We Were Early
on the Capacity Story.
Here's What We
Got Wrong.

01The Setup

Most market intelligence is forward-looking by design — predict what's coming, recommend what to do. The honest version requires occasionally looking backward to audit the track record. This is one of those.

In late 2023, our briefings began flagging the PJM capacity setup as mispriced. The argument: reserves were tightening structurally (data center load, EV demand growth, electrification), retirements of dispatchable resources were accelerating (coal, older gas), and new build interconnection queues were lengthening. Supply response was lagged.

Our conclusion at the time: 'The capacity market is mispricing reserve adequacy. The auction-to-auction price could rise sharply.' We expected gradual repricing over 2-3 auctions. We were wrong about that.

02The Data

What we got right: the directional thesis. Reserve tightness, retirement pace, demand growth — all played out as we expected. The structural drivers were correctly identified eighteen months in advance. The customers who acted on the early signals positioned themselves well.

What we got wrong: timing. We expected the price increase to be gradual (50-100% over 2-3 years), not 9x in a single auction. The market doesn't always price in known information gradually — sometimes it repriices all at once when a threshold is crossed.

What we got right and wrong simultaneously: customer impact prediction. The bill-shock conversation we said would happen is happening now (see BR-15). But we underestimated how much suppliers would price in expected future tightness, not just current tightness — that's what produced the magnitude surprise.

03The Implication

Being right about direction and wrong about magnitude is a useful failure mode. Better than being wrong about both. Worse than being right about both. The lesson isn't that the underlying analysis was bad — the lesson is that markets can reprice on a faster timeline than gradual-adjustment models assume.

For customers who took 2023 action on the early signals: they look prescient. For customers who waited for certainty: they're now paying the price of caution. The decision wasn't certain at the time — uncertain decisions made directionally well still beat correct waits.

04The Recommendation

1. Reserve margins in adjacent markets. MISO, NYISO (see BR-06), ISO-NE are showing similar structural setups. Same mispricing risk? We think yes; we're tracking, and we're more humble about timing this time.
2. Generator response to high prices. Will high clearing prices catalyze new build at scale, or will queue/permitting/economics still constrain supply response? Watching closely. Without supply response, the high-price regime persists.
3. Customer behavior shifts. Are large customers actually changing operations in response to capacity prices? Or is it pass-through? Behavior change would soften the next auction; pass-through wouldn't.
4. The 2027/28 auction window. This is when new generation could materially affect clearing, if the response works. The early data is mixed — some progress, mostly delays. We'll update views as evidence arrives.

Forecasting is hard. The directional analysis was solid; the timing was off. The honest version of intelligence work is admitting both. The customers who acted on the early signals are in better shape than the customers who waited for certainty. The customers who haven't acted yet have less margin for delay than they did in 2023. We'll keep updating views as evidence arrives — and we'll keep flagging when the views update.