Only 13% of the power capacity that entered the U.S. interconnection queue from 2000 to 2020 has actually been built. About three-quarters of it was withdrawn, and high interconnection upgrade costs contribute to many of those withdrawals. [1] Projects that do get built take far longer than they used to. In the regions with available data, the median project completed in 2025 took 61 months (just over five years) from interconnection request to commercial operation, up from 36 months in 2015. [1]

The scale of that waiting room is now larger than the power system it is trying to join. At the end of 2025, the U.S. interconnection queue held over 2,060 GW of generation and storage capacity, compared with 1,374 GW of installed U.S. generating capacity. [1] Simply put, the queue is about one and a half times the size of the system already in service, competing for connection to a grid that has not expanded fast enough to absorb it.

The pressure is especially visible in the markets now driving AI infrastructure demand. Northern Virginia, the largest data center market in the world, is served by Dominion Energy. [2] In 2024, Dominion said that connecting a data center above 100 MW could take as long as seven years, up from as long as four. [2] That wait, for power users, comes from a queue separate from the generation queue measured above.

Key Figures at a Glance

AI Interconnection Queue Management

Order No. 2023, issued by the Federal Energy Regulatory Commission (FERC) in July 2023, was the agency's most ambitious effort to overhaul the U.S. grid interconnection process in decades. [3] Designed to cut mounting backlogs and bring greater certainty to project timelines, the reforms replaced the traditional first-come, first-served approach with cluster studies that evaluate projects in groups. For an industry mired in years-long delays, the order raised hopes that the interconnection crisis was finally beginning to ease.

Three structural factors explain why AI interconnection queue management is a multi-year problem regardless of administrative reform.

1. Grid infrastructure investment has lagged demand for a decade

   The IEA expects global grid spending to reach about $550 billion in 2026, up nearly 20% from a year earlier, after years of lagging behind spending on generation. [4] Part of the rise reflects tight supply chains pushing up the cost of transformers and cables. [4] In the U.S., the pace of new high-voltage transmission has fallen sharply. The country built about 1,780 miles of new lines a year from 2010 to 2014, but only about 540 miles a year from 2020 to 2024. [5] Meanwhile, annual requests to connect surged from 2013 to 2023. [1] The queue is the physical expression of that gap.
   
   

2. The study process is sequential and self-compounding

   Under the serial framework that existed before Order No. 2023, projects were studied one at a time, in order. When a project withdrew, particularly late in the process, other projects in the queue sometimes had to be restudied, which lengthened their timelines. [1] Because most proposed projects are withdrawn, that effect repeated across the queue. FERC Order No. 2023 introduced cluster studies to reduce this effect by evaluating projects in groups rather than one at a time. But implementation has been uneven across regions, and the backlog inherited under the old framework continues to weigh on the new process. PJM Interconnection, for example, has moved to cluster cycles but is still working through requests submitted before 2023. [1]
   
   

3. AI data center demand is arriving faster than reform can absorb

   Amazon, Microsoft, Google, and Meta now expect to spend more than $700 billion combined on capital expenditure in 2026. [6] Dominion Energy, which serves Northern Virginia, reports more than 50 GW of contracted data center capacity in its pipeline. [7] Order No. 2023 governs how generators connect to the grid, not how large loads do. Data center loads move through separate utility queues, and neither process was built for a demand shock of this size.
   
   

The Transformer Bottleneck

Large power transformers are built to order, which limits how quickly manufacturers can scale. Photo: ShadowInspiration/Envato Elements

Clearing the interconnection queue does not resolve the schedule problem. A project that successfully obtains a connection agreement still faces a second constraint at equipment delivery: the transformer supply chain lead time crisis.

Lead times for large power transformers, which typically took a year before the pandemic, now stretch beyond 24 months in most major markets. [8] According to Wood Mackenzie analyst Ben Boucher, generator step-up transformer lead times had passed 160 weeks (more than three years) by early 2026, up from an average of 143 weeks in 2024. [9] The underlying constraint is manufacturing capacity. Only a limited number of manufacturers can build the largest units, and key inputs such as grain-oriented electrical steel, high-voltage bushings and tap changers are constrained too. [8] Demand from renewable energy, data center power infrastructure and grid modernization has outpaced that capacity. [8]

If transformer procurement begins only after financial close, the full lead time of two years or more can sit on the schedule's critical path. Project financial models often account for the queue delay, but not the combined impact of equipment lead times.

The practical implication for procurement strategy is clear. Transformer orders need to start much earlier than pre-2022 norms suggest. Ideally, they should begin at or before interconnection agreement execution, not at financial close.

Three Implications for Project Developers

1. Reprice your schedule assumptions

   Any project schedule built on pre-2022 interconnection timelines or transformer lead times is carrying unpriced risk. The time from interconnection request to operation has nearly tripled since 2008, and transformer lead times jumped after the pandemic. [1][8] Developers and their lenders should therefore run sensitivity analysis against current market data, not historical norms.
   
   

2. Interconnection queue position is a project asset

   Given that about three-quarters of queued capacity is withdrawn, a confirmed queue position carries real option value. Projects that hold active interconnection agreements are holding an asset that took years to obtain, particularly in constrained markets. Competing projects cannot easily replicate their advantage on a compressed timeline.
   
   This value can be underrepresented in development-stage project valuations and should be explicitly modeled in capital allocation decisions. Even an executed interconnection agreement is not a guarantee: 41% of the generation and storage capacity with agreements signed from 2000 to 2022 had withdrawn by the end of 2025. [1]
   
   

3. Early-stage risk intelligence is a structural advantage

   Many of the factors that stretch an interconnection wait are visible before they become embedded in the project schedule. But the window to respond is often short.
   
   Projects that use AI to monitor interconnection queues during early-stage development can act inside that window. Instead of treating queue status as a static metric reviewed once a quarter, they continuously track changes in queue positions, study revisions, and competing load applications that signal emerging risks. That is the information advantage.
   
   

Interconnection Is Now a Core Assumption

The data in this article reflects the current shape of the market, not a forecast. For developers with assets in or approaching the queue, interconnection is no longer a background assumption in the model. It needs to be built into the model: shaping timelines, procurement decisions, financing horizons, and the value of the project long before construction begins.