Interconnection queues explained: why new power waits years

An interconnection queue is the waiting line for new power plants and batteries that want to plug into the transmission grid. Each project is studied in turn for the upgrades it needs. Berkeley Lab data, as reported in July 2026, show 2,061 gigawatts waiting at the end of 2025, and most requests are never built.

Data centers need that power, and the queue is one of the reasons electricity has become the pacing item for AI buildouts. This post explains how the line works, what the latest numbers say, and why it matters for anyone planning large computing capacity. It covers the generation side. A companion post covers the separate line forming for large electricity users.

Key numbers at a glance

• 2,061 GW of generation and storage was in active US interconnection queues at the end of 2025, down 10 percent from the end of 2024 (Lawrence Berkeley National Laboratory, "Queued Up" 2026 Edition, as reported by the American Public Power Association on 1 July 2026).
• The peak was nearly 2,600 GW at the end of 2023 (same source).
• More than 750 GW of requests were withdrawn during 2025 (same source).
• Natural gas in the queue rose 86 percent to 253 GW in 2025. Solar fell 19 percent to 773 GW, storage fell 16 percent to 749 GW, and wind fell 19 percent to 220 GW (same source).
• For requests made between 2000 and 2020, status at the end of 2025: 13 percent were online and 75 percent had been withdrawn (same source).
• More than 40 percent of projects withdrew even after signing an interconnection agreement (same source).
• The median wait from request to signed agreement was well over 3 years in 2025, and request to commercial operation was over 5 years in the subset of regions with data (same source).
• In the PJM region, average time from application to commercial operation rose from under 2 years in 2008 to over 8 years in 2025 (RMI analysis, published 4 November 2025 and updated 11 November 2025).
• PJM's most recent studied group, called TC1, drew over 100 GW of applications; 40 GW were studied in the first phase and 17 GW received draft agreements (RMI, same dates).
• ERCOT, the Texas grid operator, reported about 452 GW of generation interconnection requests as of late May 2026, alongside 438 GW of large load requests (ERCOT's Jeff Billo at a board presentation, as reported by New Project Media).
• 27 July 2023: the date FERC issued Order No. 2023, which moved generator studies to clusters; it took effect on 6 November 2023 (ACORE fact sheet; Federal Register).

What an interconnection queue actually is

The US high voltage grid is run in regions. In most of the country a regional transmission organization or independent system operator, often shortened to RTO or ISO, runs the grid and its markets. PJM, MISO, SPP, CAISO, NYISO and ISO New England are examples. Texas has its own operator, ERCOT. Some western and southeastern areas are run by individual utilities instead.

When a company wants to build a solar farm, a gas plant, a wind project or a large battery, it files an interconnection request with the grid operator or utility that owns the nearby lines. The request says where the project is, how many megawatts it will inject, and when it hopes to come online. Then the study work begins.

The studies answer a simple question: if this plant is connected, what breaks, and what does it cost to fix? Adding power at one spot can overload a line, a transformer or a substation somewhere else. The studies find those problems and assign the fix, called a network upgrade, along with its cost, to the project that triggered it.

A queue exists because projects interact. A line that can handle one new plant may not handle ten in the same area. So studies cannot be done in complete isolation, and the results for one project can change when another drops out.

From first come first served to cluster studies

For years most regions studied projects one at a time in the order they arrived. That sounds fair, but it had a flaw. A project at the front of the line could drop out and force everyone behind it to be studied again, because the upgrade list changed. With hundreds or thousands of requests, the line stopped moving.

On 27 July 2023 the Federal Energy Regulatory Commission, the federal regulator of interstate transmission, issued Order No. 2023 to change the approach. The rule was published in the Federal Register on 6 September 2023 under docket RM22-14-000 and took effect on 6 November 2023. ACORE, a renewable energy industry group, describes it as a rule of about 1,500 pages that replaces "first-come, first-served" with "first-ready, first-served." Its core idea is to study projects in groups, called clusters, and to make developers prove they are serious. A summary published by the Compute Law Blog, citing FERC's explainer and law firm analyses, describes the main mechanics:

• Requests that arrive within a 45 day window are studied together as one cluster.
• The transmission provider must finish the initial cluster study within 150 calendar days or owe a penalty of $1,000 per business day.
• Developers post study deposits and a commercial readiness deposit, and total readiness deposits reach 20 percent of the estimated network upgrade cost by the time the agreement is signed.
• Withdrawal penalties rise as a project moves along, from the greater of the study deposit or twice the study costs, up to 5 percent of upgrade costs at restudy, 10 percent at the facilities study, and 20 percent after the agreement is signed.
• Penalties can be waived if the upgrade estimate jumps sharply, by 25 percent or more in a cluster report, or by more than 100 percent in a facilities study.

