The most important sentence in FERC’s October 9 rejection of PJM’s large-load package was not written by the Commission. It was written by Commissioner David Rosner, concurring, and it reads like a headline the trade press would not have dared to write: “homes, hospitals, and schools should never be curtailed before data centers.”
The order itself, 197 FERC ¶ 61,019 in Docket No. ER26-3515-000, was unanimous — all five commissioners, Swett, Rosner, See, Chang, and LaCerte, with no dissent — and its holding was jurisdictional. PJM’s Interim Resource Adequacy Service (IRAS), which would have let new large loads connect without matching new capacity in exchange for accepting curtailment first in emergencies, was rejected along with the mandatory new-capacity designation, the Bring Your Own New Capacity (BYONC) rules, and the exclusion of uncovered new loads from the capacity demand curve. What survived was the definition of a Large Load — end-use customer load of 50 MW or more at a single electrical site, with affiliated facilities within a one-mile radius counted as one — and the Large Load Registry, both effective October 12, with a compliance filing due within 30 days of the order.
But concurrences, in practice, are where a commissioner tells you what the order did not quite say. Rosner’s concurrence does something unusual: it agrees with the rejection, then writes the exam any refiling will be graded against. PJM spokesman Jeffrey Shields has said the grid operator intends to refile in a way that meets FERC’s requirements while “honoring the concept that existing electricity customers should not be unfairly burdened,” with no timeline given. Whenever that second attempt arrives — and PJM has told FERC it plans Section 205 filings in the separate large-load show-cause proceeding by early to mid-November — it will be measured against the test Rosner laid down. That test has four parts: jurisdiction, sequencing, discrimination, and price honesty.
The jurisdictional line
The first part of the test is the order’s holding, and Rosner’s concurrence assumes it rather than relitigating it: federal tariffs govern wholesale markets; states govern retail service and retail rates. PJM’s transmittal letter said it would tell electric distributors only how many megawatts to shed, leaving customer selection to the states. FERC found the tariff did something else — it appeared to place obligations directly on the Eligible New Large Loads themselves, which are for the most part retail customers, to dictate which specific class of retail customers gets curtailed, and to set up a compensation framework FERC read as establishing a retail rate with money collected from other retail customers. A federal tariff that decides which retail customers are switched off first has crossed the Federal Power Act’s most basic boundary.
This is why Rosner’s assignment of the curtailment-order question to the states is not merely a preference but a jurisdictional conclusion. “We need state regulators to establish clear rules for which customers should be curtailed first when circumstances require it,” he wrote, and the order itself told the states they have “a crucial role to play in this challenge by designating which retail customers must be curtailed when the system is stressed.” For PJM’s second attempt, the line is now bright: the grid operator may give distributors a number, but the moment the tariff tells distributors which retail customers to cut, it is writing a state tariff in federal ink. Any refiling that wants to retain some form of priority curtailment must keep the retail dispatch decision on the state side of the line — and get thirteen state commissions to actually write the rules.
The sequencing test
The second part of Rosner’s test is the one most likely to surprise PJM’s designers, because it turns on the ordering of emergency tools rather than their existence. IRAS would have begun acting after PJM dispatchers issued a NERC Energy Emergency Alert Level 1 (EEA1), directing load reductions from eligible new large loads generally before pre-emergency demand response was deployed. Rosner called this a flip of the accepted paradigm: “flips this paradigm, proposing to force IRAS load off the system in the earliest stages of emergency conditions, before PJM deploys demand response.” Demand response customers, he reminded everyone, are the customers paid to volunteer for exactly this purpose. A market that pays for voluntary flexibility and then cuts involuntary load before calling on the volunteers has its priorities inverted.
The sequencing objection pairs with a reliability-standards problem the order flagged. Under Reliability Standard EOP-011-4, firm load may be involuntarily curtailed only after the reliability coordinator initiates an Energy Emergency Alert Level 3 (EEA3). PJM’s proposal would have involuntarily curtailed certain firm load earlier, at EEA1, which FERC said appears to ignore the standard’s plain language. The Commission’s own guidance for a second attempt pointed toward a voluntary non-firm service — a customer agreeing in advance to interruptible power — an approach that leaves alone the service of customers who signed up for firm supply. Rosner endorsed the sequencing principle directly: mandatory measures “should only be taken when necessary to keep the system reliable, after voluntary demand response resources have been exhausted.”
Put together, the rule for the refiling is: voluntary first, mandatory last, and never mandatory before the standard says mandatory is permitted. PJM can design a service that pays large loads for interruptibility and deploys it early — that is, after all, what demand response is. What it cannot do is force specific retail loads off the system at EEA1 while the volunteers who were paid to be first are still standing by.
The discrimination question
The third part of the test is the one the order found most under-evidenced, and the one that will demand the most work from PJM’s economists. FERC found that PJM had not offered sufficiently detailed evidence that treating load-serving entities differently based on their retail customers’ characteristics — size, vintage, whether they brought new capacity — is not unduly discriminatory. PJM’s capacity market has historically bought capacity for all network load alike. A regime that sorts loads by size and arrival date and curtails one class first needs a record showing why those characteristics matter to resource adequacy, and FERC found that record wanting. It rejected the New Large Load definition without prejudice — the invitation to try again, with evidence.
