NERC Reliability Standards and Large Data Center Loads: Ride-Through, Load Loss and Registration

The North American Electric Reliability Corporation writes and enforces the mandatory reliability standards for the bulk power system, subject to approval by the Federal Energy Regulatory Commission. Those standards have always applied to generators, transmission owners, balancing authorities and other registered entities. Load, the customer side of the meter, was assumed to be diverse and well behaved: thousands of motors, lights and appliances that do not all react the same way at the same instant.

Large data centers break that assumption. A single campus can draw hundreds of megawatts, and many campuses near each other use similar electrical designs that respond identically to a disturbance. NERC’s incident review of the 2024 Virginia event noted that the grid has not historically experienced simultaneous load losses of this size in response to a fault, and that operators had not anticipated it.¹ The grid is planned to survive the sudden loss of a large generator; a sudden loss of load of similar size pushes voltage and frequency the other way, and planners had few models for it.¹

That is why NERC formed a Large Loads Task Force, why grid operators are rewriting interconnection requirements, and why developers should treat reliability performance as part of the large-load interconnection process rather than an internal design matter. Utilities’ own view of these loads is covered in our guide for utilities.

At about 7 p.m. Eastern on July 10, 2024, a lightning arrestor failed on a 230 kV transmission line in Northern Virginia. The line’s auto-reclosing scheme tried three times from each end, producing six successive faults in 82 seconds before the line locked out.¹ Each fault caused a brief voltage dip. Data centers along the system saw repeated dips, and their protection logic moved them off the grid: roughly 1,500 MW of load, all of it associated with data centers, disconnected across 60 points and 25 substations.¹⁶

Frequency did not rise enough to cause concern, but voltage rose, and operators had to act to bring it back within normal limits.¹ The load did not come back on its own; many facilities stayed on backup power until staff transferred them back, which took hours in some cases.¹⁰

The 2026 event was larger. At 7:56 a.m. on July 22, 2026, a mechanical failure on a 230 kV line in Ashburn caused about 3,800 MW of large load to disconnect, more than 3 GW and roughly 3% of PJM demand at the time.² PJM reported a measurable frequency change but no bulk power system reliability impact, and Dominion said the data centers’ own control systems, not the utility, initiated the transfer.²¹¹

Similar events have occurred in Texas. Cryptocurrency mining loads in ERCOT were involved in 25 load-loss events from November 2023 onward, each between 100 MW and 400 MW.⁶

Ride-through is the ability of a facility to stay connected and keep drawing power through a short voltage or frequency disturbance instead of disconnecting. Generators have long been required to ride through specified disturbances. Data centers have been designed for the opposite goal: protect the IT equipment at all costs, which means jumping to the UPS and generators at the first sign of trouble.

NERC’s review described the common logic. A UPS takes over instantly when the grid disturbs. Many facilities also count disturbances, and a typical setting is three within one minute; once that count is reached, the site transfers its load off the grid and stays off until operators move it back manually.¹¹⁰ That is sensible for one building. Repeated across dozens of buildings on the same lines, it turns an ordinary cleared fault into a gigawatt-scale load loss.

Fixing this is mostly a design and settings question: UPS and transfer schemes that tolerate defined voltage dips, coordinated protection, and controlled, staggered reconnection. Some approaches add on-site battery storage or power electronics to buffer the grid connection. Our guide to microgrids and battery storage covers those options.

The response has moved from study to mandate in about two years. NERC published its incident review in early 2025.¹ Its Large Loads Task Force issued a white paper in July 2025 describing the new class of loads (AI data centers, cryptocurrency mines, hydrogen electrolyzers and electrified industry) and the risks they pose.³ On May 4, 2026, NERC issued a Level 3 Essential Action alert covering computational load modeling, studies, instrumentation, commissioning, operations, protection and control; it was only the third Level 3 alert in NERC’s history.⁷

The seven essential actions ask registered entities to develop modeling requirements for computational loads, improve system studies, update what counts as a “qualified change,” set up a commissioning process, enhance protection and coordination, install dynamic fault recorders, and establish communication with the load.⁴ NERC said the urgency came from load reductions of more than 1,000 MW happening in seconds, a timescale comparable to a large generator trip.⁴

FERC then made it mandatory. Its July 16, 2026 order directs NERC to file new or modified standards, glossary changes and Rules of Procedure revisions, including registry criteria for computational load entities, by December 31, 2026, plus an informational filing on the next phase by March 1, 2027.⁵

Registration is the most consequential change for owners. Until now, a data center dealt with its utility, and the utility dealt with NERC. Under the proposed criteria, a large computational load would itself register and carry compliance obligations. The spring 2026 draft described a Computational Load Entity as a load of 20 MW or more, connected at 60 kV, with more than 1 MW of IT load; comments were due May 15, 2026.¹² Those thresholds were proposals and could change before FERC approves final criteria.

Registration of qualifying entities is expected in 2027, with further standards development that year and compliance obligations taking effect only after FERC approval.⁵ Most hyperscale and AI campuses would clear a 20 MW threshold easily, so owners should expect registration, compliance programs and audits to become part of operating a large site.

Grid operators are not waiting for national standards. In Texas, ERCOT’s NOGRR 282, “Large Computational Load Ride-Through Requirements,” adds frequency and voltage ride-through requirements for large loads. It was designated a board priority in December 2025, recommended by ERCOT’s board on June 2, 2026, approved by the PUCT on July 9, 2026 and effective August 1, 2026.⁹ That sits alongside the interconnection and curtailment rules covered in our guide to Texas Senate Bill 6.

In PJM, the 2026 event put ride-through on stakeholder agendas, and Dominion and PJM jointly reviewed the large-load transfer.¹¹ For sites in Virginia and elsewhere in the RTO, see our guide to PJM data center site selection. Other utilities are writing ride-through, modeling and telemetry requirements into their large-load interconnection rules and electric service agreements.

1. 01Ask the utility which ride-through, modeling and telemetry requirements apply now, and which it expects to add once NERC standards are filed.
2. 02Have the electrical engineer confirm the UPS, transfer and protection settings can meet those requirements, and document the settings for the utility’s load study.
3. 03Plan for commissioning tests and data sharing (dynamic models, fault recorders, communications) as conditions of energization.⁴
4. 04Budget for a compliance program if the site will exceed the proposed registration thresholds.¹²
5. 05Consider whether flexible load or on-site storage could help the site meet new rules and earn faster service.

For landowners, none of this changes whether land can hold a data center, but it does change what a buyer will ask the utility, and it adds months of technical work in the energization timeline. If you want a view of how a parcel’s grid position fits these rules, you can get a site reviewed.