$CEG $VST $TLN $NRG
EXECUTIVE SUMMARY
https://t.co/9TH3D5hcZ6
Monitoring Analytics’ “Analysis of the 2027/2028 RPM Base Residual Auction Part A” frames the 2027/2028 PJM capacity outcome as the intersection of a large, data-center-driven demand shock with a supply stack that was unable to clear the target reliability requirement even at an administratively constrained maximum price. The auction’s defining quantitative facts are: 1) a material cleared-capacity shortfall versus the RPM reliability requirement; 2) clearing at the administratively defined maximum price across the footprint; and 3) a very large portion of total capacity-market revenues explained by embedded and, more importantly, forecast data center load embedded in the 2025 load forecast used for the demand curve. The report’s core contribution is a set of controlled counterfactuals that isolate the impacts of (a) the restricted Variable Resource Requirement (VRR) curve introduced via the Pennsylvania Governor–PJM agreement (maximum and minimum price “collar”) and (b) the inclusion/exclusion of data center load from the peak-load forecast, holding the underlying offer stack constant.
Under actual 2027/2028 BRA results, total RPM market revenues were $16,411,578,225 at the restricted maximum price of $333.44/MW-day UCAP, with cleared UCAP (generation and DR) of 134,478.1 MW and a reserve margin of 14.4% versus a 20.0% IRM, implying a shortfall of 6,516.6 UCAP MW versus the RPM reliability requirement (and 8,452.4 ICAP MW versus the IRM in ICAP terms). The report estimates that removing the restricted VRR curve (i.e., reverting to the filed/accepted “unrestricted” VRR curve without the agreement’s cap/floor) would have increased 2027/2028 RPM market revenues to $26,324,850,846, a +$9,913,272,621 change (+60.4%) versus actual, with cleared UCAP rising modestly to 135,271.0 MW (+0.6%). The MMU’s alternative “IMM VRR curve” (lower maximum defined as the lower of Gross CONE and 1.5x Net CONE; no minimum price) produces an intermediate revenue estimate of $19,676,240,030 (+19.9% versus actual) with cleared UCAP of 135,187.4 MW (+0.5%).
The incremental capacity-revenue impact attributed to data center load is large even under the restricted VRR curve: excluding 17,071 MW of embedded plus above-embedded data center load from the 2027 peak forecast reduces RPM market revenues to $9,913,924,713, implying an incremental $6,497,653,512 (+65.5%) attributable to including that data center load in the forecast used for demand. However, the report emphasizes that the agreement’s minimum price materially compresses the measured incremental impact in these restricted-curve counterfactuals by holding clearing prices above the levels that would otherwise occur at lower demand; under an unrestricted VRR curve without a minimum price, excluding the same 17,071 MW reduces revenues to $6,876,999,976, implying an incremental $19,447,850,870 (+282.8%) from including data center load. This divergence is central: the restricted VRR curve simultaneously suppresses upside in scarcity (via the cap) and props up revenues in non-scarcity counterfactuals (via the floor), reducing both the level of revenues under scarcity relative to the “unrestricted” construct and the apparent incremental effect of data center load when comparing to a low-demand baseline.
AUCTION OUTCOME AND RELIABILITY IMPLICATIONS
The 2027/2028 BRA outcome is characterized by a systemwide scarcity condition rather than localized LDA constraints, under the administrative collar. PJM’s public BRA materials state that the 2027/2028 auction cleared at $333.44/MW-day UCAP and that all modeled LDAs cleared at the same price, indicating no constrained LDAs under the auction’s zonal constraint modeling. The PJM BRA report also notes that the RTO failed the market structure test (a 3-pivotal supplier test), triggering mitigation for existing generation resources, yet clearing still occurred at the capped maximum price, consistent with a binding administrative ceiling rather than a market-clearing price discovered on a downward sloping portion of the demand curve.
