For the better part of a decade, the story about America's interconnection queue has been the same: it keeps growing, and the wait keeps getting longer. In 2025, that trend broke. The national queue actually contracted, falling roughly 10 percent year over year.
It would be reasonable to read that as progress. The data underneath says otherwise. The queue got smaller mostly because projects gave up, not because they got built.
Lawrence Berkeley National Laboratory's Queued Up: 2026 Edition tracks every generator seeking transmission interconnection in the United States. As of the end of 2025, roughly 2,060 gigawatts of generation and storage were actively waiting in line, spread across about 8,200 projects. For scale, total installed U.S. generating capacity at the same date was about 1,374 GW. There is still roughly half again as much capacity waiting to connect as the grid currently carries.
The composition of that decline is where it gets interesting:
Every clean resource category fell by roughly a fifth. Natural gas nearly doubled. That is not a queue clearing out in an orderly way. It is a queue whose composition is shifting, and the shift tracks closely with rising load forecasts from data centers and electrification, which are pulling developers toward resources that can promise firm, dispatchable output on a schedule.
Here is the statistic that should reframe how anyone reads a project's queue position. Of all the capacity that submitted interconnection requests between 2000 and 2020, only 13 percent had reached commercial operation by the end of 2025. Seventy-five percent had been withdrawn. The remaining 10 percent was still, years later, sitting in the queue. Counted by number of projects rather than megawatts, the picture is only modestly better: 19 percent built, 71 percent withdrawn, 9 percent still waiting.
Put plainly: a project entering the interconnection queue is considerably more likely to be abandoned than to be built.
This matters enormously for how landowners should interpret what a developer tells them. "The project is in the queue" is often presented, and often sincerely believed, as meaningful progress. Historically it has been closer to a lottery ticket. A signed lease option tied to a project in an early queue position is not the same thing as a project that will exist.
The timeline for the survivors is not fast either. For projects that reached commercial operation in 2025, the median duration from interconnection request to commercial operation was more than five years. In 2008, that figure was under two years.
The dominant reason projects withdraw is money, specifically network upgrade costs.
When a generator connects to the grid, the utility studies what transmission infrastructure has to be built or reinforced to accommodate it. Under most regional rules, the developer pays for those upgrades. Berkeley Lab's February 2026 analysis of interconnection costs found a revealing pattern in projects studied between 2018 and 2024:
Withdrawn projects faced interconnection costs roughly three and a half times those of projects that made it. The cost estimate is frequently the thing that kills the project.
Two structural trends make this worse. First, network upgrades have gone from about 35 percent of total interconnection costs in the 2000s to roughly 85 percent in the 2018 to 2024 period. The expense is no longer mostly about the physical tie-in to the substation; it is about reinforcing a grid that was not built for where new generation wants to locate.
Second, the variance is extreme. A separate Berkeley Lab analysis of organized markets, published in early 2025, found that a quarter of studied projects faced interconnection costs below $25/kW while another quarter faced costs more than ten times that. A developer cannot reliably price this risk in advance, which is precisely why so many enter the queue speculatively and exit once the study results arrive.
Resource type matters too. Across 2018 to 2024 studies, average interconnection costs ran about $509/kW for solar, $504/kW for wind, and $437/kW for storage, against roughly $150/kW for natural gas. One important caveat on those figures: they average all projects, including the withdrawn ones carrying the highest study estimates, which were never actually paid. Looking only at projects that made it to completion, the same analysis shows solar at about $216/kW, storage at $151/kW, and wind at $103/kW. The gap between those two sets of numbers is essentially a measure of how much cost estimation is doing the filtering. Renewable projects tend to site where the resource is, which is often where the transmission is weakest.
There is a less-discussed backlog behind the first one. Roughly 549 GW of capacity already holds a draft or fully executed interconnection agreement but has not reached commercial operation. That includes 256 GW of solar, 161 GW of storage, 76 GW of wind, and 45 GW of gas.
That is a large number, and it deserves more scrutiny than it usually gets. These are projects that have cleared the study gauntlet. They know their network upgrade cost, because the study is done and the agreement assigns it. Whatever else is holding them up, it is not uncertainty about the number.
So the question worth asking is whether the number itself is the problem. Two readings are possible, and they have very different implications.
Under the first reading, an executed interconnection agreement means the cost cleared the developer's hurdle rate, and the delay is about everything downstream: financing, equipment procurement, offtake, permitting, supply chain. Under the second, the developer signed the agreement to hold the position, then found that at the assigned upgrade cost the project no longer pencils, particularly once tax credit timing, equipment prices, and achievable PPA rates moved underneath it. In that case the 549 GW is not a pipeline waiting to be released. It is a queue of projects quietly deciding whether to eat the cost or walk.
The public data does not settle this. Berkeley Lab reports how much capacity holds an agreement without reaching operation, but the published series does not break out how much of that capacity subsequently withdraws, or attribute post-agreement withdrawals to cost. What the cost data does establish is that the pattern holds strongly at the earlier stage: withdrawn projects carried cost estimates roughly three and a half times those of completed ones, and the gap between all-project averages and completed-project averages is wide enough that cost estimation is visibly doing the filtering.
