The Bottleneck Stalling Clean Power

After two decades in which power demand barely moved, the grid now faces the steepest load growth in a generation, driven by data centers, new factories, and the electrification of cars and buildings. The usual worry is whether there are enough new power plants to be built in time. That worry skips past a quieter and arguably larger obstacle. Even when a project is financed and ready to build, it often cannot get permission to plug in. At the end of 2025, more than 2,000 gigawatts of new generation and storage were sitting in line waiting to connect to the American grid, which is more than the entire existing fleet of power plants currently operating (Lawrence Berkeley National Laboratory, 2026). The biggest constraint on new power is not the power. It is the plug.

A line that stretches for years

The process behind the pile-up is called interconnection. Before any large power plant or battery can join the transmission system, the grid operator must study how it will affect the network, what upgrades it will require, and who should pay for them. That study, once a routine step, has become a multi-year wait.

According to Lawrence Berkeley National Laboratory, which tracks these queues in its annual Queued Up report, roughly 2,060 gigawatts of generation and storage were seeking connection at the end of 2025 (LBNL, 2026). It is worth pausing on what is in that line, because it cuts against the usual framing. Solar and battery storage make up the bulk of it, but natural gas capacity in the queue jumped 86 percent in a single year, to 253 gigawatts. The bottleneck is not slowing one kind of energy. It is slowing all of it, firm and renewable alike, which is the first clue that this is a problem of infrastructure rather than an argument about which technology ought to win. The average wait to get through has stretched to about five years, up from roughly twenty months in 2005, and in the hardest-hit regions the delays run considerably longer (Brattle Group & Grid Strategies, 2026).

Why the jam formed

The interconnection system was built for a grid that no longer exists. The rules most operators still use date to around 2003, designed for a world in which a handful of large power plants applied to connect each year. That system could afford to study projects one at a time, in the order they arrived. It was never meant to absorb thousands of smaller solar and storage projects at once, and it has buckled.

Two design flaws turned a growing backlog into gridlock. The first is that studies were done one project at a time, so a single speculative proposal near the front of the line, one that might never be built, could stall everyone behind it. The second is subtler and more revealing. Under the old approach, each project is assigned the cost of whatever grid upgrades its connection happens to trigger, even when those upgrades are really needed because of the cumulative growth of everyone in the queue. That produces a perverse dynamic. A project unlucky enough to be the one that tips a substation past its limit can be handed a bill large enough to kill it, while the projects behind it inherit upgrades they never paid for. Developers respond rationally, filing speculative applications to hold a place and walking away when the costs land on them. The queue fills with proposals that will never break ground. In 2025 alone, more than 750 gigawatts of requests were withdrawn, which is part of why the total queue actually shrank by about 10 percent that year (LBNL, 2026). Much of what looks like a backlog is really churn.

The collision with surging demand

This would be a manageable engineering problem if demand were flat, as it was for most of the last two decades. It is not flat anymore. Utility forecasts now project U.S. peak demand growing by roughly 166 gigawatts over five years, by 2030, a six-fold jump from what was projected just three years earlier, with data centers responsible for more than half of the increase (Grid Strategies, 2025). The timing could hardly be worse. The queue is seizing at precisely the moment the grid needs to move fastest, and the newest and most impatient customers, the data center developers, are now crowding into the same congested lines as everyone else.

Who pays for the delay

It would be easy to file all of this under technical problems for engineers and regulators. It is not. When new supply cannot connect quickly enough to keep pace with demand, prices rise, and those increases land on ordinary electricity bills. The clearest case is PJM, the grid serving 65 million people across the mid-Atlantic and Midwest, where the price paid to guarantee future generating capacity leapt from about $29 per megawatt-day to about $270 per megawatt-day in a single year, with two transmission-constrained zones clearing higher still. The total cost passed on to consumers rose from $2.2 billion to $14.7 billion (PJM Interconnection, 2024). A meaningful share of that jump traces back to supply that could not get connected in time to compete. The queue is invisible to almost everyone who pays for it, and yet it shows up in what they pay.

