A single data center can consume as much electricity as tens of thousands of homes, changing how grids fail and who loses power first
Karen Fox / Getty Images
In July 2024, a single piece of equipment failed on a transmission line in Northern Virginia, and 60 data centers consuming 1,500 megawatts of power went offline at once. The facilities in Fairfax County all switched to backup generators en masse, forcing grid operator PJM Interconnection and local utility Dominion $D to scramble to prevent widespread blackouts across the region, according to a North American Electric Reliability Corporation (NERC) incident report.
The incident illustrates something that power-systems researchers have been warning about: When massive, concentrated electricity loads cluster in a single area, the physics of grid failure change. Problems that once would have spread out across a large network can instead pile up in one spot, raising the risk of deliberate power cuts to nearby homes and businesses.
Woody Zhu, an assistant professor at Carnegie Mellon University who studies power systems and data centers, recently told Quartz that rapid data center growth in regions not designed for this level of sustained load can create grids that operate in a "near-permanently stressed regime."
Here's what to know.
Electrical grids are designed around the assumption that demand is diverse and distributed. Thousands of homes, offices, and factories draw power in varying amounts at different times. That variation gives grid operators flexibility: When one part of the system is stressed, the load can be balanced across other parts with spare capacity.
Large data centers break this pattern. Unlike most commercial or industrial customers, they consume enormous amounts of power around the clock. A single modern facility can consume as much electricity as tens of thousands of homes. AI-focused data centers are even more energy-intensive because they rely on high-performance computers that run continuously, according to Zhu, writing in The Conversation.
"Rather than stress propagating gradually through diversified demand, stress can localize sharply around specific nodes," Zhu said. This means a single equipment failure near a data center cluster can produce a domino effect. When increased demand for power is concentrated in one area, a single equipment failure can trigger a domino effect, Zhu has found in his research on power grid resilience.
The Northern Virginia incident in July 2024 showed the reverse side of this problem. Traditional UPS systems instantly transfer the data center load to the battery during a grid disturbance and do not transfer it back until the grid is stable for a defined period. While this protects IT equipment, it also creates sudden, large demand drops that pose risks for grid operators, according to Schneider Electric. John Moura, NERC's director of reliability assessment and system analysis, has warned: "As these data centers get bigger and consume more energy, the grid is not designed to withstand the loss of 1,500MW data centers. At some level, it becomes too large to withstand unless more grid resources are added."
When grid stress becomes acute, operators turn to load shedding: the intentional disconnection of electricity demand to ensure the available supply meets what remains, preventing a cascading collapse of the power grid. For a household, this means the power goes off. Not because of a storm or a downed line, but because a grid operator has decided the system cannot serve everyone at once.
When demand for electricity approaches available supply, it is sometimes necessary to temporarily interrupt electricity delivery to maintain the integrity of the electric grid. Several factors can lead to load shedding, including extreme weather, sharply increased electric demand, unplanned outages at generation plants, and transmission constraints, according to Entergy $ETR, a major U.S. utility. Load-shedding power cuts can last from a few minutes to a few hours.
Americans have experienced this on a catastrophic scale. During Winter Storm Uri in February 2021, ERCOT ordered 20,000 MW of rolling blackouts to prevent grid collapse, the largest manually controlled load shedding event in U.S. history. More than 4.5 million people in Texas lost power, some for as long as four days, according to a joint report by the Federal Energy Regulatory Commission and NERC. The storm contributed to at least 210 deaths, according to the Texas Comptroller.
The concern with data-center-driven grid stress is not that the next outage will match that scale, but that the conditions for forced load shedding become more frequent and more localized. Zhu warned that without changes, rapid data center growth risks "cascading outages or forced load shedding" in nearby communities.
The North American electric grid faces intensifying reliability risks over the next decade as demand growth driven by data centers and artificial intelligence threatens to outpace additions to the grid, according to NERC's 2025 Long-Term Reliability Assessment, published in January 2026. Thirteen of 23 North American assessment areas face elevated or high resource adequacy risks over the next five years.
New data centers account for most of the projected increase in North American electricity demand over the next 10 years, NERC's report found. Several parts of the country face "high risk" for electricity supply shortfalls by the end of the decade, including states served by MISO starting in 2028, and parts of Texas, the northwestern U.S., and the mid-Atlantic starting in 2029, according to The Hill.
PJM Interconnection, which serves 65 million people across 13 states, projects peak load growth of 32 gigawatts between 2024 and 2030, with data centers responsible for 94% of that increase, according to PJM's 2025 long-term load forecast. Data center development is heavily concentrated in specific areas, such as Northern Virginia's "Data Center Alley" and parts of Ohio.
The issue is not just total demand but how fast it arrives. On May 4, 2026, NERC issued a rare Level 3 "Essential Actions" Alert in response to repeated events in which 1,000 or more megawatts of computational load dropped off the bulk power system in seconds, according to Carbon Direct's analysis of the alert. Grid instability issues caused by data centers "could become quite severe — to the point of creating widespread blackouts," Ben Inskeep, program director for the Citizens Action Coalition, told Utility Dive.
Reliability problems are often portrayed as major, systemwide blackouts. In reality, as Zhu has written, grid stress shows up in more subtle ways: Voltage fluctuations, equipment overheating, and longer recovery times after storms or heat waves. For communities near data center clusters, the risk is not a single dramatic failure but a grinding reduction in the margin that separates normal operations from emergency.
Data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume 6.7% to 12% by 2028, according to a report from Lawrence Berkeley National Laboratory published by the Department of Energy. The growth trajectory makes the grid-reliability question more urgent with each passing year.
"Long-term sustainability will require more proactive planning," Zhu said, calling for coordinated siting decisions, transmission expansion, local generation, storage, and adaptive demand management. "The system is changing faster than the infrastructure needed to support it," NERC's Moura said, according to Utility Dive.
The question is whether planning can keep pace with construction. So far, it has not.
Join 500,000+ readers who start their day with Quartz.
By subscribing, you agree to our Terms of Service and Privacy Policy.
