Key Takeaways

  • An Ashburn transmission fault prompted data centers to transfer to backup power, removing more than 3 GW of demand from the PJM grid within seconds.
  • PJM and Dominion Energy stabilized the system without bulk power reliability impacts, although voltage disturbances were reported as far away as Chicago.
  • The event could accelerate new planning, coordination, and technical standards for data centers and other large electrical loads.

A transmission-line fault in Ashburn, Virginia, has exposed a less familiar data center risk for the electricity system. The concern is not simply how much power hyperscale facilities consume. It is what happens when several of them stop drawing that power almost simultaneously.

The fault caused a transmission line to go out of service last Wednesday, according to Dominion Energy. In response, several data centers disconnected from the grid and moved to backup systems. Data Center Knowledge reported that more than 3 GW of load, about 3% of PJM system demand at the time, disappeared within seconds.

That is an enormous and unusually concentrated demand change. Power grids continuously balance electricity generation and consumption, so a sudden loss of load can push frequency and voltage away from their expected ranges. PJM recorded a measurable frequency change but said the incident produced no bulk power reliability impacts.

Still, the effects were visible. NBC Washington reported that lights flickered across a broad area, with disturbances experienced as far away as Chicago. System operators brought conditions under control within minutes.

Each data center involved in the event likely behaved exactly as designed. Uninterruptible power supplies, generators, and associated protection systems are intended to shield computing equipment and customer workloads from utility disturbances. But when many facilities in one geographic cluster make the same protective move at nearly the same moment, an effective site-level response can become a regional grid-management challenge.

Amazon Web Services said its facilities transferred to backup power after the local utility transmission event. The director for energy and water at Amazon Web Services stated those systems performed as designed and prevented customer impact. Amazon Web Services also emphasized that data centers did not cause the original transmission-line fault.

Dominion Energy offered a similar assessment of the grid response. In a statement reported by 7News, the utility said its operations team stabilized the situation within minutes and that system protections worked as intended. Dominion Energy plans to share information with its data center customers as part of efforts to improve reliability.

No serious outage occurred this time. But what if a comparable load drop coincided with another transmission failure, an extreme-weather event, or tight operating conditions? The CEO of ON.energy warned that electrical systems may activate additional protections when they encounter volatility, potentially worsening a disturbance and, in a more severe scenario, contributing to rolling blackouts.

There is precedent for concern. A similar 2024 Virginia incident followed the failure of a 230 kV lightning arrestor. About 1,500 MW of demand from roughly 60 data centers dropped off the grid, nearly prompting power cuts. The latest Ashburn event involved about twice that amount of lost load, underscoring how quickly the region’s risk profile is changing.

Growth adds urgency. NERC has said data center demand is increasing at double-digit annual rates in some U.S. regions, while the U.S. Energy Information Administration projects that national data center electricity consumption could roughly double by 2030 under high AI-adoption scenarios. Ashburn’s concentration of hyperscale infrastructure makes it an early test case, but the issue extends to other markets attracting AI clusters and cloud campuses.

PJM, which manages electricity movement across 13 states and the District of Columbia, is working with NERC on a Large Loads Action Plan. That effort is expected to inform standards for planning, protection settings, modeling, and operational coordination. Existing references include NERC’s TPL and PRC reliability-standard families, along with IEEE Std 1547 for distributed-energy interconnections.

For operators such as Amazon Web Services, Digital Realty, and Equinix, the operational question is becoming more nuanced. Backup power still protects availability, but transfer thresholds, restoration timing, telemetry, and communication with utilities increasingly matter at system scale. The Ashburn fault showed that keeping a data center online and keeping the regional grid stable are related engineering goals, not separate ones.