The Duck Curve

What is an interconnection queue?

Before a new solar farm, wind farm, battery or gas plant can send power onto the high-voltage grid, the grid operator or utility that runs the wires has to work out whether the grid can take it, and what has to be built if it cannot. The interconnection queue is the list of projects waiting for that answer. At the end of 2025 about 8,200 projects were in it across the US, representing 1,312 GW of generation and about 749 GW of storage, according to Lawrence Berkeley National Laboratory's Queued Up: 2026 Edition. That 2,061 GW is more than the entire installed US power plant fleet, which Berkeley Lab puts at 1,374 GW.

Most of it will never run. Berkeley Lab found that of the capacity that asked to connect from 2000 through 2020, only 13% had reached commercial operation by the end of 2025; 75% had been withdrawn and 10% was still waiting. Part of the reason is that asking was easy: the old rules, FERC says, let a developer proceed through the process "without having to show sufficient evidence of ability to develop its project."

How a project moves through it

Requestdeposit, site controlCluster studyupgrades identifiedFacilities studycosts assignedAgreementwho builds, who paysOperationafter upgradeswithdrawn: 75% of capacity requested 2000–2020, before the rule
The path under FERC Order No. 2023. The withdrawal share, from Lawrence Berkeley National Laboratory's Queued Up: 2026 Edition, covers requests made before the rule took effect.

A developer files an interconnection request for a specific point on the grid. The transmission provider then runs a series of studies that, in Berkeley Lab's words, "establish what new transmission equipment or upgrades may be needed and assigns the costs of that equipment." The studies end in an interconnection agreement, the contract that sets the operating terms and who pays for what. Then the upgrades get built, and only then does the project run.

It takes a while. The median project that reached commercial operation in 2025 had spent 61 months between its interconnection request and operation, Berkeley Lab found, against 36 months for projects built in 2015 and 22 months in 2008. Of the 2,061 GW in the queue, 549 GW already held a draft or executed agreement at the end of 2025 and had not yet been built.

Who pays for the upgrades

The expensive part is usually not the plant's own connection but the network upgrades: new lines, transformers and substation equipment elsewhere on the grid that the new project would overload. In the federal process the studies assign those costs to the projects that trigger them, and the number can arrive late. When ISO New England finished its transitional cluster study in August 2026, it put an estimated $65.3 million on a 125 MW battery in Tewksbury, Massachusetts, $56.3 million of it network upgrades; the developer withdrew it, as The Duck Curve reported, and an ISO New England spokesperson told WBUR the average across the study's 23 projects was $178 million. Berkeley Lab says limited transmission capacity has produced high upgrade costs that contribute to high withdrawals.

Small projects face the same problem in miniature. In Massachusetts, home solar connects under the utilities' state-approved tariffs, and the customer whose system happens to overload a neighborhood transformer has paid for the whole replacement. In August 2026 the state's regulators approved a replacement that takes effect November 1, 2026: a flat $225 fee paid by every applicant with a listed inverter-based solar or battery system of 25 kW or less, with the pooled money covering up to $10,000 of upgrades for Eversource and National Grid customers ($1,450 for Unitil) and anything above that still falling to the applicant (Duck Curve coverage). The question is the same at every scale: whether the project that tips the grid over its limit pays for everyone who comes after it.

What FERC changed

For most of the grid's history, requests were studied one at a time in the order they arrived, "first-come, first-served." The Federal Energy Regulatory Commission's Order No. 2023, issued July 28, 2023 and effective November 6, 2023, replaced that with "first-ready, first-served." Per the commission's explainer, the rule, as revised and clarified by Order No. 2023-A in March 2024, has transmission providers:

  • study requests in batches, or clusters, with a 150-day cluster study followed by a facilities study;
  • split network upgrade costs within a cluster by a "proportional impact method," according to how much each project contributes to the need for each upgrade;
  • require larger study deposits, 90% site control at the request and 100% at the facilities study, commercial readiness deposits that grow toward a share of the project's upgrade costs, and penalties for withdrawals that raise others' costs or delay them;
  • meet firm study deadlines, with penalties for missing them, in place of the old "reasonable efforts" standard; and
  • publish a heatmap of available transmission capacity, so developers can see where the grid has room before they apply.

Whether the reforms work is still open. Berkeley Lab says most regions have moved to cluster studies or are moving, and that it is too early to measure the full effect. The deadlines answered a real failure: of the 2,179 interconnection studies completed in 2022, 68% were issued late, FERC says. The stricter process cuts both ways. In PJM's first reformed cycle, a 750 MW project that combined advanced nuclear, fuel cells and gas was rejected for an unresolved modeling deficiency, and FERC upheld the rejection on September 24, 2026 (Duck Curve coverage).

One big grid sits outside all of this, and there the developer does not pay for the upgrades at all. The Texas grid operator, ERCOT, is not under FERC, and under its "connect and manage" approach, Berkeley Lab notes, generators do not pay for network upgrades.

What it is not

The queue is not a forecast. "Ultimately, most of this proposed capacity will not be built," Berkeley Lab says of its own count. It is also not the queue for data centers. Berkeley Lab's report covers only resources that supply electricity; large loads go through separate processes that are only now getting their own rules, including the ride-through requirements PJM proposed after about 3,800 MW of load in northern Virginia, which Dominion's preliminary findings identify as data centers, switched to backup power during a single July 2026 fault (Duck Curve coverage).

Where the queues, the cost fights and the large-load rules stand today is the job of the maintained Interconnection queues and costs in the US reference page; the news as it happens is under #interconnection.

Sources