On October 15, 2026, the North American Electric Reliability Corporation said its stakeholders had approved three new reliability standards for “computational loads” on a final ballot. It was the first time the rules written to keep the bulk power system reliable have addressed data centers, artificial intelligence facilities, and cryptocurrency mining operations as their own category of grid actor. The package now goes to the NERC Board of Trustees for adoption, expected this quarter; if the board adopts it, NERC will file the standards with the Federal Energy Regulatory Commission for final U.S. approval, with separate coordination with Canadian authorities.
That sentence deserves unpacking, because the grid’s reliability rulebook has always been written for the supply side. NERC’s standards obligate generators, transmission owners, balancing authorities, and the other registered entities that move power around. Loads were, for planning purposes, something close to weather: big, aggregate, and taken as given. A rule that treats a data center as an entity with enforceable obligations toward the grid — obligations to model itself, to coordinate its protection equipment, and to comply with an operator’s instructions — is a genuine institutional turn. This article explains what the three standards would require, why July made them inevitable, and what they leave unresolved.
What got approved, and who it covers
The package — Project 2026-02 — comprises three new reliability standards, targeted revisions to two existing facility-interconnection standards (FAC-001-5 and FAC-002-5) to avoid overlapping compliance obligations, and implementation plans phasing in compliance over months and years.
The standards would apply to computational-load sites of at least 50 megawatts of total connected load, supplied through electrical equipment connected to the bulk power system at 100 kilovolts or above — including sites served from multiple points of connection. The criteria were raised from NERC’s initial 20 MW / 60 kV proposal after stakeholder comment, and multiple buildings can constitute a single site. These rules are for hyperscale-class facilities, not commercial customers. A separate registration track advanced October 8, when NERC’s board approved proposed revisions to its Rules of Procedure creating two new functional entity categories — Computational Load Owner (CLO) and Computational Load Operator (CLOP) — and authorized expedited development of a data request to support registering them, pending FERC approval of the criteria. At multitenant facilities, a centralized or shared operating function could serve as the accountable operator.
The October 15 ballot answers a FERC order issued July 16, 2026 — in Docket No. RD26-7-000, Reliability Standards Pertaining to Computational Load Integration, 196 FERC ¶ 61,031 (2026). FERC acted under Section 215(d)(5) of the Federal Power Act, the provision that lets the Commission direct NERC to file a standard addressing a specific reliability matter. It told NERC to file new or modified standards for computational-load integration by December 31, 2026, revised Rules of Procedure — including registration criteria — by the same date, and a March 1, 2027 informational filing describing the next phase of the work. The ballot result means NERC is on pace to meet the year-end filing deadline. The standards are not yet in force, and will not be until FERC approves them and the phased compliance dates arrive.
CLO-001-1: interconnection studies and modeling
The first standard, CLO-001-1, is about planning. It would require coordinated interconnection studies, commissioning procedures, and planning-model data for qualifying sites. Transmission planners and planning coordinators would study new interconnections and changes to existing sites that meet a planning coordinator’s definition of a “qualified change,” a term the coordinators must define within three months of the effective date.
The modeling requirements address the industry’s most basic problem with these facilities: existing load models do not capture what data center equipment and controls actually do during grid disturbances. Owners would have to provide modeling data covering equipment tripping, ride-through capabilities, and recovery after disturbances — plus information about the amount and expected use of the computing infrastructure itself. They would also have to verify that configurable, site-specific model parameters reflect installed equipment, and provide updated verification documentation within 180 days of changes to hardware, software, firmware, or settings that alter dynamic response.
That last clause is where the standard bites hardest. A data center is not a static load; servers are swapped, UPS firmware is updated, transfer-switch settings are retuned. Under CLO-001-1, a change that alters how the facility behaves electrically becomes a modeling obligation with a deadline. Transmission owners, distribution providers, and generator owners would have six months from the effective date to establish their interconnection requirements; the remaining study, commissioning, and owner obligations follow at nine months.
