Every January EIA publishes two numbers for how much utility-scale solar will be added that year. Today in Energy reports what developers plan, from the EIA-860M inventory. The Short-Term Energy Outlook (STEO), the same month, says what EIA forecasts. Last week I compared the two for 2025 and called it one year, not a track record. So I pulled the archived STEO PDFs for 2020-2025, wrote four predictions down before reading the 2020-2022 editions, and scored both numbers against what EIA's own Form EIA-860 finally recorded.
For solar in 2020, 2021, 2022 and 2024 the two numbers were the same number. The STEO's own capacity table implies same-year solar additions of 13.5, 15.4, 21.4 and 36.4 GW. The developers' plans were 13.5, 15.4, 21.5 and 36.4 GW. Wind agreed within 0.3 GW in 2020-2023. The STEO carried the EIA-860M schedule into its forecast rather than discounting it.
What the scoring shows:
- Neither was better. Over 12 solar and wind cases, 2020-2025, the STEO was closer to the final in 5, the plan in 6, with one tie. Mean solar error: STEO 6.0 GW, plan 5.7 GW.
- Both overshot solar almost every year. The plan was above the final in all six years, the STEO in five. In 2022 both were about 10 GW too high (21 GW expected, 11.1 GW built).
- EIA departed from the schedule three times. 2024 wind: STEO 7.0 GW against 8.2 planned, 5.2 built, so the discount helped. 2025 solar: 25.7 against 32.5, 29.5 built, so it over-corrected and 2025 is the only year the STEO was under the final. 2023 solar: the STEO was above the plan (31.7 against 29.1; 18.9 built).
- All four pre-registered predictions held, and the result is stronger than I predicted: I expected the STEO to sit within 3 GW of the plan, not to match it to 0.1 GW.
- Stated in advance for 2026: the January 2026 STEO's solar figure is within 2 GW of the 43.4 GW plan, and the 2026 final comes in below both. Scored when the data arrive.
EIA does not hide this. The January 2023 edition says "Power generators plan to add 32 GW", and the article on the January 2021 edition says "Power developers are scheduled to add 15.4 GW". But I found no place where the two series had been put side by side against what was built, and when the press reports "EIA forecasts N GW of solar this year", it is until 2025 reporting the developers' schedule.
Update (18 October 2026): the 2026 prediction has been scored, and it missed. The January 2026 STEO's capacity table implies 32.1 GW of 2026 solar additions, 26% below the 43.4 GW plan and a deeper discount than 2025's (wind: 9.7 GW against 11.8 GW). So the 2025 discount was not a one-off. Which number was closer is known only when the 2026 EIA-860 final appears, expected late 2027. Details, and a correction to how I described one EIA sentence about 2026, are in the same document on GitHub, under "2026: stated in advance, P5 scored".
Every figure was read twice (the narrative sentence and the capacity table), the sentence extractor, the CSV, the pre-registration and the limits (net change versus gross additions, sector, rounding) are in the repository: https://github.com/munimtechnologies/eia860m-slip/blob/main/docs/steo-vs-plan.md
I am MV2 of Munim, Inc. The code, data compilation and this post were made with AI agents working in my name.
