Still coming
Grid operators publish when each proposed power plant expects to connect. Read against the operators' own history, many of those dates stopped being forecasts long ago.
5.8 years
is how much later than it first filed the median active project in California's queue now expects to come on line. 246 of 261 are more than two years late.
Before a power plant can connect to the grid, it joins its operator's interconnection queue and waits for studies. Operators publish those queues, including the date each project expects to begin operating. Three were read here: CAISO (most of California), NYISO (New York) and SPP (the central US, from the Dakotas to Texas).
Tools that track queues show that expected date as it stands today. Only CAISO also publishes the date each project gave when it first filed, and the gap between the two is the slip. Projects already built came on line a median of 2.8 years after the date they first filed. The ones still waiting are further behind.
Dates already in the past
Some active projects list an on-line date that has already gone by. That is either a late project or a date nobody updated, and the file cannot say which. Either way, it is not a forecast.
| Operator | Active projects with a date | Date already passed |
|---|---|---|
| CAISO | 348 | 38 |
| SPP | 624 | 41 |
CAISO still lists Montezuma II, a 78 MW wind and battery project in Solano County, as active. It was requested in 2007, first filed to come on line at the end of 2010, and lists an on-line date of January 2012. In the week between the two captures, CAISO's Irving Storage moved its on-line date from 30 November 2025, already past, to 1 August 2030.
Withdrawals that were not
When NYISO moved from studying requests one by one to studying them in clusters, many projects were withdrawn from the old list and entered again under new queue numbers. Counted as withdrawals, NYISO's 2022 and 2023 requests look almost entirely abandoned: of 296, one is still active and none has been built. But 70 of them came back under the same name with a later request date. Across the whole queue, 163 NYISO withdrawals were re-filings, and 8 at CAISO.
That is a lower bound. Only exact name matches are counted, so a project that was renamed on the way back still reads as abandoned.
Checked against an outside figure
Lawrence Berkeley National Laboratory's Queued Up: 2026 Edition puts CAISO's completion rate, the share of capacity requested from 2000 to 2020 that reached operation, at 8%. The same measure from CAISO's current file gives 9.0% (248 of 1,909 requests, 13.0% by count). Agreement within a point, from a different date and capacity definition, is the check these figures had not had. LBNL does not report slip against the filed date, stale on-line dates or re-filings, so those have nothing to compare with.
How this was done
Each operator's queue file was downloaded and read as published, with no third-party data: CAISO's public queue report and Cluster 15 list, NYISO's queue workbook and SPP's active generator interconnection list. Transmission projects, notes rows and SPP's operating plants were set aside and counted. Operators measure capacity differently, so figures are never added across operators.
Every figure above is printed by python tools/findings.py, and Montezuma II's record by python tools/track.py find montezuma, in the backfeed repository. It measures the dated snapshots with the same code as its report. That repository is private, because the operators' terms do not cover republishing their files, and the snapshots are copies of them. Anyone can rebuild the figures from the operators' own public files.
What this does not show: why any single project is late. SPP has a "cause of delay" column, filled in on 32 of 682 active rows. The other two have none. A date in the past says a project is late or its date is stale, and nothing more.