The MMU’s reserve margin decomposition highlights the severity of the reliability shortfall within the RPM-cleared portion of the system: forecast peak load ICAP of 164,579.0 MW, RPM peak load ICAP of 152,262.1 MW (after subtracting FRR peak load ICAP of 12,201.9 MW and PRD ICAP of 115.0 MW), and a forecast pool requirement factor (FPR) of 0.9260 imply an RPM UCAP reliability requirement of 140,994.7 UCAP MW. Cleared UCAP (generation and DR) of 134,478.1 UCAP MW yields a 6,516.6 UCAP MW shortfall, while the implied reserve margin of 14.4% sits 5.6 percentage points below the 20.0% IRM when measured as “reserve cleared in excess of IRM” in ICAP terms. PJM’s own reporting frames the shortfall at 6,623.2 UCAP MW versus the RTO reliability requirement when FRR procurement is included, and flags that the shortfall triggers an investigation under the tariff’s reliability backstop framework, with potential escalation if shortfalls persist across subsequent BRAs.
A critical nuance for interpreting the shortfall is the role of the price cap versus underlying supply scarcity. PJM reports that 809.6 UCAP MW did not clear because offers exceeded the capped maximum price; in a simulated rerun without the cap/floor, cleared UCAP rises by 792.6 MW to 135,270.7 UCAP MW, consistent with the MMU’s scenario result that removing the cap would have pulled a limited tranche of above-cap offers into clearing. The residual shortfall remains large even after adding back this increment, implying that the majority of the shortfall is not explained by offers marginally above the cap but by insufficient eligible/offered capacity relative to requirement, retirements, derates, and/or constraints that prevented sufficient supply from being available at any price within the auction construct.
DEMAND-SIDE DRIVER: DATA CENTER LOAD AND LOAD FORECAST STEP-CHANGES
The MMU report positions data center load as the dominant driver of the recent step-change in PJM peak-load forecasts and, by extension, the rightward shift of the RPM demand curve. Historical actual peaks are presented as relatively stable over 2014–2024 (not exceeding 150,000 MW and averaging 144,493 MW), followed by a new peak of 156,256 MW in June 2025. The 2025 long-term load forecast then projects 2027 summer peak load of 164,186 MW, an increase of 6.5% from 2025 to 2027, representing a marked departure from prior forecast vintages that implied sub-1% annual growth.
The “large load adjustment” component of the forecast is identified as the mechanism by which data center demand is incorporated. In the 2025 load forecast report, PJM projected large load additions plus growth from existing large load facilities of 8,453 MW in 2026, of which 7,892 MW (93%) is attributed to data center load, and 13,668 MW in 2027, of which 13,018 MW (95%) is attributed to data center load. The MMU’s decomposition of data center load distinguishes “embedded” load (existing at the time of the forecast) from “above embedded” (forecast growth in existing facilities plus new facilities). For the 2027/2028 BRA, the report states 13,018 MW of above-embedded data center load and 17,071 MW of total embedded-plus-above-embedded data center load in the 2027 peak forecast used for the auction, implying 4,053 MW of embedded data center load.
This embedded/above-embedded split is economically and risk-relevantly important. Embedded load is structurally “in the system” and tends to be treated as a baseline demand obligation for capacity procurement. Above-embedded load is fundamentally forecast-based and therefore subject to material uncertainty around timing, realization rates, and geographic distribution. The MMU explicitly treats forecast uncertainty as a core problem when such large adjustments are inserted into a 1-shot auction design that sets multi-year forward capacity prices and obligations, and notes that the locational distribution of the 17,071 MW is across multiple LDAs and is accounted for in the scenario analyses.
SUPPLY-SIDE CONDITIONS: STATIC OFFER STACK, LIMITED ENTRY, AND STRUCTURAL TIGHTNESS
The MMU’s counterfactual framework holds the offer stack and non-demand parameters constant to isolate the effect of demand curve changes and load forecast adjustments. That design choice makes the scenarios highly interpretable but also highlights an important real-world caveat: if market rules and demand curve parameters differed ex ante, offer behavior, entry decisions, and retirement reversals would likely differ materially.
PJM’s BRA report indicates the supply side did not expand nearly enough to meet the increase in reliability requirement. Net new entry is reported at 774.0 UCAP MW, and offered capacity increased by 955.8 UCAP MW while reliability requirement increased by 6,295.4 UCAP MW; cleared RPM capacity increased by only 370.6 UCAP MW. This asymmetric dynamic is consistent with a market in which near-term supply elasticity is low due to interconnection constraints, permitting lead times, gas deliverability constraints, and the compressed auction schedule (the 2027/2028 BRA was held roughly 18 months before the delivery year rather than the historical 3-year forward cadence). A low-elasticity supply response implies that even moderate forecast errors or discrete demand shocks can produce very large price and reliability outcomes, particularly when the demand curve is relatively steep near the reliability requirement.
