PJM and ERCOT are navigating a capacity rollercoaster

PJM’s problem is a timing mismatch rather than a mystery. From the 2008 financial crisis until the last year or two there was essentially no demand growth, so capacity cleared at $30 to $50 per megawatt-day and new plants were built only to replace retiring ones or to satisfy state renewable mandates and corporate procurement. Then demand growth arrived, in Segal’s framing almost overnight in power-market terms, and recent auctions have cleared a little over $300 against a cap. Asked whether that is a failure of market design or of foresight, Segal answers only the second half: the best minds watching this market did not anticipate the speed, and a large gas plant takes four to five years from recognizing the need to delivering power. The signal is unambiguous now, with two consecutive auctions at the cap, but the signal cannot compress construction time.

So the interesting question is what can respond fast, and the answer is uncomfortable. Demand response is the quickest resource, yet participation did not grow between auctions despite roughly ten times the value on offer. Segal offers two partial explanations rather than one: the opportunity has not been socialized to a wide enough set of customers, and the effective load carrying capacity adjustment means demand response never sees the $300-plus gross price. He also expects churn, with some participants leaving once they are actually called on and others entering for the money. Meanwhile the proposal from the White House and governors to make large data center loads pay for the supply and integration costs they create rests on two things in his account: these loads are extraordinarily large and new generation costs have doubled or tripled in a short planning horizon, so allocating those costs to the customers creating them spares everyone else, and without that allocation the loads lose their social license. The mechanics are undetermined and headed for a stakeholder process, but he expects it to push hyperscalers toward bilateral procurement of resources with the least impact on everyone else. For a 2027-2028 shortfall, he says the only viable capacity may be demand response, some batteries, or adding backup fuel to existing gas plants to raise their accredited capacity.

Texas supplies the twist. Going into the recent winter storm the market braced for a repeat of Winter Storm Uri, forward prices ran well over $1,000 per megawatt-hour, and then nothing happened. Segal’s reading is that the high forward price was not a failed prediction but the mechanism itself: it told generators to secure fuel and be ready, and told retailers to call their large customers with an incentive to cut. Across both PJM and ERCOT, more than 10 gigawatts of load that the weather said should appear simply did not. He calls it miraculous, and it is the strongest evidence in the episode that price-driven flexibility is now a real planning resource. But the same fact cuts the other way for anyone who invested in scarcity. If the market has learned that 10 gigawatts will step aside, the probability of sustained peak pricing falls, and with it the spreads merchant batteries earn. Segal confirms that directly: the event was a disappointment for merchant storage, the forward signal turned into oversupply and weak spread clears, and the air is coming out of the balloon. ERCOT has become cyclical, with the battery ancillary services market having gone from lucrative to oversupplied, and day-ahead arbitrage not currently paying for new batteries while forward markets do not support new gas on market pricing alone. His resolution is a maybe: large loads may start signing longer-term arrangements to hedge their own risk, which would fund development and pull the market back toward equilibrium.