What the queue is and why it is long

Each year Lawrence Berkeley National Laboratory publishes a report called Queued Up, which counts the power plant and battery projects waiting to connect to the transmission grid in the United States. The 2025 edition, covering the end of 2024, found roughly 2,300 gigawatts of generation and storage capacity in active queues, nearly twice the capacity of the entire existing American power plant fleet. It also found that the typical project reaching commercial operation between 2018 and 2024 had spent more than four years getting there, against less than two years for projects built in the early 2000s. Those two findings, taken together, describe the problem better than any adjective could: an enormous amount of proposed supply, moving slowly, most of which will never be built.

What is an interconnection queue? It is the line a project joins when it asks a transmission provider for permission to connect. Joining is cheap. Leaving the queue with a signed agreement is not. Between the two sit engineering studies that ask what would happen to the network if the project were switched on. Would a line overload somewhere? Would voltage sag at a substation two counties away? If the answer is yes, the study identifies network upgrades, and the project is usually asked to pay for them.

The trouble starts with the order of operations. For most of the past two decades, queues were served first come, first served. Each project was studied roughly in sequence, and each study assumed the projects ahead of it would be built. When an earlier project withdrew, as many did, the assumptions behind later studies fell apart and they had to be redone. Developers learned to file many speculative requests, since holding a place in line was worth more than the application fee. The queue swelled with projects that were never going to be built, and the engineers studying it spent much of their time on ghosts.

Berkeley Lab's data make the attrition plain. Only a minority of the capacity that has ever entered American queues has reached commercial operation; most requests are eventually withdrawn. The 2025 edition also reported that total active capacity fell during 2024, driven by fewer new requests and a large wave of withdrawals. That is not necessarily bad news. Some of the withdrawn projects were speculative placeholders, and clearing them out can shorten the wait for the projects that remain.

So far this describes generators. Why does it matter for a data center, which consumes power rather than producing it? For two reasons. First, a new large load often needs new supply to serve it, and that supply has to pass through the generator queue. A campus promised power in a given year is implicitly betting that the plants and batteries expected to serve it will clear their studies on time. Second, large loads also need transmission upgrades of their own, and the studies for those upgrades compete for the same engineering staff and the same substation capacity.

Consider what a four-year median wait implies for someone planning a facility. A developer who begins site selection today is, in effect, bidding on the grid as it will look several years from now. The site with the shortest queue in the region today may not be the site with the earliest energization date, because the queue position of the supply that would serve it matters as much as its own.

Faraday's read is that the queue is the most honest document in the power sector. Press releases announce gigawatts; the queue records which of those gigawatts are still paying study deposits. Reading it well means looking past the headline total to the projects that have signed interconnection agreements, posted security and begun construction. Those are the projects that change what a region can serve.

Berkeley Lab has since published a newer edition covering the end of 2025, and the specific totals will keep moving. The mechanism described here, a cheap place in line followed by slow and expensive studies, is what the reforms in the next lesson set out to change, with results that are still coming into view.