Ep. 026 - PJM's $12B Modeling Mistake Is Hitting Ratepayers Again (Datacenter, Energy)

Summary

This episode argues that PJM, the grid operator serving 13 states plus D.C., has made major modeling and market-design mistakes in its capacity auctions, leading to an estimated $12 billion in avoidable ratepayer costs across the first two auctions discussed. The central technical critique is that PJM models winter capacity too much like summer capacity, and uses backward-looking outage assumptions that fail to properly credit winterized plants and colder-weather performance improvements, especially for gas turbines. The speaker also says PJM runs auctions with too little lead time for new plants to interconnect, which constrains supply and amplifies clearing prices across a uniform-price auction. Later, the episode turns to an emergency auction and governance fight, with concerns that long-term contracts and data center liabilities could leave existing ratepayers exposed if the expected load does not materialize. The speaker frames the situation as a mix of bureaucratic over-centralization, regulatory capture, and political pressure, while arguing that better modeling and more direct contracting could materially reduce costs.

Notes

• PJM is the regional grid operator for 13 states plus D.C., with only partial coverage in some states.
• The region is described as the largest grid by electricity delivered in the U.S.
• The episode frames PJM as central to data center power pricing concerns.
• Capacity auctions are meant to buy reliability, not actual energy output.
• The market covers the whole system, including existing paid-off plants.
• Because every plant participates, small price changes can have a very large systemwide cost effect.
• The speaker says PJM’s supply curve has shifted left because fewer plants can join while demand rises.
• The headline claim is that $12 billion of spending was avoidable.
• The estimated avoidable amount is said to come entirely from the first two auctions.
• The speaker splits the avoidable amount into $7 billion and $5 billion across the 2025/26 and 2026/27 auctions.
• PJM is said to be slow to connect new generation to the grid.
• The proposed auction lead time is around three years; one to two years is too short.
• Winter improves plant performance, especially for gas turbines because colder air is denser.
• PJM is criticized for modeling winter as if it were summer all year.
• That seasonal mis-modeling can cause PJM to buy too much capacity.
• Winterization and weatherization are the other major capacity-input variable.
• Storm Elliott in 2022 and the 2014 polar vortex are cited as winter outage examples.
• The speaker says PJM’s reliability model is backward-looking and extrapolates past failures.
• A better design would pay more to winterized plants and reduce overbuying.
• The auction procures capacity, measured in megawatts, not generation.
• The market’s purpose is a reliability margin for the worst day of the year.
• PJM’s auction uses a demand curve and a supply curve that intersect to set price and quantity.
• Uniform clearing means all cleared participants are paid the marginal price.
• That uniform-price design means a few marginal megawatts can swing total auction cost sharply.
• The speaker argues nuclear plants are credited around the low-90% range and gas plants around 80% capacity credit.
• Better winterization could justify a higher capacity credit for gas plants.
• The speaker says PJM is too centralized and too large to function well.
• Broad deliverability studies can slow interconnection and encourage transmission overbuild.
• The episode says battery deployment is a major untapped lever for lower electricity prices.
• Data centers are framed as a potentially positive load because they can spread high fixed grid costs.
• The next issue is an emergency auction and a projected future shortfall.
• The speaker says a 6.8 GW shortfall may actually be closer to 3 GW after better winter modeling.
• PJM’s board overrode a membership-backed auction design.
• PJM has over a thousand members split into five voting groups, with a 66% equivalent threshold.
• The board chose long-term contracts extending to 2042.
• PJM is described as a pass-through vehicle with no capital base of its own.
• The proposed load contract could run 11 to 15 years, out to 2042 or 2043.
• The price cap is described as $555 per megawatt-day, implying up to $21 billion in exposure.
• The speaker warns that data centers may not take the liabilities and ratepayers could be left holding the bag.
• The membership-supported alternative would have had data centers take liability directly.
• The speaker says the board rejected that alternative because it feared insufficient power procurement.
• The episode explicitly describes the dynamic as classic regulatory capture.
• The governance structure is portrayed as a dense stack of complex contracts rather than a simple central authority.