These figures are the blog's summary, and the order text in the Federal Register is the authority for anyone who needs exact terms. ACORE's fact sheet is consistent on the direction: readiness deposits that rise as a project moves forward, withdrawal penalties that also rise, and penalties on transmission providers that miss study deadlines, which ACORE puts at $1,000 to $2,500 per day depending on the type of study. The rule also ended the old "reasonable efforts" standard, under which a late study carried no fixed consequence, and it requires providers to evaluate alternative transmission technologies, such as advanced conductors, in their studies.

ACORE also records the problem the rule was written to fix. Citing Berkeley Lab, it says that at the end of 2022 at least 10,000 projects, with 1,350 GW of generation and 680 GW of storage, were seeking connection, and that the average time from request to commercial operation had risen to 5 years in 2022 from 3 years in 2015. Citing FERC's own analysis in the order, it adds that two thirds of the interconnection studies completed in 2022 were delayed.

Berkeley Lab notes that the reforms took effect in most regions in 2024 and that it "remains too early to fully understand the effects of FERC Order 2023." That is the right level of caution. Early signs are visible, since the lab reports record volumes of draft or executed agreements, but the long run effect is not yet known.

What the 2026 Berkeley Lab numbers say

The Lawrence Berkeley National Laboratory publishes the standard yearly count, called "Queued Up." The figures below come from the American Public Power Association's 1 July 2026 write up of the 2026 Edition, which covers data through the end of 2025.

Put next to the earlier editions, the numbers trace a clear arc. Roughly 2,030 GW of generation and storage was waiting at the end of 2022, by ACORE's citation of Berkeley Lab. The total peaked at nearly 2,600 GW a year later, just as Order 2023 was taking effect. By the end of 2025 it was back to 2,061 GW, about where it stood three years earlier.

The headline is a small decline. Active capacity fell to 2,061 GW from a peak of nearly 2,600 GW at the end of 2023. That is not because developers lost interest. More than 750 GW was withdrawn in 2025 alone. Part of that is the new rules working as designed, since deposits and penalties push out speculative projects.

The mix changed too. Gas rose 86 percent to 253 GW, while solar, storage and wind all fell. The report is a measure of requests, so it shows what developers are asking for, not what will be built. Still, a jump in gas requests lines up with the push to find firm power for large new loads, and with trade press reports of gas projects tied to data centers in Texas.

By region, the report says the non ISO West held 567 GW and ERCOT 408 GW, and that capacity fell in every region except ERCOT and SPP.

The most important statistic is how few projects finish. Of requests made from 2000 to 2020, only 13 percent were online at the end of 2025 and 75 percent were withdrawn. The remainder were still active or suspended. Even with a signed agreement in hand, more than 40 percent of projects withdrew. So a gigawatt in the queue is not a gigawatt coming. Treat the queue as a ceiling on ambition, not a forecast of supply.

Why projects drop out

Several things cause withdrawals, and they are worth separating because they point to different fixes.

Speculation. Under older rules it was cheap to file. Developers sometimes filed many requests at many sites to see which one got a cheap upgrade bill. Deposits and readiness tests are meant to stop that.

Cost surprises. A project may enter the line expecting a small upgrade and then receive a bill that makes it uneconomic. The Order 2023 waiver rules for large cost jumps address this.

Restudies. When a project ahead of you leaves, your upgrade list can change. Cluster studies reduce this but cannot remove it.

Equipment and permitting. Even a project with a signed agreement needs transformers, breakers, land rights and local permits. Reuters reported on 9 July 2026 that generator step up transformer lead times passed 160 weeks by the first quarter of 2026, up from an average of 143 weeks in 2024 (reporting that cites Wood Mackenzie data). Those delays push dates and raise costs.

Market conditions. A project's economics can shift between filing and agreement, especially when it takes several years.

Why the line got so long in the first place

RMI's November 2025 analysis of PJM gives a concrete picture. The study cycle for TC1, which covered projects that entered between 2018 and 2020, took 544 days, and PJM finished it on 19 September 2025. Over 100 GW applied. 40 GW were studied in the first phase, and 17 GW ended with draft interconnection agreements. RMI also reports that the average time from application to commercial operation in PJM grew from under 2 years in 2008 to over 8 years in 2025.

That timing matters because of a mismatch. RMI notes that PJM runs capacity auctions two to three years ahead, while bringing new generation online has been taking much longer. If the market asks for power in three years and the line delivers it in eight, prices rise. RMI reports capacity prices in PJM went from $29 per megawatt day to the $330 per megawatt day cap across two auctions, and that the total capacity bill rose from $2.2 billion in the December 2022 auction to $16.1 billion in the latest one it cites. These are RMI's figures and RMI is an advocacy and research organization, so read them as one analysis, not an audited tally.