CLO-002-1: operational data and communications
The second standard is about operations. CLO-002-1 would let transmission operators, balancing authorities, and reliability coordinators specify the operational information they need for planning, real-time monitoring, and reliability assessments, and would obligate computational-load owners and operators to satisfy those specifications. It would also require communications capabilities between facilities and grid entities, with predefined protocols, failure procedures, and training on oral operating instructions.
Its most consequential provision is Requirement R8: computational-load operators would have to comply with operating instructions from their transmission operator, balancing authority, or reliability coordinator — unless the action cannot be physically implemented or would violate safety, equipment, regulatory, or statutory requirements. R9 requires notification when an operator cannot comply. Read plainly, this makes a data center operator answerable to the grid operator the way generators already are. NERC has been careful about one implication: the standard does not automatically convert firm electricity service into interruptible service, and nothing in it changes the service priority agreed upon for computational-load sites.
But the multitenant question lingers. Where a facility operator lacks direct control over tenants’ computing equipment, the standard’s exceptions — inability to physically implement, or violation of safety and equipment requirements — do not expressly recognize contractual limits on the operator’s authority over tenant gear as grounds for noncompliance. The compliance timeline is six months for the grid entities to establish their data specifications, and twelve months for owners and operators to meet the data, communications, instruction-response, and training requirements.
CLO-003-1: protection coordination and disturbance monitoring
The third standard reaches into the equipment itself. CLO-003-1 would require computational-load owners to establish and follow a process for developing protection settings, so that protective systems operate in the intended sequence during faults — including coordination across equipment owned by different entities, with identified coordination issues resolved before the settings are implemented. And it would require disturbance recording: sequence-of-events data for specified circuit-breaker positions, fault recordings, and continuous dynamic disturbance data.
NERC’s reasoning is that ordinary operational telemetry may not capture the rapid load responses and protection-control interactions at these facilities, so higher-resolution recordings would support event analysis, model validation, and disturbance-behavior assessment. Responsibility follows ownership: computational-load owners handle sequence-of-events data for breakers tied to equipment they own, while the owner of the main power transformer supplying the site takes the fault and dynamic recordings. Protection-coordination requirements begin after nine months. The recording requirements carry the longest transition — 36 months for sites already operating when the standard takes effect, and for later sites, 15 months after the effective date or their in-service date, whichever is later.
The July that made this inevitable
Three events, in sequence, turned a research project into a deadline. The first was a Level 3 Essential Action Alert — NERC’s most urgent alert category — issued May 4, 2026 with seven essential actions (responses due August 3) on computational-load modeling, studies, commissioning, protection and control, disturbance monitoring, and operational coordination. The alert rested on incident reviews: a July 2025 NERC assessment documented successive Eastern Interconnection transmission faults in July 2024 that triggered the near-simultaneous loss of roughly 1,500 MW of data center load through customer-side protection, pushing frequency to 60.053 Hz before recovery within four minutes. A January 2026 review examined 26 events involving cryptocurrency mining loads in ERCOT between January 2023 and September 2025; each cut consumption by more than 100 MW, with facilities losing 17% to 95% of pre-disturbance load after normally cleared transmission faults.
The second was the FERC directive itself, on July 16, which concluded that the pace of data center growth left no room to wait on NERC’s voluntary timelines.
