Analysis · interconnection queue
What the interconnection queue actually tells us
The interconnection queue is a pipeline of proposals, not a forecast of plants. Most requested capacity never gets built, waits vary sharply by region, and the resource mix is shifting. Here is what the LBNL “Queued Up” dataset shows.
Most requested capacity is never built
Of the capacity requested in interconnection studies between 2000 and 2020 — old enough for the outcome to be settled — 13.0% reached commercial operation while 75.7% withdrew, and 11.3% is still active. These shares are of requested capacity (capacity-weighted, so each megawatt counts equally). This is why a queue total can never be read as a build forecast.
How long the queue takes, by region
| Region | Median request → operation | Median request → agreement | Completed projects |
|---|---|---|---|
| CAISO | 6.0 yrs | 3.7 yrs | 225 |
| NYISO | 4.6 yrs | 3.4 yrs | 83 |
| SPP | 4.1 yrs | — | 272 |
| Non-ISO Southeast | 3.8 yrs | 1.8 yrs | 214 |
| ERCOT | 3.7 yrs | 1.7 yrs | 470 |
| PJM Interconnection | 3.2 yrs | 3.0 yrs | 1,180 |
| MISO | 3.0 yrs | 2.5 yrs | 476 |
| Non-ISO West | 2.0 yrs | 1.5 yrs | 175 |
| ISO New England | — | 2.8 yrs | 0 |
| National | 3.6 yrs | 2.2 yrs | 3,095 |
Durations use requests that carry both a request date and an online (or agreement) date with a positive elapsed time.
Withdrawal rates by request cohort
Grouping requests by the five-year window in which they entered shows how much of each cohort has since been built, withdrawn, or is still active. The percentages below are the share of requests (count-weighted, so each project counts equally).
| Request cohort | Requests | Operational | Withdrawn | Active |
|---|---|---|---|---|
| 2000–2004 | 1,752 | 29.1% | 69.7% | 1.1% |
| 2005–2009 | 4,483 | 23.2% | 75.0% | 1.8% |
| 2010–2014 | 4,487 | 22.9% | 74.7% | 2.4% |
| 2015–2019 | 9,053 | 15.9% | 67.9% | 16.2% |
| 2020–2024 | 15,141 | 2.6% | 60.5% | 36.8% |
Count-weighted and capacity-weighted rates differ: because withdrawn projects skew larger than average, the capacity-weighted withdrawal rate for 2000–2020 (75.7% of requested capacity) runs higher than the count-weighted rates in the table.
The shift from renewables toward gas
In this dataset — active and suspended requests, primary components — requested capacity by resource type is:
| Resource type | Requested capacity | Share |
|---|---|---|
| Solar + Battery | 456.42 GW | 24.3% |
| Solar | 431.17 GW | 22.9% |
| Battery | 418.43 GW | 22.2% |
| Gas | 260.54 GW | 13.8% |
| Wind | 220.37 GW | 11.7% |
| Other | 45.88 GW | 2.4% |
| Offshore Wind | 24.93 GW | 1.3% |
| Nuclear | 10.39 GW | 0.6% |
| Geothermal | 4.84 GW | 0.3% |
| Coal | 4.22 GW | 0.2% |
| Hydro | 4.06 GW | 0.2% |
According to LBNL's Queued Up analysis, total capacity in U.S. queues fell about 10% year over year through the end of 2025, and the mix tilted toward gas: proposed gas capacity rose 86% to roughly 253 GW, while solar (773 GW, −19%), storage (749 GW, −16%), and wind (220 GW, −19%) all declined. LBNL also reports that more than 40% of capacity withdraws even after signing an interconnection agreement, and that projects completing in 2025 waited a median of over five years from request to commercial operation. Those figures are LBNL's national totals (active requests, hybrid-inclusive) and are not computed by GridAlmanac.
Methodology & sources
The backbone is the Lawrence Berkeley National Laboratory “Queued Up” project-level dataset (end-of-2025 edition). We include active and suspended requests only; withdrawn and already-operational requests are excluded from the aggregate tables. Capacity is each request's primary component (LBNL mw_1); LBNL's headline 2,061 GW additionally counts the storage component of hybrid projects, which we never mix in.
We matched 782 operational queue requests to operating plant records by state, resource type, capacity, timing, and name/county signals; matched plants show a "time in queue" figure with a median of 4.6 years. These pages are refreshed annually with each LBNL edition.
Source: "Queued Up" (Rand et al., Lawrence Berkeley National Laboratory, 2026 edition) (data as of December 31, 2025). DOE-funded; reused with attribution. Browse the national queue.