PJM Interconnection has proposed new data-center ride-through rules intended to keep large computational loads connected during defined grid disturbances instead of abruptly transferring thousands of megawatts to backup generation. The regional grid operator filed the tariff revisions with the Federal Energy Regulatory Commission on November 16 in docket ER27-522-000.
The proposal is not yet an approved rule. It is also separate from PJM’s Fast Energization filing, which would change how new generation and large loads move through the connection-study process. Fast Energization concerns how projects get studied for connection; ride-through concerns how a large load behaves after it is operating.
That distinction matters because a data center can create a reliability challenge not only by adding demand, but also by removing a very large block of demand almost instantly when its equipment detects a disturbance.
What data-center ride-through means
Ride-through is the ability to remain connected while voltage or frequency temporarily moves outside its normal range. Power plants have long faced technical standards that define when they should stay online during a fault. PJM’s proposal would establish comparable requirements for customers taking transmission service to supply a Large Computational Load.
A ride-through rule does not require equipment to ignore every unsafe condition. It creates a defined region of voltage, frequency and time in which the load should continue operating rather than disconnecting or switching to another source. Protection systems may still act outside that region, and a transmission owner may require stricter performance where local conditions justify it.
PJM’s stakeholder presentation on Large Computational Load ride-through places the technical requirements in the tariff and PJM Manual 14B. The customer requesting service would be responsible for ensuring that the computational load meets them.
Why PJM proposed data-center ride-through rules
The operational concern became concrete on July 22, when about 3,800 MW of data-center load in Virginia transferred away from the grid during a transmission event, according to PJM’s review with Dominion Energy. That was the largest large-load transfer PJM had recorded.
PJM said the bulk power system did not experience significant reliability impacts, but the sudden change created an imbalance and voltage and frequency movement that operators had to manage. The lesson is not that the July event caused a grid failure. It is that simultaneous protection settings across many facilities can turn what appears to be a local equipment decision into a regional operating event.
A rapid loss of demand is different from ordinary conservation or planned demand response. Grid operators continuously balance generation and consumption. If several gigawatts of load disappear before generators can reduce output, supply temporarily exceeds demand and frequency can rise. If the facilities later return together, the opposite imbalance can occur.
Backup generators and uninterruptible power systems protect computing equipment, but an uncoordinated transfer can shift the reliability problem onto the larger network. PJM’s proposed rules are meant to make that response more predictable.
What PJM’s filing would require
PJM’s November 16 filing proposes tariff revisions covering ride-through, other operational requirements and annual transmission-service reporting for Large Computational Load. A PJM compliance matrix shows the ride-through provisions in Part III and Part XI of the tariff.
The design places responsibility on the network customer or eligible customer arranging service for the large load. In practice, compliance may involve facility protection settings, power-electronics controls, uninterruptible power supplies, backup-generation transfer logic and coordination with the transmission owner.
Readers should not infer that every existing data center would immediately receive identical new settings. The final obligations depend on what FERC accepts, the effective tariff language, the applicable service agreement and the technical manuals used to implement the standard.
The filing also does not complete PJM’s work on how quickly a large load can increase or decrease demand. PJM’s compliance matrix says it plans to develop ramp-rate requirements using a study to be conducted in 2027. Ride-through addresses the immediate response to a disturbance; ramp-rate rules would address how load changes over time.
How this differs from a faster grid connection
PJM filed the ride-through proposal alongside a broader package of large-load reforms. That timing makes it easy to blend several distinct policies together.
The Fast Energization proposal would coordinate studies for new generation and large loads and aims to reduce the projected study period. It does not set the electrical behavior of a data center during a fault.
Ride-through rules do not accelerate construction or guarantee a connection date. They establish operating performance that a connected facility may need to demonstrate. A project can therefore face both questions: when can it connect, and how must it behave once connected?
Future & Fact’s guide to the four grid-access milestones explains why a study request, completed study, signed service agreement and physical energization are separate forms of evidence.
What to verify in a data-center power claim
A company announcement that mentions backup power, batteries or flexible demand does not by itself establish ride-through compliance. Project documents should eventually identify the requested service, the responsible utility or transmission owner, applicable protection standards and any operating limitations.
Useful questions include:
- Which PJM transmission service and utility agreement cover the facility?
- Are voltage and frequency ride-through settings specified?
- When may the load transfer to backup generation, and how quickly can it return?
- Can the facility reduce demand under instruction, and is that flexibility contractually defined?
- Has FERC accepted the relevant tariff language, or is it still proposed?
Those checks complement Future & Fact’s guide to verifying power-supply claims for an AI data center.
What happens next
FERC can accept the filing, reject it, require revisions or allow parts to take effect under applicable procedures. PJM will also need to translate accepted tariff provisions into manuals, agreements and implementation steps.
Until that process is complete, the accurate description is that PJM has proposed ride-through rules for Large Computational Load. It has not imposed a final nationwide standard, and the filing does not prove that every data center currently responds to disturbances in the same way.
The proposal nevertheless marks a shift in grid planning. Large data centers are no longer being treated only as customers that consume a great deal of electricity. Their second-by-second operating behavior is becoming part of the reliability record.


