PJM’s next capacity auction, the Base Residual Auction for the 2029/30 delivery year, closes December 15. It is the fourth and final auction under the price collar, and the first conducted after the Federal Energy Regulatory Commission’s October 9 order split PJM’s large-load package in two. FERC accepted the 50 MW Large Load definition and the registry that goes with it, effective October 12, and rejected the two frameworks that were supposed to do something with that visibility: the Interim Resource Adequacy Service and Bring Your Own New Capacity.
That split is the frame for December. The auction will test three things at once: whether registry visibility changes how load is forecast and bid, whether the suspended Reliability Backstop Procurement — the parallel channel meant to buy roughly 6.8 GW the regular auction could not — leaves a hole that bidding behavior fills or widens, and whether another year of load-forecast growth pushes the shortfall further out of reach of the supply that can actually clear. None of those questions has a price answer yet. All of them will be visible in the auction’s parameters before the clearing price is ever announced.
The starting point: a shortfall that grew
Start with the baseline December inherits. The July auction for the 2028/29 delivery year cleared at $325/MW-day across the entire footprint — the collar’s cap, for the third consecutive auction — and still procured only 138,318 MW of unforced capacity, plus 10,864 MW under Fixed Resource Requirement arrangements. Against a reliability requirement of 156,013 MW, that left a shortfall of 6,831 MW and a reserve margin of 14.7 percent against a 20 percent target. It was the second consecutive auction in PJM history in which the region as a whole missed its reliability requirement; the 2027/28 auction had missed by roughly 6,500 MW.
The new supply the price summoned was, as this desk has noted, a rounding error against that gap: about 317 MW of new generation and about 208 MW of upgrades to existing resources, roughly 525 MW combined. A one-year commitment at a capped price cannot finance a power plant, and three auctions at the cap have now demonstrated that proposition three times.
December’s target is larger. PJM’s planning parameters for 2029/30 set the reliability requirement at 158,712 MW, 2,699 MW higher than the year before, with a forecast peak of 171,916 MW, 3.6 percent above the prior auction’s 165,954 MW. To meet the new requirement, procured capacity would have to rise roughly 9,530 MW above the 2028/29 total of 149,182 MW. Between the last two auctions, offered supply rose by about 3,447 MW. Closing a gap nearly three times the recent annual supply increase would require a break in the pattern, not a continuation of it.
What the collar does to December
The collar continues through this auction, and that matters more than the level of the cap itself. PJM’s planning parameters for 2029/30 put the cap at $337.83/MW-day and the floor at $181.91/MW-day — figures that reflect the collar’s indexation rather than the approximately $325 and $175 levels described when FERC approved the extension in April 2026. The mechanism is the same: a ceiling that has bound every auction it has governed, and a floor that has never been tested, because the market has never been long enough to fall toward it.
The practical consequence is that December’s clearing price is, within a range, already known. If supply remains short of the requirement — and the requirement has risen while the supply pipeline has not demonstrably widened — the auction will again clear at or near its cap, and the demand curve’s slope will again be irrelevant. The interesting question is not whether the price hits the ceiling. It is what the auction reveals underneath a price that cannot move: the size of the shortfall, the quantity of new supply that clears, and whether any locational delivery area separates from the system-wide price. In July, zero areas were constrained and every zone cleared at exactly $325. A flat map for a fourth time would confirm that the collar has extinguished locational signals entirely; a separated zone would tell planners where the constraint actually binds.
There is a second consequence worth stating plainly. The collar is time-limited, and December is its last auction. The first auction with no collar will be for the 2030/31 delivery year, on the normal three-year-forward schedule PJM is working back toward. December’s results report, with its uncapped simulation, will be read as a briefing document for that post-collar design debate, not just a record of 2029/30.
What the registry changes, and what it does not
The October 9 order, 197 FERC ¶ 61,019 in docket ER26-3515-000, is best read as a visibility decision, not a market-design decision. The accepted pieces — a Large Load definition capturing any end-use load of at least 50 MW at a single site, with affiliated sites within one mile aggregated, plus a Large Load Registry — give PJM and market participants a structured record of the loads driving the forecast. FERC’s stated reasoning was that the definition balances insight into substantial loads against burdening smaller ones, and that the registry will assist load forecasting, transmission planning, and operational management.
Visibility is not nothing. The load forecast is the auction’s demand side: the reliability requirement is built from it, and the forecast’s credibility has been disputed, with critics arguing that speculative pipeline load inflates the requirement and supporters arguing that under-forecasting large loads is the greater reliability risk. A registry populated with reported large loads gives PJM a firmer basis for distinguishing committed load from pipeline speculation — and gives intervenors, state commissions, and the market monitor a dataset to test the forecast against. If the December forecast proves more disciplined than its predecessors, the registry will deserve some of the credit. If it does not, the registry at least makes the argument evidence-based rather than rhetorical.
But the order explicitly did not give PJM the tools that would have converted visibility into obligation. IRAS would have required load-serving entities serving new large loads either to designate new capacity covering the load’s peak or to accept curtailment during capacity shortfalls. BYONC would have let new large loads bring their own new capacity into the market under locked accreditation. FERC rejected both, finding that aspects of the proposal intruded on retail rate regulation — the filed tariff appeared to place obligations directly on eligible new large loads and to set a compensation framework resembling a retail rate — and that PJM had not shown why a customer’s size or vintage should matter to resource adequacy when the capacity market has historically bought capacity for all network load alike. The rejection of the demand-curve exclusion for eligible new large loads, which would have applied beginning with 2029/30, is the piece with the most direct December consequence: PJM had sought the October 12 effective date precisely so it could change the demand-curve parameters for this auction. That change is now off the table. December runs on the existing curve.
