Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

A Fallen Power Line Exposed a Growing AI Data Center Grid Problem: 3.1GW Vanished in 30 Seconds

A downed power line near Washington, D.C. triggered a cascading grid event as over 3.1 gigawatts of data center load disconnected simultaneously from the PJM grid, causing voltage spikes from Northern Virginia to Chicago. Experts warn the problem will worsen as data center power demand grows from 6% to an expected 24% of PJM load by 2040.

Published
一根倒下的电线暴露AI数据中心电网隐患:3.1GW负荷瞬间消失,未来或更糟
Image source: techcrunch.com

When a power line went down outside Washington, D.C. this week, it triggered a chain reaction that exposed a growing vulnerability in the electrical grid: mass simultaneous disconnection of AI data centers. Within 30 seconds, 3.1 gigawatts of data center load vanished from the PJM grid, causing voltage to spike from Northern Virginia to Chicago and lights to flicker across the region.

According to PJM Interconnection data, the failed power line caused data centers to switch to backup power, removing their load nearly simultaneously. PJM's grid briefly recovered before additional loads dropped off. At peak, the grid had an extra 3.49 gigawatts of electricity it couldn't handle. It took 11 minutes before the system stabilized. The disconnected data centers represented roughly 3% of total PJM demand at the time.

Northern Virginia, within PJM's territory, hosts the world's highest concentration of data centers. PJM manages grids from New Jersey to Illinois, serving 67 million customers as the largest grid operator in the United States.

"It's the canary in the coal mine," Ricardo de Azevedo, CTO at ON.Energy, told TechCrunch. These sorts of events involving large loads like data centers are "happening more and more," he added.

The problem stems from the grid's need for near-perfect supply-demand balance. When data centers in Northern Virginia detected voltage fluctuations from the failed line, they all independently switched to backup power within seconds. What started as a small supply drop became a large demand drop, sending supply surging.

"We need to figure a way for these loads that are located next to each other to sequentially either disconnect or reconnect," Ali Zain Banatwala, senior market models specialist at the Independent Electricity System Operator, told TechCrunch. "A more orderly process would allow grid operators to develop more robust procedures in advance."

An alternative approach is being developed by ON.Energy, which has created an uninterruptible power supply system for entire data center campuses. The system hides the data center behind a bank of batteries with sophisticated power conversion equipment, presenting the grid with one consistent, well-behaved load. ON.Energy is currently installing 3 gigawatts of its systems at four data center campuses.

Grid operators are waking up to the problem. ERCOT, for example, is moving to require large loads like data centers to "ride through" disruptions. But the clock is ticking — this week's event was twice as large as a similar 2024 incident when 60 data centers simultaneously disconnected 1.5 gigawatts. By 2040, data centers are expected to account for 24% of PJM's load, up from 6% today. Without intervention, the risks will only escalate.

Why it matters

This week's grid event reveals a systemic vulnerability in AI data center infrastructure: uncoordinated mass disconnection. With data center power demand projected to quadruple by 2040, the incident serves as a critical warning for grid operators and data center developers to implement sequenced reconnection protocols.

AI InfrastructureData CenterPower GridPJM
Back to realtime news

Nearby Updates

All