On Monday, October 26, the North American Electric Reliability Corporation’s Board of Trustees adopted the first mandatory reliability standards written for the load that is remaking the grid: data centers, AI compute clusters, and cryptocurrency mining facilities, collectively called computational loads. The package now goes to the Federal Energy Regulatory Commission for final approval. Nothing in it is enforceable yet, and nothing in it buys a megawatt of capacity or allocates a dollar of cost. But it is the first time the reliability framework has treated a class of customer as a grid actor in its own right, with obligations to model itself, to talk to its operator, and to record what it does under stress.
Start with the count, because the first reports of the vote were easy to misread. The package contains five standards: three new ones — CLO-001-1 on interconnection studies, modeling, and verification; CLO-002-1 on operational data and communications; and CLO-003-1 on protection coordination and disturbance monitoring — plus revisions to two existing facility-interconnection standards, FAC-001-5 and FAC-002-5. The standards passed their final industry ballot on October 14, and the Board adopted them at a special meeting twelve days later. NERC says it will file the package with FERC; as of publication the filing has not yet been made, and FERC’s order directing the work sets December 31 as the deadline.
The standards apply to a defined slice of the load wave. A computational load site — one location, or several buildings treated as one site — is covered when its total connected load is at least 50 MW and it is supplied through equipment connected to the bulk power system at 100 kV or above. That threshold was raised during development from an original proposal of 20 MW at 60 kV, narrowing the rules to the largest facilities. The obligations fall on both sides of the interconnection: on the computational load owners and operators, and on the planning coordinators, transmission planners, transmission owners and operators, balancing authorities, and reliability coordinators they connect to.
CLO-001-1 is the planning standard, and it is the one that will change interconnection studies first. It requires coordinated interconnection studies, commissioning procedures, and planning model data for each covered site. Owners must provide data on how their equipment trips, rides through, and recovers during disturbances, and must document model verification showing that the site-specific parameters represent the equipment and settings actually in service. As balloted, verification documentation must be updated within 180 days of a hardware, software, firmware, or setting change that alters the site’s dynamic response, with initial verification due within 365 days of the in-service date. The FAC revisions align transmission owners’ interconnection requirements and studies so the same obligation is not imposed twice.
CLO-002-1 governs the operating relationship once a site is energized. It sets data and communications requirements between the site and its transmission operator, balancing authority, and reliability coordinator, and requires operators generally to follow operating instructions unless doing so is physically impossible or would violate safety, equipment, regulatory, or statutory requirements — with notice to the issuing entity when they cannot comply. The standard also requires communications capability with predefined protocols and personnel training.
CLO-003-1 is the forensic standard. It requires a process for developing and coordinating protection settings across entities, plus three kinds of recording: sequence-of-events recording for specified circuit breakers, fault recording, and continuous dynamic disturbance recording, with responsibility assigned by equipment ownership. The drafts posted for comment specified the technical detail — input sampling of at least 960 samples per second, dynamic recording of at least 60 observations per second, time synchronization to UTC within plus or minus one millisecond, and data in IEEE C37.111 (COMTRADE) format, provided within 15 calendar days of a request. Those figures come from the drafts and should be confirmed against the final text FERC approves. The direction is unmistakable: after a disturbance, the system will have a record of what a large load did.
Registration is the other half of the framework, and it is already at FERC on a separate track. On October 21, NERC filed a petition in docket RR27-1-000 seeking approval of revisions to its Rules of Procedure that would add two new functional entity categories to its compliance registry — Computational Load Owner and Computational Load Operator — and requested expedited action. Comments and motions to intervene are due November 20. The petition answers who must register; the standards answer what registered entities must do. Neither has effect until FERC approves it.
The reason all of this is moving on a compressed schedule is documented in NERC’s own incident record. In May, NERC issued a Level 3 Essential Actions Alert — its highest-severity category — after several incidents in which data centers unexpectedly dropped load or oscillated demand rapidly, threatening the stability of regional grids. The alert set out seven actions grid entities must implement while permanent rules are created. NERC had already published incident reviews on January 8, 2025, and January 9, 2026, the second on voltage-sensitive cryptocurrency mining load reductions. The standards are, in substance, the permanent version of the alert: the same risks, converted from urgent instructions into auditable requirements.
The schedule is not NERC’s own. On July 16, FERC issued an order in docket RD26-7-000, 196 FERC ¶ 61,031, directing NERC under section 215(d)(5) of the Federal Power Act to file one or more new or modified reliability standards addressing computational-load integration by December 31, 2026, along with related glossary changes and registry-criteria revisions. The order calls the October package the Phase I “essential actions” standards and directs NERC to submit an informational filing with a work plan for additional standards by March 1, 2027. FERC’s order noted that NERC “has documented multiple grid disturbances in which computational loads have caused or contributed to the instability of the Bulk-Power System.” The Board’s October 26 vote is NERC meeting a federal deadline two months early.
The slow half of the response
That deadline discipline is worth pausing on, because the large-load story this autumn has mostly been told in the language of procurement and price. PJM’s Reliability Backstop Procurement — refiled October 29 to buy the 6,831 MW its July capacity auction failed to procure, at a $555/MW-day ceiling and 15-year terms — is a purchase: a one-time, expensive, litigated answer to a forecast-driven shortfall that analysts now estimate may have shrunk to roughly 630 MW. The NERC standards are the other half of the response, and they work on a different clock. A procurement buys capacity for a delivery year. A standard rewrites what it means to connect load to the system, permanently, for every site that meets the threshold, in every region.
