πŸ“° Key Highlights

In mid-July 2026, a power line near Washington, D.C. failed. The grid should have self-healed within seconds, but this time it took over 10 minutes to stabilize because more than 3 gigawatts (GW) of data centers nearly simultaneously cut their power draw. IoT sensor startup Ting Labs detected severe voltage fluctuations across the PJM grid (spanning Northern Virginia to Chicago). The event didn’t cause a widespread blackout, but triggered flickering lights across the region. Northern Virginia is both the most data-center-dense area within PJM’s territory and globally. After the line failure, data centers detected the power fluctuation and triggered automatic switchover to backup power; within about 30 seconds, 3.1 GW of load vanished from the grid. The grid briefly appeared to stabilize, then more load dropped offline β€” at peak, there was 3.49 GW of excess power on the grid, and it took another 11 minutes to return to stability. The disconnected data centers accounted for about 3% of PJM’s total power demand at the time. Grid operation requires near-perfect supply-demand balance; if data centers disconnect from the grid in synchronized fashion within seconds, supply spikes while demand plummets, causing dramatic voltage swings that trip protective relays. ON.Energy CTO Ricardo de Azevedo called the incident a “canary in the coal mine,” warning that such events are becoming more frequent β€” and this wasn’t the first time: a similar event hit the PJM grid two years ago. PJM is the largest grid operator in the U.S., serving 67 million customers from New Jersey to Illinois.

πŸ’¬ JudyAI Lab Perspective

A single line failure on the U.S. PJM grid caused more than 3 GW of data centers to nearly simultaneously disconnect and switch to backup power, triggering severe voltage swings that took over 10 minutes to stabilize β€” showing that data center power consumption has grown large enough to shake an entire regional grid.

This incident highlights a design blind spot that’s often overlooked: data center backup power systems were originally designed to protect the facility itself, with no consideration of how synchronized load drop-off affects overall supply-demand balance on the grid. As AI compute demand pushes data center density higher, the fault-tolerance mechanism at a single site can actually become an amplifier for systemic risk. ON.Energy’s CTO called the incident a “canary in the coal mine,” implying similar events will become more frequent β€” and this is already the second occurrence on the same grid within two years.

For AI builders, it’s worth thinking about whether the infrastructure your projects depend on carries similar single-point cascading risks β€” rather than waiting for an incident to find out.

πŸ“… Source Info

πŸ”— Further Reading