The lever almost no one is pulling

Faced with numbers like these, the instinct is to build, and to build faster. But some of the most interesting recent work points in a different direction, toward using the grid the country already has more cleverly before pouring concrete for a new one. A widely discussed 2025 study from Duke University’s Nicholas Institute found that the existing system could absorb between 76 and 126 gigawatts of new demand with no new power plants at all, provided those new loads agreed to be curtailed for a small fraction of the year, on the order of a fraction of one percent of their operating hours (Norris et al., 2025). The intuition is that the grid is built to survive rare extreme peaks, which means it sits underused most of the time. A data center willing to ease off during the few dozen hours a year when the system is most strained could plug into that spare headroom instead of waiting years for new capacity to be built.

The finding is genuinely important, and genuinely limited, and both halves matter. It suggests a large share of near-term demand growth could be met faster and cheaply than the build-everything approach assumes. But it applies to flexible loads, and many data centers, especially those training AI models, are reluctant to pause for even an hour. It does not, by itself, clear the queue for the generation projects still waiting to connect. What it does is reframe the problem. The bottleneck is not only a shortage of wires and plants. It is also a failure to wring full value from the ones already standing.

The fixes, and their limits

The core problem is finally getting serious attention. In 2023, federal regulators issued a landmark overhaul of the interconnection process, and its reforms began taking effect across most regions the following year (Brattle Group & Grid Strategies, 2026). The central change is to study projects in batches rather than one at a time, and to require developers to put real money down before entering the queue, so that speculative applications are filtered out before they can clog the line. The early signs are encouraging. The queue shrank in 2025 and the deadwood is clearing. But a 2026 progress report evaluating those efforts concluded that the process, while improving, remains too slow and too unpredictable to keep pace with the demand now arriving. It is too early to call the fix a success.

The deeper problem is the wires

Even a perfectly run queue would run into a harder wall. You cannot connect a power plant, of any kind, if there is no transmission line with room to carry its output, and much of what gets blamed on interconnection is really a shortage of high-voltage wire in disguise. Here the numbers are sobering. The United States built only about 888 miles of new high-voltage transmission in the most recent year on record, a pace badly out of step with the scale of demand now coming (Grid Strategies, 2025). Building those lines is one of the slowest undertakings in American infrastructure, routinely taking a decade or more, not because the engineering is hard but because of the tangle of siting approvals across state lines and the drawn-out fights over who should pay for a line that benefits an entire region. Fixing the queue speeds up the paperwork. It does not string the wire. Until the country builds transmission far faster than it does today, the bottleneck will keep reappearing in new forms.

The heart of the matter

There is a tendency to picture the future of energy in terms of its most photogenic parts: rows of solar panels, spinning turbines, and sleek reactors. Almost no one pictures an interconnection study or a transmission right-of-way. Yet the evidence keeps pointing to those pieces of grid machinery as the place where the real constraint lives. A country can finance all the power plants it wants, but if they cannot get connected for the better part of a decade, the pace of the whole system slows to the speed of the queue. Whether the steepest rise in electricity demand in a generation can be met will depend less on developing new ways to generate power than on fixing the underlying systems that determine whether that power can reach the grid.

References

Brattle Group & Grid Strategies. (2026). Interconnection progress report. Advanced Energy United. https://advancedenergyunited.org

Grid Strategies. (2025). National load growth report 2025. Grid Strategies LLC. https://gridstrategiesllc.com/wp-content/uploads/Grid-Strategies-National-Load-Growth-Report-2025.pdf

Lawrence Berkeley National Laboratory. (2026). Queued up: 2026 edition. Characteristics of power plants seeking transmission interconnection as of the end of 2025. U.S. Department of Energy. https://emp.lbl.gov/queues

Norris, T. H., Profeta, T., Patino-Echeverri, D., & Cowie-Haskell, A. (2025). Rethinking load growth: Assessing the potential for integration of large flexible loads in US power systems. Nicholas Institute for Energy, Environment & Sustainability, Duke University. https://nicholasinstitute.duke.edu

PJM Interconnection. (2024). 2025/2026 base residual auction report. PJM Interconnection LLC. https://www.pjm.com