The distinction matters because the two halves solve different problems. The backstop exists because the forecast was wrong — the three-year forward auction bought for load that has since been canceled or delayed, and the market could not summon supply at the price the collar allowed. The standards exist because the load itself behaves differently from anything the planning models were built for: it can drop hundreds of megawatts in seconds, oscillate, and recover on timers the operator cannot see. A capacity market can pay for enough megawatts and still fail if the load it serves trips off and back on faster than the reserves can respond.
They do not, however, answer the questions the procurement fights are actually about. The standards say nothing about who pays for the transmission and generation a large load requires, nothing about whether a data center should bear the cost of the capacity its forecast created, and nothing about flexible service — the idea, now central to FERC’s show-cause proceedings, that a large load could accept interruptible or curtailable service in exchange for faster interconnection. Those are commercial and jurisdictional questions, and they are being fought in tariff proceedings, not in standards development.
What the standards would actually require
For the entities they bind, the practical effect is a new compliance program with a phased clock. As balloted, obligations begin on staggered schedules after the effective date that follows FERC approval: planning coordinators define qualified changes and modeling data specifications within three months; transmission owners, distribution providers, and generator owners update interconnection requirements within six months; the remaining study, commissioning, and owner obligations under CLO-001-1 follow at nine months. CLO-002-1 gives operating entities six months for data specifications and owners and operators generally twelve months. CLO-003-1 sets nine months for protection coordination, fifteen months for recording at new sites, and thirty-six months for recording at existing sites. The full package phases in over roughly three years.
That phasing is also the standards’ most honest feature: they concede how much of the industry is not ready. Many sites do not have the recorders, relays, and historians CLO-003-1 contemplates, and many do not have dynamic models that have ever been verified against their own equipment. The 180-day reverification rule and the 365-day initial verification window are, in effect, a multi-year project to build the evidence base that planners need — and the evidence base that the next disturbance investigation will demand.
There is a limit worth naming plainly. As drafted, model verification checks that a model’s parameters match the design and settings of the equipment. It does not require the model to be compared against how the facility actually behaved during a disturbance — and some of the parameters that matter most, such as UPS transfer thresholds, disturbance counting and lockout logic, and recovery timers, can be hard to verify from design data alone. The drafts allow owners to state which parameters could not be verified. That gap is why the recording requirements matter as much as the modeling ones — but it also means the standards, as adopted, are a framework for producing evidence, not a guarantee the evidence will be used to correct the models. That correction is part of the Phase II work NERC says is underway.
What comes next is already underway
Phase II is not a vague promise. NERC’s announcement says additional standards development is underway to address computational load performance, planning and reliability studies, and event reporting and model quality — the harder questions the foundational package deliberately deferred. A new project, 2026-03, has opened a 30-day informal comment period on a standard authorization request for disturbance performance, closing November 20, the same day comments are due on the registration petition. The March 1, 2027, work plan filing to FERC will set out the rest of the sequence.
Meanwhile the commercial track is running on its own deadline. On June 18, FERC issued six show-cause orders under section 206 of the Federal Power Act — one to each RTO and ISO, in dockets EL26-67 through EL26-72 — preliminarily finding their tariffs unjust and unreasonable because they do not adequately address large-load interconnection, and identifying five categories of reform: application and study processes, cost transparency and cost-shifting protections, co-location and behind-the-meter generation, new transmission services for flexible large loads, and studies of generation serving electrically proximate large loads. FERC granted abeyance in all six dockets on August 14, and responses are now due November 16 for five RTOs and November 20 for SPP.
The two tracks converge on the same underlying facts. NERC’s long-term assessment put summer peak demand growth at 224 GW over ten years, a 24 percent increase from the 2025 peak, driven overwhelmingly by the load the standards now cover. PJM’s experience this year — a forecast that justified a 6.8 GW backstop and then fell 6.5 GW in a single round of utility revisions — shows what happens when the commercial machinery has to price load it cannot yet see clearly. The standards are, among other things, an attempt to fix the seeing: if planners have verified models and disturbance records for every 50 MW site on the system, the next forecast revision has a better chance of being right, and the next auction has a better chance of buying what is actually needed.
That is the claim to test as the package moves through FERC. The Commission can approve the standards, approve them with modifications, or remand them. What is already fixed is the direction: data centers and other computational loads are now a distinct reliability category, with their own standards, their own registry, and their own compliance clock. The question for the next two years is whether that category produces better planning — or merely better-documented surprises.
What to watch
November 16: show-cause responses from five RTOs, and the first test of whether the operators’ tariff reforms answer FERC’s five reform categories. November 20: comments due on NERC’s registration petition in RR27-1-000, and on the Project 2026-03 disturbance-performance authorization request — the same day SPP’s show-cause response is due. December 31: NERC’s deadline to file the standards package with FERC. Then the Commission’s review, the effective date, and the phased clocks — three months to three years — that turn adopted standards into enforceable obligations. And on March 1, 2027, the work plan that says what Phase II will require, and when. The headlines this autumn belonged to the procurement. The standards are the part that will still be binding when the procurement is a footnote.