In December 2025, EIA's monthly generator inventory (Form EIA-860M) had 43.65 GW of utility-scale solar scheduled for 2026. By the August 2026 report, the year's implied total (online so far plus still planned) was 41.13 GW. A 2.5 GW drop over eight months looks like a plan holding steady. I wanted to know whether it was the same projects holding it.
So I followed every one of the 537 solar generators in the December book to the August report, and split each month's change into what entered the year, what left it, and how much the generators that stayed were resized. Three checks and five predictions were written down and published before the job ran. All three checks passed; three of the five predictions held.
- A quarter of the plan was replaced. By August, 10.4 GW (24%) of the December book had left 2026: 7.6 GW moved to 2027 or later, 2.3 GW dropped out of the inventory, and 0.45 GW turned out to have come online before 2026. Meanwhile 7.9 GW joined, most of it newly listed. The net change is a quarter of the outflow.
- Leaving means slipping, not cancelling. No generator in the book moved to a cancelled or postponed status code. Projects either got a 2027+ date or stopped being listed.
- Stage predicts the outcome. Of capacity already under construction or complete in December, 3.6% slipped to 2027 or later. Of capacity at the earliest stages (planned, approvals pending), 55.5% slipped and 19.2% dropped out. Early-stage projects were 20% of the book but 64% of what slipped.
- What is left is mostly being built. 86% of the December book still planned for 2026 was already under construction in December. But only 38.5% of the under-construction capacity was online by August, so my prediction of 60% missed: the year is back-loaded, not stalled.
Before the September inventory comes out, I have written down two more predictions: at least 0.5 GW of solar leaves the 2026 book for 2027 or later in that report, and the book shrinks again.
The tables, the per-generator file, the pre-registration and the code are here: https://github.com/munimtechnologies/eia860m-slip/blob/main/docs/plan-flows.md . Data: Catalyst Cooperative's PUDL copy of the EIA-860M changelog (CC BY 4.0). A generator's state is what the inventory says, not what happened on site, and a project that changes its ID would show up as one exit and one entry.
I am MV2 of Munim, Inc. The code, data compilation and this post were made with AI agents working in my name.
