Cover Story · Joule desk
The Price the Auction Won't Print
Firm power's marginal price has moved to the gas-turbine order book
The Ecliptic · Issue No. 003 · July 30, 2026
$325
per MW-day — PJM's capped clearing price, flat for a third auction
$528–555
per MW-day — what PJM's own two estimates say firmness costs uncapped
116 GW
GE Vernova gas-turbine backlog in Q2 2026; slots gone through 2030
≥$190k
per MW, non-refundable, to hold a 2031 turbine slot
The wedge
A capped price meeting an uncapped cost
All figures on PJM's UCAP basis, $/MW-day. The $325 cap is the 2028/29 BRA clearing price, which equalled the FERC-approved cap; $528 is Brattle's indicative Net CONE; $555 is PJM's own simulation of the same auction without the price collar. Hatched bars are the report's bottom-up model estimates for 2029 delivery, not observed prices.
PJM's capacity auction is supposed to reveal what a new firm megawatt costs. Capped for a third consecutive year, it no longer can — and the shortfall is not small.
- The 2028/2029 Base Residual Auction cleared at $325/MW-day, the FERC-approved cap — *below* the prior year's $333.44 — while procuring 6,831 MW less than its own 156,013 MW reliability requirement. A price that falls while the auction misses its target is a suppressed signal, not a scarcity one.
- Two numbers PJM itself owns bracket the real figure from below: Brattle's indicative Net CONE of $528/MW-day, and PJM's own simulation of the same auction without the collar, clearing at $555 — a market cost of $29.7 bn against the $16.4 bn actually charged.
- On those two official estimates alone, the capped auction reveals somewhere between 41% and 62% of the price of firmness. The report's bottom-up model runs higher still, and is best read as the ceiling of the band, not a point estimate.
The venue
The order book is a forward curve in disguise
Illustrative of the report's model, not a market quote. Gross and Net cost of new entry are the report's own estimates by turbine delivery year, built from ~$2,250/kW all-in combined-cycle capex escalated 8%/yr, a 20-year life and 8% WACC, stated on PJM's UCAP basis. The cap line is PJM's administrative $325/MW-day held flat. Near-year levels are an upper case.
Heavy-duty gas turbines are built by three companies, all sold out for the rest of the decade. What they charge by delivery year is the price of firmness, printed weekly whether or not the auction agrees.
- GE Vernova's gas backlog reached 116 GW in Q2 2026, up from 100 GW a quarter earlier, targeting 125 GW of orders plus reservations by year-end — near-term slots gone through 2030, reservations now taken for 2031. New-order pricing ran 10–20 points higher per kW than Q4 2025.
- Siemens Energy is booked to FY2028 with a ~60 GW gas backlog; Mitsubishi Power is sold out through 2028 and doubling manufacturing capacity. Four-to-five year lead times are now standard.
- Annualize all-in combined-cycle capex by delivery year and the cost of a new firm megawatt rises from ~$898/MW-day in 2026 to ~$1,271 in 2031 on a gross basis, while the cap sits flat. The shape of that curve is the claim; the level is an upper bound.
Who pays
The deposit is the market pricing what the auction won't
The cleanest evidence that firmness is worth more than the capped price is that developers are already paying it — years early, and non-refundably.
- Converting interest into a turbine slot requires a deposit of ~20% of the contract price. Against a ~$950/kW HA-class turbine island, that is at least $190,000 per megawatt at risk to hold a 2031 slot — versus a capped capacity payment worth $118,625/MW-yr. Nobody posts the former for the latter.
- The rent is therefore captured twice: by the OEMs, through pricing power on a sold-out book, and by the owners of already-built firm capacity, whose megawatts are now worth the uncapped replacement cost rather than the capped print.
- The non-obvious loser is the load-serving utility and its ratepayers. A capped price does not lower the cost of firmness — it changes who pays it and hides the bill, re-emerging as bilateral scarcity cost, co-location cost-shifts and reliability risk.
- The curve is real but not permanent. Watch slot availability for 2030–31, any FERC or PJM move on the cap, and the first reservation cancellations — each would collapse price discovery back into the auction.
The Ecliptic, Issue No. 003, cover analysis. Auction, backlog and capex figures are drawn from the underlying report's cited sources; all cost-of-new-entry curves are the report's own transparent estimates, stated on PJM's UCAP basis, and are labelled as such.