For most of the history of the North American grid, reliability standards have been written for the supply side. Generators, transmission owners, balancing authorities and planners must register with the North American Electric Reliability Corporation and comply with mandatory rules on everything from frequency response to cyber security. Customers, even very large ones, have sat outside that regime. On July 16 the Federal Energy Regulatory Commission moved to change that for one class of customer. In Docket RD26-7-000, it directed NERC to file one or more new or modified Reliability Standards addressing risks associated with the integration of computational loads, together with any new definitions, by December 31, 2026.
The order also directs NERC to develop changes to its Rules of Procedure, including registry criteria, so that computational load entities can be registered, with those revisions also expected by December 31. By March 1, 2027, NERC must submit an informational filing with a workplan setting out the next steps for additional standards. The Commission accepted that NERC may file in phases, with initial "essential actions" standards, which it calls Phase I, due by year end.
Why data centers are a reliability problem
The case for standards rests on a specific technical behavior. Data centers are sensitive to voltage disturbances. When a fault on the transmission system causes a voltage dip, data center protection systems can transfer the facility to backup power within fractions of a second, removing the load from the grid. One data center doing that is a minor event. Dozens doing it at once, in a region with a high concentration of facilities, is a sudden loss of load on the scale of a large power plant tripping, but in reverse.
NERC documented exactly that in an incident review published on January 8, 2025. At about 7:00 p.m. Eastern on July 10, 2024, a lightning arrestor failed on a 230 kV transmission line, starting a disturbance with six faults. Coincident with it, the same local area saw roughly 1,500 MW of load reduction, and the review found that it was exclusively data center type load. The area had a high concentration of data centers. When NERC talked to the owners, it learned that the facilities had transferred to backup power systems in response to the disturbance. The grid stayed stable, but operators had to respond to a sudden surplus of generation, and the event showed that a growing class of customers can behave like a single large, fast-acting resource that no standard governs.
NERC published a second review on January 9, 2026, on voltage-sensitive cryptocurrency mining load reductions, adding to the evidence that computational loads, a category broad enough to include crypto mining and AI training, share these characteristics.
From action plan to mandatory rules
The FERC order traces the path from the Department of Energy's October 2025 proposal on large load interconnection to this point. NERC filed comments in that proceeding supporting swift Commission action and stating that reliability standards are important for an economy heavily reliant on the electric grid. It described a large load action plan. Then, in supplemental comments on March 20, 2026, NERC said it had concluded that it should accelerate that plan and begin updating registry criteria and developing initial standards that include large loads, with initial new or modified standards filed by December 31, 2026, and further revisions during 2027.
In other words, FERC's order formalizes and makes mandatory a timeline NERC had already proposed. That is significant for two reasons. It removes any doubt about whether the work will proceed. And it ties NERC to dates, which matters because standards development normally takes years, while data center construction is measured in months.
What standards might require
The order does not write the standards, but the incident evidence points to the likely content. The first is ride-through: requirements that large computational loads remain connected through defined voltage and frequency disturbances rather than dropping offline, or that they reconnect in a controlled way. Similar ride-through standards were developed for inverter-based generation after solar and wind plants tripped unexpectedly during faults. The second is data: requirements that large loads provide accurate models of how their equipment behaves during disturbances, so that planners can study them. ERCOT has already made dynamic models a condition of entering its batch study for large loads, which shows how quickly that requirement is spreading. The third is coordination: requirements that load owners communicate with transmission operators about operating changes, backup generation transfers and planned ramps.
Registration criteria will decide who must comply. A threshold set too low would sweep in ordinary industrial customers. A threshold set too high would miss the clusters of mid-sized facilities that caused the 2024 event. NERC will have to define a computational load entity in a way that captures the reliability risk without creating a compliance burden for loads that pose none.
Who bears the cost
Mandatory standards impose real costs. Ride-through capability may require changes to power electronics, uninterruptible power supplies and protection settings. Modeling and data requirements add engineering work. Compliance programs add staff. For hyperscalers with large engineering teams, these are manageable. For smaller colocation operators and crypto miners, they could be more significant. There is also a question of retrofits. Facilities built under current practices may not meet new ride-through requirements, and the standards will need to decide whether to apply only to new connections or to existing loads over a transition period.
The alternative is worse. If computational loads keep growing and keep dropping offline together during faults, the system will need more fast reserves and more conservative operating limits, the cost of which falls on everyone. Making the loads themselves responsible for their behavior is the cheaper path and the fairer one.
The larger shift
The order marks a conceptual shift in how the grid is governed. Large loads have become large enough, concentrated enough and fast enough that they affect bulk system reliability in the same way generators do. The regulatory response is to bring them inside the reliability framework. That will not happen overnight, and the first standards will be narrow. But by the end of 2026, a data center campus may be a registered entity with compliance obligations, not just a customer with a service agreement.