A Record Twenty Years in the Making
Sit with that date for a moment. The largest grid operator in the country, serving 67 million people across 13 states, went twenty years without exceeding its 2006 peak. Then AI demand arrived, and the record fell in the same week the National Weather Service forecast highs of 102 to 104 degrees across Maryland and Virginia. PJM’s own operations update describes the curtailment authority as a measure to be used “as a last resort prior to voltage reduction or load shed.” In plain terms: before rolling blackouts, data centers go first.
The exemption list is instructive. Hospitals, 911 call centers, water treatment plants, air traffic control towers, and defense installations are protected. Data centers are not. And this was no one-time improvisation. It is the third round of 202(c) curtailment orders PJM has received in 2026 alone, following Winter Storm Fern in January and an unseasonable May heat wave. Twice a season is not an emergency measure. It is a pattern.
Curtailment Is Now an Operating Condition
These orders work precisely because data centers carry backup generation. But there is a difference between backup power and dispatchable power. Diesel and standby systems are engineered for rare, short-duration outages, not for scheduled service as a grid resource every time temperatures spike or a polar vortex arrives. Operators who committed to four-nines uptime are now planning around a risk that did not exist in their models three years ago: a federal curtailment signal with 15 minutes’ notice, arriving several times a year.
The deeper problem is that the shared grid cannot add supply on the timeline demand is arriving. Interconnection queues stretch five years or more, new transmission takes a decade, and the demand curve steepens every quarter. When reserves get thin, the policy priority is clear and, frankly, correct: homes and hospitals come first. But that clarity cuts both ways. If you are a data center on the shared grid, you now know exactly where you stand in line.
Independence, Literally
This is the problem Next Century-Power was built to solve. Our model is dedicated, on-site generation: 1GW+ gas turbine microgrids in an islanded, single-tenant configuration, transitioning over time to 1.2GW pods of four integrated 300MW small modular reactors. Power that never touches the shared grid cannot be curtailed off of it. There is no 202(c) exposure, no 15-minute signal, no place in anyone’s load-shed sequence. Four-nines uptime is engineered in through redundancy, not bolted on through diesel. And because on-site generation bypasses the interconnection queue entirely, it arrives years sooner than a grid connection would.
This Independence Day marks 250 years since the Declaration was signed. Secretary Wright said this week that maintaining reliable power in PJM territory is “non-negotiable.” For the 67 million people PJM serves, that is true. For the data centers inside its footprint, this week proved the opposite: grid power is very much negotiable, and the terms are not theirs to set. Energy independence is available to any operator willing to generate power where it computes.
