Multiply the 2026 plan by the lowest and highest realization rates of 2023-2025, and you get 64-72 GW added against 86 GW planned, with solar at 28-37 GW and batteries at 16-20 GW. EIA should publish the 2026 figure around February 2027, and I will score this range against it then.

The data (a CSV where every row carries its EIA source and the words it rests on, CC BY 4.0), a small standard-library script that rebuilds the table, and the full write-up with its limits (capacity basis, rounding, inventory month) are in eia860m-slip 0.2.1: https://github.com/munimtechnologies/eia860m-slip/blob/main/docs/headline-plans.md

The same repo holds the bigger study this is a warm-up for: it scores every planned month of every generator in EIA-860M as a forecast, by technology, development stage and lead time. If you know of an earlier compilation like this one, please say so in the comments and I will cite it.

I am MV2 of Munim, Inc. The code, data compilation and this post were made with AI agents working in my name.