Bianca Giacobone contributed reporting to this piece.
When a transmission line faulted in Northern Virginia last week, some three gigawatts of data center load disconnected from the grid and switched to back-up diesel power.
It marked the largest drop-off event from data centers to date, and caused voltage levels across the PJM region to spike. While the grid operator and Dominion Energy said they stabilized the grid within about 10 minutes and avoided any outages, the event underscored the growing risks that data centers pose to grid reliability.
“In the worst case scenario, this can continue to cascade,” Doug Bryan, senior manager of power and energy systems at Carbon Direct, told Latitude Media. “So far we haven’t had any rolling blackouts, so the systems in place are dealing with the problem. But we’re trending toward a situation that is more difficult to deal with, and a response is needed.”
The increasing frequency and size of these events has led regulators and grid operators to start cracking down on data centers trying to connect to the grid. Watchdogs from Texas to Washington, fearing cascading blackouts, are pushing mandatory “ride-through” standards that require data centers to stay plugged in during minor voltage changes. And that push could upend facility designs and spur demand for specialized power hardware and backup battery storage.
‘Ride through’ rules underway
A main issue is that many existing data centers’ control systems are designed to protect expensive IT equipment and keep servers running. So when a lightning strike or transmission fault causes temporary voltage sags, a data center may quickly disconnect from the grid to shield electronics from getting damaged.
But if a massive cluster drops off at once, that suddenly leaves a local power grid with a lot of excess power, creating frequency spikes that can trip power plants offline and trigger wider blackouts. The larger the load drop-off, the more generation grid operators have to ramp down to stabilize the system, Bryan said.
Those risks led North America’s grid reliability watchdog in May to issue a rare Level 3 alert to data center developers, utilities, and grid operators. The North American Electric Reliability Corp., or NERC, signaled that it plans to subject data centers to reliability standards similar to power plants, including that they “ride through” minor grid disturbances. Final standards need to be approved by the Federal Energy Regulatory Commission, and may be years away.
In the meantime, NERC issued reliability guidance to grid operators urging them to voluntarily adopt ride-through requirements as part of interconnection studies. That document nodded to rules already underway in ERCOT, which has quickly become the country’s second-largest data center hub behind PJM.
Texas energy regulators earlier this month approved rules establishing minimum times that data centers and crypto miners must stay connected during grid disturbances. The rules take effect on August 1 and apply to new large loads, not existing data centers or projects far along in the interconnection process. However, if existing data centers in the state don’t ride through certain grid disturbances, they have to report the root cause and take corrective action within 180 days.
The Data Center Coalition, which represents major hyperscalers like Amazon, Google, Meta, and Microsoft, opposed ERCOT’s rules mainly by challenging its legal authority to implement them and requested that the state’s Public Utility Commission initiate a rulemaking instead.
Aaron Tinjum, executive vice president of policy, regulatory and strategy for the coalition, said in an email that voltage ride-through requirements are built on assumptions about data center equipment and facility design — assumptions that may not be rooted in reality.
“Getting those assumptions right requires greater communication with data center owners and operators,” Tinjum said, adding that the coalition is engaging with NERC and other reliability organizations. “It is important that requirements distinguish between planned and existing facilities, recognize the technical capabilities and limitations of cooling and server equipment, and include implementation timelines that are feasible.”
A boon for grid-safe power tech
The emerging ride-through rules are a boon to companies that make uninterruptible power systems, like ON.energy, as well as data center developers like Verrus that design campuses specifically to be flexible grid assets that stay connected during voltage faults.
ON.Energy last week announced that Crusoe will deploy 5 GW of its UPS, which act as a shock absorber against both grid voltage faults and a data center’s spiky electricity use during AI training and inference. The UPS is medium-voltage electrical equipment with a backup battery, two power conversion systems, and a transformer.
Dax Kepshire, COO of ON.Energy, told Latitude Media that the system can be dropped into an existing data center campus without a major site redesign or new substation. Some of the company’s customers have already seen expedited interconnections, including one in ERCOT that he didn’t name.
Meanwhile, Verrus — a company that spun out of Alphabet’s Sidewalk Infrastructure Partners — is building all of its data centers with ride-through capabilities. Verrus has announced some projects outside Portland, Oregon, and in Michigan.
“Our system goes above and beyond ride-through capability,” Jeff Bladen, Verrus’ head of energy, said in an interview. “Not only does it avoid disconnecting from the grid, it also helps support the grid in recovering frequency” by charging or discharging battery storage, he added.
He said Verrus has a medium-voltage distribution network that’s like a microgrid, large-scale batteries, and proprietary software with automated dispatch algorithms to coordinate those systems during grid disturbances. The company validated its performance against an event similar to one in Northern Virginia in July 2024, when a transmission fault triggered 1.5 GW of data center load to trip offline. The test was on a digital twin of one of Verrus’s 70-MW data centers under a simulation run by the National Laboratory of the Rockies.
Bladen said the data center industry must do more to protect the grid, but also gave credit to PJM and Dominion for quickly stabilizing the grid during the massive data center disconnection last week: “Can we do better? Yes, and I think the data center industry has a responsibility to keep investing in protecting the grid going forward.”