The PJM report’s statement that the auction failed the 3-pivotal supplier test is also notable for market-power and risk considerations: structural market power conditions can coexist with scarcity, and mitigation can constrain offers, but if demand intersects supply on a capped segment, the binding constraint becomes administrative rather than purely competitive. This is consistent with the MMU’s emphasis that it is misleading to interpret recent outcomes as “ordinary” supply-demand equilibria; the presence of a binding maximum price and a non-economic minimum price are direct market design interventions that change both level and distribution of outcomes.
VRR CURVE DESIGN: RESTRICTED VS UNRESTRICTED VS IMM
The report’s most policy-relevant content is the quantification of how the restricted VRR curve altered outcomes relative to both the tariff-based unrestricted VRR curve (Quadrennial Review 2023 parameters) and the MMU’s proposed IMM VRR curve. The restricted VRR curve introduces both a maximum price and a minimum price for 2026/2027 and 2027/2028, derived from the Pennsylvania Governor–PJM agreement. For 2027/2028, the maximum is $333.44/MW-day UCAP and the minimum is $179.55/MW-day UCAP, calculated by converting the agreement’s $325 and $175 UCAP values into ICAP terms using a dual-fuel CT ELCC and then reconverting using an updated ELCC of 0.77. The MMU notes the maximum is $30.34/MW-day (8.3%) below its stated recommended maximum of 1.5x Net CONE ($363.78/MW-day) for Rest of RTO.
Mechanically, the restricted VRR curve changes the demand curve in 2 ways that matter for interpreting both realized and counterfactual outcomes.
Scarcity regime (supply intersects demand at low quantity): a lower maximum price reduces the clearing price and total revenues when scarcity drives the auction to the upper flat segment. In 2027/2028, both the MMU’s scenario analysis and PJM’s own simulation indicate the auction would have cleared at a much higher maximum without the cap (Rest of RTO at $529.80/MW-day and DOM at $542.83/MW-day), implying the cap was materially binding and suppressed revenues by roughly $9.9B.
Non-scarcity / lower-demand regime (supply intersects demand near or to the right of the reliability requirement): the minimum price prevents clearing prices from falling below $179.55/MW-day even if supply surplus would otherwise drive a lower clearing price. The MMU explicitly states this minimum price is unique, high, and lacks an economic rationale in the record presented, and that it compresses the apparent incremental effect of reduced demand in the data center counterfactuals by forcing a higher clearing price than would occur absent the floor. Regulatory skepticism about the economic logic of a minimum price is also reflected in Commissioner See’s dissent, which criticizes the floor as uneconomic and potentially harmful to market confidence, even while supporting the cap as a transitional measure.
The MMU’s IMM VRR curve construct is an attempt to replace both the high cap and the floor with a maximum tied to the lower of Gross CONE and 1.5x Net CONE, and with a $0 price at the maximum quantity point, reintroducing a more conventional scarcity-pricing framework without a hard floor. The IMM curve therefore sits between the restricted and unrestricted curves in scarcity pricing (for Rest of RTO, $363.78 versus $333.44 versus $529.80) and is materially lower than unrestricted maxima in zones where Gross CONE is high.
QUANTITATIVE DECOMPOSITION OF RESULTS
The report’s scenario design enables a clean decomposition of revenues into 2 orthogonal “levers”: (a) demand curve collar (restricted vs unrestricted vs IMM) and (b) data center load inclusion (embedded + above embedded; above embedded only; none). The results have several implications that are directly relevant for forward capacity price distributions, policy risk, and asset valuation sensitivity.