If that same dynamic extends past agreement execution, then completion rates and online dates are being set by network upgrade cost allocation more than by anything happening in the study queue, and interconnection reform that only speeds up studies will move the completion rate very little. This is a testable question and a useful one to press developers and regulators on directly: of capacity that reached an executed agreement in the last five years, how much has since withdrawn, and what were the assigned upgrade costs on those projects versus the ones that got built?
The 86 percent jump in gas capacity entering the queue is the clearest signal in the dataset that data center development is well past the announcement stage. Developers do not pay study deposits on speculative load. EIA now forecasts U.S. electricity demand rising through 2027, driven largely by large computing facilities, and describes it as the strongest four-year growth in power demand since 2000.
But interconnection is only one of the two lines these projects are standing in. The other is the turbine order book, and it is longer.
EPRI's March 2026 assessment puts lead times for large gas turbines at more than five years from order to delivery, with an average delivery year of 2031 for an order placed today. Smaller units run 18 to 36 months. Prices have moved accordingly, from roughly $2,000/kW to roughly $3,000/kW in about six months. Global orders in 2025 reached 846 units and 100.3 GW, with 427 units and 43.1 GW in the United States alone, more than double 2024 by capacity.
The manufacturers' own disclosures say the same thing. GE Vernova reported in July 2026 that gigawatts under contract for gas had grown from 100 to 116 GW in a single quarter, split between 53 GW of backlog and 63 GW of slot reservation agreements, and told investors it is mostly sold out through 2030 and expects to have sold more than half of its 2031 production slots by the end of this year. Capacity is expanding, from a 20 GW annualized run rate now to a target of 30 GW by 2030, but that expansion is itself a multi-year build. Siemens Energy and Mitsubishi are adding capacity on similar timelines, and industry estimates suggest a manufacturer doubling its own production lifts total industry output by something in the range of 15 to 40 percent, constrained mainly by skilled labor.
Put the two lines side by side and a timeline problem appears. A gas project entering the interconnection queue today faces a median five-plus year path from request to commercial operation, and a turbine order placed today lands around 2031. Those clocks can run concurrently if the developer has secured a slot, and the serious ones have. But slot reservation agreements are themselves a scarce commodity now, and a developer holding a queue position without one is not five years from operation. The binding constraint has partly moved off the grid and onto the factory floor.
That has three consequences worth watching. Announced data center load may arrive on a schedule that the generation fleet cannot match, which pushes the balancing burden onto existing units, demand response, and whatever can be sited fast. Turbine cost inflation of the magnitude above changes the economics that made gas the default answer in the first place, and narrows the gap against storage and renewables that can be delivered sooner. And the same load growth that pulled gas into the queue is a reason to expect the clean resource categories to come back, because in a market where the firm option is priced at $3,000/kW and dated 2031, the resource that can be energized in 2028 has an argument it did not have two years ago.
Two reform efforts are worth understanding.
FERC Order No. 2023 (2023) replaced the old first-come, first-served serial study process with a "first-ready, first-served" cluster model across FERC-jurisdictional regions. It raised commercial readiness requirements and imposed penalties for withdrawal, on the theory that speculative applications were clogging the process for serious ones. Results through 2026 have been mixed: several regions are signing interconnection agreements at record pace and the storage queue has begun to shrink, while backlogs persist elsewhere.
In California, CAISO's Interconnection Process Enhancements 5.0 is now approved and imminent. CAISO's Board adopted it on April 30, 2026, the tariff amendment was filed in June, and FERC issued its order on the revisions on August 4, 2026. The changes apply beginning with Cluster 16, whose application window opens October 1, 2026. Among them: deliverability allocations are capped at the lesser of 50 percent of a load-serving entity's forecasted resource adequacy load share or 500 MW; projects seeking timeline extensions beyond seven years must show an executed power purchase agreement matching the project's deliverability status; customer-requested extensions are capped at three years cumulative; and distributed energy resources are folded into the standard cluster scoring and study process. Commercial readiness deposits also shift to after project ranking, and pre-application reports are eliminated for small generators at or under 20 MW.
The through-line in both reforms is the same: raise the bar for entry so that queue position means something. Neither one touches the cost allocation question, and neither one builds a turbine.
If you are a landowner evaluating a lease or option offer, the queue data suggests a few concrete questions. Where is this project in the queue, and in which cluster? Has the interconnection study been completed, and what did the network upgrade cost estimate come back at? Is there an executed interconnection agreement, or just a request? Is there an offtake agreement? For a gas project, is there a turbine slot, and for what delivery year? A developer with completed studies, a signed PPA, and equipment secured is in a materially different position than one holding a queue number. Option payments are real money, but a multi-year option on land tied to a class of project that has historically been built less than one time in five deserves terms that reflect that risk.
If you write or influence policy, the cost data points somewhere specific. The single largest driver of project failure is network upgrade cost allocation, and the variance in those costs is what makes the risk unpriceable. Jurisdictions that have moved toward socializing or averaging network upgrade costs, as Texas and Australia have in different forms, have changed developer behavior measurably. Grid-enhancing technologies, "connect and manage" style operational approaches, and proactive transmission planning all address the same underlying problem, which is that the grid is being asked to accommodate a generation fleet in locations it was never planned around.
The queue shrinking is not, by itself, evidence that any of this is working. It is evidence that the filter is getting tighter. Whether that filter is selecting for the right projects, and whether the projects it selects can get equipment in time to matter, is the question worth watching over the next two years.