This is where the second attempt will be graded hardest, because the underlying facts are genuinely difficult. The case for sorting loads is that the loads PJM is sorting are qualitatively different from ordinary load growth: they are large, lumpy, ramping fast, and — in PJM’s telling — driving an imminent shortfall that began with the 2027/28 delivery year. But “different” is not the same as “demonstrably different in a way that justifies differential treatment under the Federal Power Act.” PJM will need to show, with data rather than adjectives, why a 50-MW load that arrived in 2027 may be curtailed first while a 50-MW industrial load that arrived in 2019 may not — and why the distinction tracks resource adequacy rather than revenue convenience.
The accreditation question sits alongside it. The BYONC rules would have locked in a resource’s capacity accreditation — its credited contribution to reliability — for the life of its arrangement. FERC found that lock-in insufficiently justified, and the concern is structural: accreditation is the market’s judgment about what a resource is actually worth on a hot afternoon, and freezing it for BYONC resources while the rest of the market’s accreditation evolves would create a two-tier system with no demonstrated basis. Any refiling that offers loads a way to escape curtailment by bringing their own supply must price that supply with the same accreditation the market applies to everyone else, or justify in detail why the BYONC class deserves a different one.
The demand-curve warning
The fourth part of Rosner’s test is the passage that has traveled farthest in the trade press, and it deserves the attention. IRAS would have excluded uncovered new large loads from the capacity market’s demand curve starting with the 2029/30 delivery year. Rosner warned of what that exclusion would do to the market’s price signal: “The result could be a capacity market signaling that PJM has more than enough capacity while the region sheds thousands of megawatts of demand on an involuntary basis. That’s not only economically inefficient, but also likely to prevent supply and demand from coming back into balance.”
The honest version of this point is that excluding load from the demand curve does not make the load go away. The excluded loads still draw power in normal conditions and still get cut when conditions tighten — they are real demand wearing an invisibility cloak. Remove a large and growing class of load from the forecast, and the curve reports that less capacity is needed than the physical system actually requires. Prices soften, investment signals weaken, and the market’s one job — translating scarcity into a price that calls forth supply — degrades precisely when scarcity is most acute. In a market already clearing at its $325/MW-day cap with 6,831 MW of shortfall, an honest demand curve is one of the few instruments left that is not already broken. A second attempt that revives any form of demand-curve exclusion will have to answer Rosner’s question: why should the market be told to underprice the very loads it may later be forced to shed?
What a defensible second attempt looks like
Rosner’s concurrence does not just veto; it sketches. The shape of a survivable refiling is visible in the four tests, and it looks substantially different from what PJM filed on August 13.
First, it would be voluntary where IRAS was mandatory. The Commission’s own guidance points toward a non-firm transmission service in which a load agrees in advance to interruptible power — the approach it raised in the large-load show-cause proceeding — rather than a federal tariff imposing curtailment priority on retail customers. The bilateral market is already moving this way: Constellation Energy’s 890 MW nuclear uprate deal with Google, announced October 6 as a direct response to PJM’s bring-your-own-power proposal, has now outlived the proposal it answered. Opt-in interruptibility is the IRAS idea translated across the jurisdictional line.
Second, the curtailment order would be written by states, not PJM. Ohio has already shown what state action looks like: the Public Utilities Commission’s take-or-pay rule for AEP Ohio’s largest data-center customers, approved in July 2025, requires loads above 25 MW to pay for at least 85 percent of reserved power monthly for up to 12 years — and speculative interconnection requests fell away once reserving power came with a price. Rosner’s test assigns the emergency-sequencing rule to the same layer of government: state regulators write which customers are curtailed first, PJM operates within those rules.
Third, any bring-your-own-capacity mechanism would have to justify its economics with primary data — accreditation that moves with the market, not a lock-in; discrimination defenses built on a record that ties size and vintage to resource adequacy; cost causation that lands on the loads driving the shortfall. The December Base Residual Auction for the 2029/30 delivery year, which will now model all forecast demand after FERC rejected the exclusion, will be the first real-world test of whether an honest demand curve can begin closing a gap two capped auctions could not.
The calendar for all of this is compressed. The 30-day compliance filing in this docket falls on the week of November 9, PJM has told FERC it plans Section 205 filings in the show-cause proceeding by early to mid-November, the backstop refiling is expected by October 29, and the December auction runs regardless. Two weeks into October, PJM’s large-load strategy has lost one of its two spans — the curtail-first mechanism rejected, the backstop procurement suspended. What remains is the registry, which PJM now must populate with the project-level visibility commissioners asked for, and Rosner’s concurrence, which has quietly become the design brief. The test is whether PJM can write a tariff that a commissioner who believes homes, hospitals, and schools should never be curtailed before data centers can nevertheless vote to accept. That is a narrower question than the one PJM asked in August. It is also the only one that matters now.