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

Summary

This episode dissects PJM, the 13-state-plus-DC grid serving 66 million people, as a capacity market whose auction design can turn small modeling errors into billions in extra costs. Robert Boswell argues that PJM overestimated how much capacity it needed by modeling winter reliability too conservatively and by running auctions with too little lead time for new plants to be built. He estimates $12 billion in avoidable spend across the 2025/26 and 2026/27 auctions, with another potentially expensive emergency auction on the way. The discussion also covers how capacity prices are set, why winterization and plant reliability matter, and why data centers are being blamed for costs they may not actually be driving. The guests frame the core issue as governance and market design: PJM is large, complex, and increasingly unable to translate actual system conditions into efficient procurement.

Notes

• PJM is the regional grid at the center of the episode: 13 states plus DC, with some states only partially included.
• It serves 66 million people and is described as the largest grid by electricity delivered in America.
• The capacity market applies to the whole system, including fully paid-off plants.
• That system design makes small marginal price changes extremely expensive in aggregate.
• PJM delayed auctions and is now running them with very short lead times.
• Lead-time mismatch is a core bottleneck because new generation takes years to build and interconnect.
• Boswell says PJM is inflating demand while undercounting available supply.
• Across the last four auctions, PJM spent $63 billion.
• Boswell estimates $12 billion of that was avoidable.
• He breaks the avoidable spend into $7 billion in 2025/26 and $5 billion in 2026/27.
• Boswell argues capacity auctions should really be run with about three years of warning.
• But PJM has effectively been using only one to two years of warning.
• Winter conditions make plants, especially gas turbines, more efficient.
• PJM’s model is criticized for acting like it is summer all year.
• Its winterization approach is described as backward-looking, based on historical failures.
• Boswell wants winterized plants to be rewarded with more capacity credit and more revenue.
• Capacity is being paid for as reliability margin, not generation.
• At clearing, all participants get the market price, even if they bid much lower.
• Reliability ratings determine how much auction volume a plant gets.
• A well-winterized gas plant can bid as more reliable and earn more.
• Boswell says PJM is overbuying because it does not recognize the reliability of what it already has.
• Data centers are framed as a potentially cheaper use case for electricity, not a more expensive one.
• The logic is simple scale economics: more demand spreads fixed grid costs over more usage.
• The episode acknowledges some real buildout constraints, including wires, copper, transformers, and long lead times.
• PJM’s deliverability studies are criticized as too broad and too slow for a large footprint.
• The emergency auction is triggered by a modeled 6.8 GW shortfall in 2028/29.
• Boswell says better cold-weather and winterization modeling would cut the needed procurement to about 3 GW.
• The board chose a long-term contract design that can run to 2042.
• The emergency contract design carries a price cap of $555 per megawatt-day and could total as much as $21 billion.
• The board overrode the membership-supported design because it feared it would not procure enough power.
• PJM is described as an extremely complex stack of contracts rather than a simple market.
• About 10% of PJM’s system does not participate in the auction, mostly vertically integrated utilities.
• Boswell agrees PJM is too big to function unless governance is simplified or the system becomes more localized.
• The episode argues much of the public electricity-price debate is downstream of PJM issues.
• Batteries and flexibility are presented as an underused route to lower prices.
• Boswell says the high-price period is temporary, not permanently baked in.
• He says the current auctions run only until partway through 2029.
• Boswell argues load growth is the best way to reduce power prices if data centers can justify it.
• For data centers, he lists behind-the-meter power, backup power, batteries, and demand flexibility.
• He is skeptical of data-center demand flexibility, though he says capable teams are working on it.