Effect of the restricted VRR curve (cap/floor) holding load forecast constant:
Actual revenues of $16,411,578,225 rise to $26,324,850,846 (+$9,913,272,621; +60.4%) under the unrestricted VRR curve, with cleared UCAP rising from 134,478.1 MW to 135,271.0 MW (+792.9 MW; +0.6%). The modest MW effect relative to the revenue effect implies the policy intervention primarily changes price (and hence rents) rather than materially changing physical procurement volumes. PJM’s simulation is directionally identical: removing the cap/floor increases clearing prices to $529.80/MW-day across all zones except DOM at $542.83/MW-day, and increases cleared capacity to 135,270.7 UCAP MW. This corroboration reduces model risk around the magnitude of the cap’s binding effect for 2027/2028 under static offers.
Relative to the IMM VRR curve, actual revenues are lower by $3,264,661,805 (16.6% lower than IMM scenario), again with only a modest change in cleared MW (+709.3 MW under IMM versus actual). This indicates the restricted cap is below the MMU’s own preferred demand curve maximum and that, under scarcity, the restricted design depresses the price signal even relative to a more conservative “investment-signal” benchmark.
Effect of forecast (above-embedded) data center load under the restricted VRR curve:
Removing 13,018 MW of above-embedded data center load reduces revenues to $10,167,054,398, a -$6,244,523,827 change (-38.0%), while keeping embedded load in the baseline. The implied conclusion is that the forecast portion of data center demand, rather than the already-connected embedded portion, drives most of the demand-side pressure in the 2027/2028 construct.
Removing the full 17,071 MW of embedded plus above-embedded data center load reduces revenues to $9,913,924,713, a -$6,497,653,512 change (-39.6% versus actual). The small incremental difference between these 2 counterfactuals ($10,167,054,398 versus $9,913,924,713; a $253,129,685 gap) is an explicit artifact of the minimum price: the floor prevents clearing prices from falling as much as they otherwise would when demand is reduced, compressing the incremental revenue impact attributable to embedded data center load.
Effect of data center load under an unrestricted VRR curve (no floor; higher cap):
Under unrestricted VRR, removing above-embedded data center load yields $8,878,470,594, while removing total data center load yields $6,876,999,976. Under this construct, adding total data center load increases revenues by $19,447,850,870 (+282.8%), and adding above-embedded load increases revenues by $17,446,380,251 (+196.5%). The scale of these increments illustrates the non-linear nature of capacity-market cost allocation: a demand shock does not only add capacity procurement in proportion to the incremental load; it can raise the clearing price applied to a very large base of procured MW, thereby producing systemwide rent transfers. This is the primary reason why a 17,071 MW forecast inclusion can map to a $19.448B revenue increment under an unrestricted scarcity regime.
Effect of data center load under the IMM VRR curve:
Under the IMM VRR curve, removing above-embedded load reduces revenues to $8,360,373,447 and removing total data center load reduces revenues to $6,416,591,187. The incremental revenue attributable to including total data center load is $13,259,648,843 (+206.6% versus the no-data-center IMM baseline). The IMM curve materially reduces the scarcity rents versus unrestricted but still implies a very large revenue sensitivity to data center load inclusion, reflecting that the underlying scarcity condition remains substantial even under a more conservative maximum price.
Impact on cleared UCAP MW across scenarios:
Changes in cleared UCAP are small under demand curve changes alone (+0.5% to +0.6% when moving from restricted to unrestricted/IMM), but larger when data center load is removed (approximately -0.9% to -1.0% under restricted; -2.6% to -4.9% under unrestricted; -2.6% to -4.8% under IMM). The directionality is intuitive under static offers: lower demand reduces clearing prices and can cause higher-offer units to not clear, reducing cleared MW even as requirement falls. From a reliability standpoint, the MW deltas alone are not sufficient to infer shortfall severity under each counterfactual without the corresponding demand requirement, but the pattern confirms that price outcomes are far more sensitive than physical procurement volumes.
COST-ALLOCATION AND “WHO PAYS” DYNAMICS
The report repeatedly frames the revenue increments as “increases in costs to other PJM customers,” reflecting the capacity-market settlement mechanism in which load-serving entities fund capacity payments and new loads increase system requirements. The MMU’s cumulative estimate across the 2025/2026, 2026/2027, and 2027/2028 BRAs is $23,100,955,341 of incremental capacity-market revenues attributable to data center load inclusion, with the 2027/2028 component at $6,497,653,512. The year-by-year pattern shows the 2027/2028 incremental effect is smaller than 2025/2026 and 2026/2027 in the MMU’s accounting, which the report attributes to the restricted VRR curve’s minimum price and a higher baseline price environment even in counterfactual scenarios without data center load (including a higher reliability requirement for MAAC in 2027/2028).
