PJM moves toward "ride-through" standards after 3.8 GW computational load trip
US grid operators test new reliability rules for data centers after massive simultaneous outages expose vulnerability in interconnection standards.
The PJM Interconnection said Tuesday (2026-08-11) it is considering bolstering interconnection reliability requirements for "computational loads" such as data centers and crypto-mining facilities, following an event where roughly 3,800 MW of such load tripped offline simultaneously.7
The move to evaluate "ride-through" standards and other requirements marks the first major step by a US grid operator to treat data centers as a distinct reliability risk rather than ordinary commercial load. The scale of the August trip underlines why: 3.8 GW is roughly the size of several large gas plants disappearing from the system in an instant.7
Texas is already moving down the same path. The Electric Reliability Council of Texas was set to vote Tuesday (2026-06-02) on rules requiring data centers to meet new performance standards, part of a broader effort to prevent "cascading" outages if large computational facilities suddenly drop offline.2
The concern is not theoretical. Data centers can vanish from the grid in seconds, and unlike industrial motors or heating loads, they offer no natural ramp-down. ERCOT's proposed rules passed on Tuesday (2026-06-02) created winners and losers in the rush to join what one developer described as "Batch Zero" interconnection slots, but the operator finally has a framework to work with.3
The tension cuts both ways. Data centers want power fast and they want it reliable, but their hardware is highly sensitive to voltage disturbances. Utilities want control because their demand response experience shows they need gradual load reduction, yet data center operators say their equipment cannot tolerate the same curtailment profiles as a residential air conditioner or an industrial smelter.5
As one analyst put it, if the choice is having 80% of a data center working or none of it, operators will take 80%. But the regulatory structures that might allow that kind of partial-load flexibility have not been validated yet, and market certainty will take time to build.5
The stakes for PJM are enormous. The grid operator serves 67 million people from Virginia to Illinois, and its latest capacity auction showed the existing process for securing new generation is unable to keep pace with data center demand growth.6
The numbers explain why. Data centers now account for about half of incremental US electricity demand growth, according to the International Energy Agency's global energy assessment. Globally, data center electricity demand grew 17% in 2025, with AI-focused consumption surging 50%.1
The interconnection queue tells an even starker story. Only 13% of capacity that submitted requests from 2000 to 2019 had reached commercial operations by the end of 2024; 77% had been withdrawn. That is a structural failure of the current process, and it is why PJM is now looking at whether it can build the grid differently.1
The Asia Pacific region is already ahead on the flexibility front. Data center developers there are increasingly required to provide grid services support in exchange for power access, according to Wood Mackenzie research. The regulatory direction is clear: reliability obligations are becoming the price of admission.4
What remains unresolved is whether ride-through standards alone can fix the problem. A data center that trips offline during a voltage sag must also be able to ride through the recovery, and its backup generation must respond within the same window. PJM has not yet specified what its new requirements would look like, only that it is weighing them.7
ERCOT's experience suggests the implementation will be messy. The rules established Tuesday (2026-06-02) will generate appeals, design refinements and likely several rounds of stakeholder fights over who bears the cost of the new equipment and controls. The developers who moved early, however, expressed relief at finally having something concrete to plan around.3
The real test comes when a large data center trips in the middle of a hot summer evening with reserves thin. PJM's 3.8 GW event was a warning shot; the next one may not be. Grid operators are betting that ride-through standards turn a potential cascading failure into a manageable blip, but the hardware that makes that possible is still being built, and the first compliance deadline has not even been set.7,3