PJM Weighs Ride-Through Rules After 3,800 MW of Data Center Load Trips Offline in Seconds
A sudden 3.8 GW load dropout at PJM has forced grid operators across the U.S. to confront how fast computational loads can destabilize the system.
PJM Interconnection is moving to tighten interconnection reliability requirements for data centers and crypto-mining facilities after roughly 3,800 megawatts of computational load tripped offline simultaneously, the grid operator disclosed on Tuesday (2026-08-11). The scale of the event — nearly the output of four large nuclear plants vanishing from the grid in seconds — caught the attention of regulators who have spent years watching data center demand surge without equivalent scrutiny of how that demand behaves under stress.6
PJM serves 67 million people across territory stretching from Virginia to Illinois. A load dropout of that magnitude in the wrong conditions could cascade through adjacent systems, forcing generation to compensate at speeds that automation, not human operators, must manage. The proposed "ride-through" standards would require computational loads to stay connected during grid disturbances rather than switching instantly to backup power, a practice that looks sensible at the facility level but creates serious problems at the system level.6,5
The timing is awkward for PJM. Its latest capacity auction cleared at $16.4 billion, matching the record set last December and more than eight times what the grid operator has paid for capacity in prior years, according to Canary Media's reporting on the auction results. Customers in Illinois, New Jersey, and Pennsylvania are already absorbing electricity bill increases of more than 10% in part because of those soaring costs. Adding reliability obligations to data centers — which have expanded rapidly inside PJM territory — risks slowing interconnection at precisely the moment the region is trying to attract new generation.5
Texas got there first. The Electric Reliability Council of Texas voted on Tuesday (2026-06-02) to require data centers to stabilize their grid behavior as a condition of connecting, with the new rules creating what developers there called a "Batch Zero" process for working through the interconnection queue. E&E News reported that while the rules necessarily create winners and losers, data center developers expressed relief at having a defined path forward rather than regulatory uncertainty. ERCOT's approach is now being watched closely by other grid operators.2,3
The underlying demand trajectory makes regulatory action difficult to defer. Global data center electricity consumption grew 17% in 2025, according to the International Energy Agency, with AI-focused facilities growing even faster — up 50% in that year alone. In the U.S., data centers account for roughly half of all incremental electricity demand growth, per IEA figures. That growth has overwhelmed the interconnection system's capacity to process new supply. Of all the generation projects that submitted interconnection requests between 2000 and 2019, only 13% had reached commercial operations by the end of 2024; 77% had been withdrawn.1
The supply-demand mismatch is playing out in queue dynamics that Utility Dive described in detail on Tuesday (2026-08-25). Grid operators need verified as-built models of data center IT loads, cooling systems, uninterruptible power supplies, protection settings, backup generation, transfer logic, and reconnection timing. They also need real-time telemetry — real and reactive power, voltage, transfer status — to anticipate what a facility will do in a contingency before it actually does it. Without that visibility, a 3,800 MW dropout is not a hypothetical; it already happened.7,6
The Asia-Pacific region has moved further along this trajectory. A Wood Mackenzie report cited by Data Center Dynamics in June found that data center developers across the region are increasingly required to provide active grid services — frequency response, voltage support — as a condition of securing power access. That model, where computational loads earn grid rights partly by contributing to system stability rather than just consuming power, represents a more demanding version of what PJM and ERCOT are now considering.4
Whether U.S. grid operators can move fast enough is far from settled. PJM's queue problem is structural: the rules for processing new interconnection requests were designed for a grid where supply and demand grew predictably, not one where a single data center campus can add gigawatts of load in months. The capacity auction costs now flowing through to consumer bills are partly a symptom of that mismatch — new generation is clearing at elevated prices because the pipeline of deliverable projects has been so thin.5,1
The more immediate test is what PJM's proposed ride-through standards will actually require of existing data center interconnections, not just new ones. Retrofitting protection settings and transfer logic at operating facilities is technically feasible but commercially contentious. Companies that designed their backup systems to switch instantly — a sensible choice for uptime — will face real costs to modify that logic. How PJM handles the existing fleet, and whether it phases requirements or grandfathers current operators, will shape the industry's response more than any rule applied only to new entrants.6,7