A power line failure outside Washington, DC, caused lights to flicker across the region because more than 3 gigawatts of data centers stopped drawing power at nearly the same time.
Normally, the grid recovers from such an event in a few seconds. This incident took more than 10 minutes to stabilise.
Data collected by Ting Labs, a startup that monitors electricity through IoT sensors in domestic sockets, showed voltage spiking across the PJM grid from Northern Virginia to Chicago.
The event did not result in a blackout, but it highlighted the strain data centres place on the network. Experts believe this outcome will become more frequent.
The scale of the issue
Northern Virginia sits within PJM’s territory and contains the highest concentration of data centres in the world.
Ricardo de Azevedo, CTO at ON.Energy, told TechCrunch that these events involving large loads are happening more often.
He described the situation as a canary in the coal mine.
The incident echoes a similar event two years ago on the PJM grid. It could foreshadow larger problems if data centres are not built to handle power supply disruptions more effectively.
The PJM Interconnection manages grids from New Jersey to Illinois and serves 67 million customers, making it the largest grid operator in the United States.
When the power line failed this week, data centres switched to backup power. About 3.1 gigawatts of load vanished in roughly 30 seconds, according to PJM data.
The grid appeared to recover somewhat, but additional loads dropped off shortly after. At its peak, PJM’s grid carried an extra 3.49 gigawatts of electricity.
It took another 11 minutes before the system stabilised. The disconnected data centres represented around 3% of total demand on PJM at the time, Reuters reported.
A few percent may seem small, but the electrical grid must operate in a state of near-perfect balance. Supply and demand must be closely matched.
If they are not, voltages can sag or spike. The grid and connected devices tolerate small fluctuations, but larger ones trigger failsafes within the network or individual facilities, causing them to disconnect.
When data centres in Northern Virginia sensed the fluctuation from the failed line, they switched to backup power. This removed their load from the grid.
As more data centres made the switch, they removed greater amounts of load. What started as a relatively small drop in supply became a larger drop in demand. This sent supply surging and caused light bulbs to flicker.
Most data centres make decisions in a split second. The facilities that disconnected this week appear to be no different.
When the voltage dip reached them, they all decided to disconnect within a few seconds of each other, Ali Zain Banatwala, senior market models specialist at the Independent Electricity System Operator, told TechCrunch.
He said operators need to find a way for these loads to sequentially disconnect or reconnect.
A more orderly process would allow grid managers to develop better procedures in advance.
What it means
Alternatively, data centres could be built to absorb disruptions rather than turning their backs on them.
One startup, ON.Energy, has developed a product to help data centres and the grid ride through events like the one this week.
The company has created an uninterruptible power supply for an entire data centre campus. This covers servers, chillers, and other equipment.
The system essentially hides the data centre behind a bank of batteries connected to sophisticated power conversion equipment.
All the grid sees is one consistent, well-behaved load rather than the peaks and valleys from each individual part.
ON.Energy’s system allows data centres to ramp computing workloads up and down, including AI training, without bothering the grid.
More importantly, it means data centres can absorb power fluctuations. Instead of disconnecting, the system uses extra power to charge batteries.
If the flow dips, the system dispatches power to servers. It can also follow the grid’s lead within milliseconds, preventing sags or surges like those that caused the problem for PJM this week.
ON.Energy is currently installing a total of 3 gigawatts of its systems at four different data centre campuses, de Azevedo said.
Grid managers have also woken up to the problem.
ERCOT, for example, is going to require large loads like data centres to ride through disruptions, de Azevedo said.
The clock is ticking. The mass disconnection this week was twice as large as a similar event in 2024.
Back then, 60 data centres simultaneously disconnected, pulling 1.5 gigawatts of load from the grid.
At that time, data centres accounted for about 6% of PJM’s load, Synapse Energy Economics reported.
By 2040, they are expected to make up 24%.
If the problem is not addressed soon, things could get a lot worse.
