How AI Data Centers Disconnecting Simultaneously Caused a 1,000-Mile Disturbance

A July 22 fault caused 3 GW of Virginia data center load to vanish at once, sending disturbances across PJM’s 67-million-person grid

Rex Edison Avatar
Rex Edison Avatar

By

Image: Tara Energy

Key Takeaways

Key Takeaways

  • Over 3 gigawatts of data center demand disconnected simultaneously, destabilizing PJM’s grid for 10 minutes.
  • Sudden gigawatt-scale load loss causes dangerous voltage spikes, threatening grid stability more than power shortages.
  • NERC and PJM are developing mandatory standards to shift data center grid costs onto developers.

On a Wednesday morning in northern Virginia, lights flickered. Air conditioners groaned. Refrigerators made sounds they shouldn’t. No storm rolled through. No transformer blew. Instead, a transmission line went down, and the AI data centers powering your cloud storage, your AI chatbots, and your streaming queues did exactly what their protection systems told them to do: they all disconnected from the grid at once.

More than 3 gigawatts of demand — enough to power several mid-size cities — vanished almost simultaneously. The grid wasn’t ready for that.

When the Lights Blink, Follow the Algorithm

A single transmission fault set off a cascade that took ten minutes and a thousand miles to resolve.

According to Reuters, utility Dominion Energy confirmed that data centers’ internal control systems detected the fault on July 22 and automatically switched to backup power. PJM Interconnection — the largest U.S. grid, serving 67 million people from Washington, D.C. to the Midwest — registered a sudden frequency shift, the electrical equivalent of a cardiac arrhythmia. Bob Marshall, CEO of sensor-network company Ting Labs, told Reuters that his company’s 1.4 million sensors picked up voltage disturbances stretching from D.C. all the way to Chicago. “The power quality took a pretty big hit,” Marshall said. Grid disturbances normally resolve in milliseconds. This one took roughly 10 minutes.

  • More than 3 GW of demand disconnected almost simultaneously on July 22 — roughly 3% of PJM’s total load at that moment
  • Voltage disturbances registered across 1.4 million sensors from Washington, D.C. to Chicago (Ting Labs, via Reuters)
  • Grid disturbances typically correct in milliseconds; this event took approximately 10 minutes to stabilize
  • PJM covers 67 million people across the Mid-Atlantic and Midwest
  • No formal blackouts were declared, but residential customers reported flickering lights and appliance noise

No blackouts were declared. That is cold comfort when residential customers are watching their lights strobe and their appliances stutter without any obvious explanation.

This isn’t a new problem. It’s an accelerating one. In July 2024, according to Data Center Dynamics, roughly 60 Virginia data centers disconnected simultaneously, dumping around 1,500 megawatts of surplus power onto the grid and forcing emergency action to avoid cascading outages. NERC — the North American Electric Reliability Corporation — formed a task force afterward. Two years later, the same pattern repeated at more than double the scale. “The grid is not designed to withstand the loss of 1,500 MW data centers,” NERC’s John Moura warned, per Data Center Dynamics. “At some level it becomes too large to withstand unless more grid resources are added.”

The Physics Problem Nobody Budgeted For

The danger isn’t a shortage of power — it’s what happens when gigawatts disappear faster than the grid can respond.

Everyone worries about not having enough power. The new threat runs in the opposite direction. When massive loads vanish instantly, the grid oversupplies, voltage spikes, and frequency wobbles dangerously. Think of it like every driver on a six-lane highway hitting the brakes simultaneously — no collision, but pure, preventable chaos.

Grid operators want data centers to “ride through” minor faults and stay connected rather than trip offline. Data-center operators push back hard, arguing that staying online during disturbances risks damaging hundreds of millions of dollars in sensitive hardware and precision cooling systems. Both positions are rational. Neither is sustainable as AI campuses scale toward gigawatt-level power appetites. NERC is now developing mandatory modeling standards — including a technical framework called PERC1 — along with high-resolution monitoring requirements designed to give operators better visibility before the next event.

The bill is already arriving. PJM is pursuing regulatory changes to shift transmission and capacity costs directly onto data-center developers rather than spreading them across all ratepayers. States including Texas and Pennsylvania are weighing frameworks that would allow utilities to cut large commercial loads before residential customers during grid emergencies. Dominion Energy and data-center owners in Virginia have also begun adjusting facility control systems so that sites remain connected during brief grid faults, rather than triggering the kind of simultaneous disconnections that rattled the eastern grid on July 22.

The invisible infrastructure behind the AI tools you use every day now has a very visible address. And your electricity rate is about to reflect it — and if you’re wondering where else you might be paying too much without realizing it, the answer may surprise you.

Share this

At Gadget Review, our guides, reviews, and news are driven by thorough human expertise and use our Trust Rating system and the True Score. AI assists in refining our editorial process, ensuring that every article is engaging, clear and succinct. See how we write our content here →