From an economic perspective, the report highlights a structural issue: the marginal capacity cost imposed by large data center additions is highly convex because it moves the system along a scarcity curve where prices can jump to administrative maxima. In such a regime, traditional “cost causation” principles (marginal load pays marginal cost) are challenged because the capacity charge applied to a new large load can be orders of magnitude larger than the direct incremental resource additions needed to serve it, since the load shift drives a systemwide price. The MMU’s policy suggestion to require new data center load to bring new generation is an attempt to internalize this externality by converting a systemwide price effect into a more directly attributable self-supply obligation.
INTERPRETATION OF “DATA CENTER DRIVES THE MARKET” AND THE ROLE OF FORECAST UNCERTAINTY
The MMU’s conclusion that data center load growth is the “primary reason” for the tight supply-demand balance, the cleared-capacity shortfall, and high prices rests on the magnitude and recency of forecast revisions and the scenario results showing large revenue deltas when data center load is removed. The strength of this conclusion depends on 2 implicit premises:
The large-load/data-center adjustments are the dominant driver of peak forecast growth over the auction horizon, and absent them, peak load would have remained on a much slower growth trajectory. The tables on historical peak stability and the forecast step-changes support this premise.
The supply side would not have reached scarcity conditions rapidly absent the demand shock; instead, the system would have had time for market-driven adjustments (queue reforms, new dispatchable entry, demand response adaptation, delayed retirements) to occur before scarcity rents hit the maximum. The MMU argues the trajectory would have been slower and prices would rise but not “jump suddenly to the maximum price.”
A key analytical risk is that the above-embedded component (13,018 MW) is forecast-based and therefore inherently uncertain. The report treats this uncertainty as a critical design flaw when combined with a single forward-clearing mechanism. PJM itself notes in its press release that peak demand forecasts for 2027/2028 are expected to be revised downward and references the role of announced retirements potentially returning and other factors that could mitigate the shortfall before the delivery year. The presence of such statements in PJM’s own materials supports the MMU’s emphasis on forecast uncertainty, even if the magnitude of future revisions is not quantified here.
REGULATORY AND MARKET DESIGN TRAJECTORY INTO 2028/2029
The report emphasizes that the Pennsylvania Governor–PJM agreement’s cap/floor is temporary and will not apply to the 2028/2029 BRA, implying that the same underlying scarcity drivers could translate into materially higher prices and revenues if the demand shock persists and supply response remains constrained. PJM’s own simulation for 2027/2028 already provides a concrete anchor for the potential price level absent the collar: $529.80/MW-day in Rest of RTO and $542.83/MW-day in DOM, with corresponding revenue of roughly $26.3B.
Separately, PJM’s ongoing VRR curve periodic review process appears poised to adjust the maximum prices and shapes for 2028/2029 and beyond, with reported proposals implying a PJM-wide cap near $550/MW-day and zonal caps spanning approximately $483/MW-day to $785/MW-day, and an auction start date of June 30, 2026 for the 2028/2029 BRA. This introduces a second layer of price-risk asymmetry: even if the agreement’s cap/floor disappears, the “unrestricted” curve that would apply in 2028/2029 may differ from the 2027/2028 unrestricted parameters due to updated CONE estimates, revised reference resource assumptions, and new curve parameters.
The regulatory debate over the minimum price is also relevant for longer-dated risk. Commissioner See’s dissent articulates a view that a minimum price is economically unsound and undermines investor confidence, suggesting that even if transitional measures are adopted in response to political pressure, they may not be durable or may be challenged on economic grounds. The MMU’s report similarly asserts that a minimum price lacks an economic rationale and that its inclusion changes the interpretation of data center impacts by forcing higher prices in low-demand counterfactuals. The combination implies elevated regulatory uncertainty around the lower bound of capacity prices (particularly in scenarios where forecast load fails to materialize) as well as around future interventions should price outcomes be politically contentious.