The third came six days later. On July 22, at 7:56 a.m., a mechanical failure took a 230-kilovolt transmission line out of service in Northern Virginia, inside Dominion Energy’s territory — home to the highest concentration of data centers on Earth. The fault was normally cleared; the data centers’ response was not: facilities began disconnecting in waves and transferring to on-site backup power without warning. The first drop was 2,970 MW; the total reached roughly 3,800 MW — nearly 4,000 MW, the largest such event in PJM’s history, cutting PJM’s overall load by nearly 4%. PJM and Dominion dispatched generation down and restored balance within nine minutes, inside NERC’s 30-minute standard, with no unresolvable reliability impacts — though the event produced large swings in voltage and system frequency. Previous Virginia events in 2024 and 2025 had each shed about 1.5 GW; July 22 was roughly two and a half times larger. Emanuel Bernabeu, chair of PJM’s Operating Committee, put the industry’s verdict bluntly: the fault was normally cleared, and “these data centers are too sensitive to the kind of voltage that they like.” David Souder, PJM’s executive director of system operations, called it the third measurable large-load disconnection event in two years and said the pattern was “unsustainable.”
The biggest gap: ride-through
The most important omission is one NERC acknowledges openly: the package establishes no industry-wide minimum voltage and frequency ride-through requirements for computational loads. Ride-through — a facility’s ability to stay connected and operating through temporary disturbances rather than disconnecting — is the technical heart of the July 22 event. CLO-001-1 requires modeling data on ride-through capabilities, and interconnection requirements can incorporate ride-through criteria based on planning-study needs, but the rules do not say what disturbances every qualifying facility must ride through. That work is advancing on parallel tracks. NERC’s Large Loads Working Group has developed three additional Standard Authorization Requests covering computational-load disturbance performance, the study of load losses in planning and near-term reliability assessments, and event reporting and model validation. Under FERC’s order, the March 1, 2027 workplan filing will lay out the next phase, and NERC’s standards roadmap lists further potential requirements — disturbance performance, load oscillations, model validation, event reporting, co-located generation, cybersecurity — extending into 2027 and beyond.
The regions are not waiting for phase two. On September 10, PJM proposed its own ride-through requirements for large loads — facilities would have to remain connected through normally cleared transmission disturbances — and said it expects to file tariff changes with FERC in November. The proposal follows a June FERC order directing PJM to identify gaps in its tariff for large-load customers, including ride-through and ramp rates. ERCOT moved first: it implemented ride-through requirements for large data centers and crypto mining facilities on August 1. The emerging shape is federal standards setting the floor while the big grid operators write the detailed performance rules — a pattern this desk will be watching, because it determines whether “computational load” means one compliance regime or five.
What changes for planners and developers
For transmission planners, the change is that large loads become modeled, verifiable entities rather than aggregate assumptions. Planning studies for new interconnections and qualified changes to existing sites will have to account for how a facility behaves during faults — tripping characteristics, ride-through, recovery profile — with owners obligated to keep that data current. For developers and hyperscalers, the compliance surface is expanding from interconnection, siting, and power procurement into reliability compliance, and buildout timelines will increasingly run through someone else’s study queue: MISO’s proposed 120-day large-load study process and SPP’s new CHILLS service for large loads (covered on this desk October 14 and 15) are the regional counterparts of the same trend.
What to watch
Four milestones will determine what this package means in retrospect. First, the NERC board vote, expected in the fourth quarter — adoption is the likeliest outcome given the ballot result, but not certain until it happens. Second, the FERC filing, due by December 31 under the July order, and FERC’s subsequent action: approval makes the obligations mandatory and enforceable. Third, the March 1, 2027 workplan, which will show whether the hard technical questions — minimum ride-through, load oscillations, co-located generation — are moving or stalling. Fourth, PJM’s November tariff filing, the first test of whether the regional ride-through layer harmonizes with the federal floor.
The deeper story is in NERC’s own framing: these loads are “reshaping the demand profile and exposing reliability gaps in both operational and planning time horizons.” NERC’s 2025 Long-Term Reliability Assessment projected summer peak demand growing 224 GW over ten years — 24% above 2025 levels, and 69% above the previous assessment’s 132-GW projection — with AI and digital-economy data centers driving most of it. The reliability framework built for a world where load was passive weather is being rewritten for a world where the largest new loads can vanish in minutes. October 15 was the day the rewrite passed its first real vote.