Commissioner David Rosner concurred, warning against a market that reports adequate capacity while shedding load involuntarily, and stating that in an emergency, data centers not serving critical functions should be curtailed before homes, hospitals, and schools. That is a statement about emergency priorities, not a tariff rule — curtailment decisions remain with states and distribution utilities — but it signals where the Commission’s sympathy lies for any refiling, and it frames the political question December’s results will feed: if the auction again comes up short, who is curtailed first, and under whose authority?
The suspended backstop hangs over the bidding
The second shadow over December is the Reliability Backstop Procurement. FERC’s September 29 order, 196 FERC ¶ 61,245, accepted the core of PJM’s one-time procurement — including the $555/MW-day volume-weighted maximum willingness to pay — but suspended it for five months, to February 28, 2027, and opened a paper hearing on cost allocation, transmission-owner exit rules, and collateral requirements for load-serving entities. The backstop was designed to buy the roughly 6.8 GW the regular auction could not, on 15-year contracts, at a price the regular auction is not allowed to reach.
Its suspension changes December’s bidding incentives in ways that deserve more attention than they have received. A generator deciding whether to offer into December faces two channels with different prices and timelines: a one-year commitment at the collar’s cap, and a possible 15-year backstop contract at up to $555, whose cost allocation and collateral rules are unresolved and whose procurement may not happen at all. A resource that can wait — a life extension, an uprate, a mothballed unit weighing return — has a reason to hold its best offer for the backstop rather than commit it to the auction. If December’s new and returning supply looks thin, part of the explanation may be rational waiting, not physical absence.
PJM has said it intends to file a revised backstop proposal by October 29 — a stated intention, not a FERC-imposed deadline. The content of that filing, and whether FERC finds it sufficient to hold the paper hearing in abeyance, will determine whether the backstop exists as a real second channel when December’s results are digested, or remains a suspended promise. Watch the filing’s cost-allocation section most closely: FERC’s core objection was that allocation must reflect updated load forecasts capturing large-load growth rather than a snapshot of today’s load, which is the same cost-causer-pays question now being litigated at the Pennsylvania commission and in FERC’s June Section 206 proceedings. The backstop’s allocation rule, if accepted, will become a template for the broader fight.
Who pays is still the open question
That brings December back to the question this desk keeps returning to, because every other question in PJM’s capacity market is a version of it. The market monitor’s data for the first half of 2026 put data center load’s contribution to wholesale cost at $11.11 per MWh through the capacity market alone, excluding any effect on energy or transmission. The monitor’s longer accounting puts the cumulative data-center-driven increase in capacity costs across the 2025/26 through 2028/29 auctions at roughly $29.4 billion. Those figures are contested — developers and some states dispute the attribution — but they define the stakes of the allocation fight that December’s results will intensify.
There is also a quieter distributional story inside the auction that the headline price obscures. Load hedged through self-supply or bilateral contracts is not exposed to the clearing price, and PJM has noted that such load eats first in a rationed market. December’s results report will show how much capacity cleared under Fixed Resource Requirement arrangements outside the auction — 10,864 MW last time. Watch whether the FRR share grows: growth there is a vote of no confidence in the auction, cast by the participants with the resources to bypass it.
What to watch in December
Four things, in the order they will become visible. First, the auction parameters themselves: the demand-curve shape PJM publishes, the collar levels actually applied, and any locational constraints that survive the flat-price era. The parameters are the design; the price is the residue.
Second, the registry’s first footprint. PJM owes a compliance filing within 30 days of the October 9 order — day 30 falls on Sunday, November 8, making Monday, November 9 the practical date — and the filing’s reporting mechanics will show how quickly registry data becomes usable. Whether any of it informs the December forecast is a separate question; the auction’s load forecast was substantially set before the registry existed. The registry is a 2030/31 story wearing a 2029/30 timestamp, and it should be judged on that horizon.
Third, the supply response: how much new generation and how many upgrades clear, against last time’s roughly 525 MW, and how much of the clearing supply is returning capacity versus genuinely new steel. Turbine backlogs measured in years — GE Vernova reported a 116 GW gas-turbine backlog in the second quarter of 2026, and Siemens Energy has cited lead times of three years or more — mean December cannot conjure new plants. It can only reveal how much existing and near-complete capacity is willing to commit at the capped price.
Fourth, and most consequentially, the post-collar debate that December’s results will trigger. The collar expires after this auction. The governors who supported it, the states that opposed its premises, the developers who say it cannot finance entry, and the consumer advocates who say the uncapped price is unaffordable will all read the same results report and draw different conclusions. The honest reading will hold two facts at once: the cap held the 2028/29 clearing price at $325 against PJM’s uncapped simulation of about $555/MW-day, and the market it capped still missed its reliability requirement by 6,831 MW. December will update both numbers. It will not reconcile them. That reconciliation is the design work for the first auction with no collar, and it is the work the next year of this market will be judged on.—
Sources: PJM, 2028/2029 Base Residual Auction results (July 14, 2026) and 2029/2030 planning parameters; PJM, “FERC accepts proposal to extend price collar” (April 30, 2026); FERC, 197 FERC ¶ 61,019, docket ER26-3515-000 (Oct. 9, 2026); FERC, 196 FERC ¶ 61,245 (Sept. 29, 2026); PJM, RPM Auction Schedule; Monitoring Analytics, State of the Market, January–June 2026; prior Fracmoo coverage: “The Market Is Trying to Say $555. The Rules Only Let It Say $325.” (Oct. 15, 2026).