Why this matters to data centers

A data center is a load, not a generator. It does not enter the generator queue. But it competes for the same scarce things, which are transmission capacity, transformers, engineering time and new supply. Two links tie them together.

First, a large new load can only be served if there is enough generation to serve it. If the generation line takes five to eight years, a data center that wants power in two years has to find another route, such as buying existing capacity, building its own generation on site, or accepting limits on when it can draw power.

Second, load requests and generation requests are now being studied together in some places. The Department of Energy's proposal to FERC in October 2025 said large loads should be studied together with generating facilities where practical, because siting a load near a new plant can reduce the upgrades needed. The companion explainer on FERC's large load rules covers that process.

Texas shows how close the two lines now run. At a late May 2026 ERCOT board presentation reported by New Project Media, Jeff Billo, ERCOT's vice president of interconnection and grid analysis, put large load requests at 438 GW, close to the roughly 452 GW of generation requests. The same report counted 17 new gas generation requests in the previous 60 days, against 18 battery storage applications. Both totals are requests, not commitments, and the same caution about completion rates applies to each.

For readers following AI infrastructure money, the same logic applies to the big commitments. A multi gigawatt compute plan assumes the power shows up. See the companion explainers on OpenAI's compute commitments and Stargate and Anthropic's compute commitments for how those plans are described, and what a gigawatt of AI capacity costs for the money side.

What is firm and what is uncertain

Firm, based on the sources above:

• The active queue was 2,061 GW at the end of 2025 and had fallen 10 percent in a year, per the Berkeley Lab figures reported by APPA.
• Most requests historically were withdrawn and few were built.
• Order 2023 changed the process to cluster studies with deposits and penalties, and it took effect in most regions in 2024.
• PJM's average wait grew past eight years by 2025 on RMI's measure.

Uncertain:

• Whether Order 2023 will shorten timelines. Berkeley Lab says it is too early to tell.
• How much of the queue is real. Completion rates for old requests are low, so any total overstates future supply, by an amount nobody can state precisely.
• Exact Order 2023 deposit and penalty figures, which come from a secondary summary and should be checked against the order text.
• Timing for projects in non RTO areas, where data are thinner. Berkeley Lab's timing figures cover only the regions with data.

What to watch next

• The next Berkeley Lab edition, normally published about a year after this one.
• Whether withdrawals stay above 750 GW a year or slow as speculative requests clear out.
• Whether gas keeps growing as a share of the queue.
• PJM's progress on faster tracks and on the next study group, and ERCOT's batch process described in the companion Texas explainer.
• Equipment lead times for transformers and breakers, which can hold up projects that have cleared the studies.

Safeguard's SBOM and dependency scanning covers the software side of infrastructure projects, and the free scan is at https://gold.safeguard.sh/scan.

Quick check

Sources

• American Public Power Association, "Backlog of power plants seeking transmission grid connection eased somewhat in 2025: LBNL," reporting the Berkeley Lab Queued Up 2026 Edition, 1 July 2026: https://www.publicpower.org/periodical/article/backlog-power-plants-seeking-transmission-grid-connection-eased-somewhat-2025-lbnl
• RMI, "PJM's speed to power problem and how to fix it," published 4 November 2025, updated 11 November 2025: https://rmi.org/pjms-speed-to-power-problem-and-how-to-fix-it/
• Compute Law Blog, "How AI data centers connect to the grid," summarizing Order No. 2023 and the DOE proposal: https://computelaw.blog/power/how-ai-data-centers-connect-to-the-grid-ferc-power-queue/
• Reuters via Kitco, "US power companies scramble to secure equipment as surging data center demand strains supplies," 9 July 2026: https://www.kitco.com/news/off-the-wire/2026-07-09/us-power-companies-scramble-secure-equipment-surging-data-center
• New Project Media, "ERCOT large-load requests climb to 438 GW," citing ERCOT's Jeff Billo: https://newprojectmedia.com/origination-ercot-large-load-requests-climb-to-438-gw-17-new-gas-generation-filings-in-last-60-days/
• Federal Register, FERC Order No. 2023, Improvements to Generator Interconnection Procedures and Agreements, 6 September 2023: https://www.govinfo.gov/content/pkg/FR-2023-09-06/html/2023-16628.htm
• ACORE, fact sheet on FERC Order 2023, September 2023: https://acore.org/wp-content/uploads/2023/09/ACORE-Fact-Sheet-FERC-Order-2023.pdf
• US Department of Energy, Secretary of Energy letter and proposed rulemaking on large load interconnection, 23 October 2025: https://www.energy.gov/sites/default/files/2025-10/403%20Large%20Loads%20Letter.pdf

Last reviewed: 19 October 2026.