Firm power under bottlenecks

The short version

Firm power is scarce, and the equipment to build more of it is booked years out. So the market has stopped buying on lifetime cost. It buys on delivery date instead. The buyers who can lock a turbine slot, a transformer or a whole developer build first. Everyone else waits, and they pay for the wait in one of three ways: a premium for second-tier hardware, older plants kept running longer, or a share of grid costs that the largest loads no longer carry. If you sell into this market, a date you can stand behind is worth more than a lower price.

The mechanism

The mechanism is a timing mismatch. Large loads want firm power within a few years. Almost every path to new firm capacity takes longer than that. Interconnection queues run for years. Heavy-duty gas turbines are sold into the end of the decade. Large power transformers depend on a thin supply of specialty steel and skilled labor.

When time is the scarce input, three things follow:

• Procurement inverts. The buyer's real requirement becomes "can it arrive by my date?" Price, efficiency and even technology preference come second.
• Capital structures change. A 20-year contract assumes patient debt and time to underwrite. Buyers who can't wait own the asset or the developer outright, or pay to reserve a slot before there's a project to put in it.
• Old capacity stops retiring. The cheapest firm megawatt is often the one already standing. Plants slated to close get extended by utilities, grid operators or federal order, and someone pays to keep them running.

Behind-the-meter gas, turbine reservations, refurbished transformers and 1970s peaking barges all answer one question: what can show up in time?

How it developed

• April: the hardware becomes the story. My April 2 post on transformer waits argued that whoever can pay to jump the line decides who gets power. The same month, the largest loads started going around the queue (see the large-loads thread).
• May–June: gas wins on calendar, not cost. My May 8 post on the turbine-blade bottleneck argued buyers were choosing on deployment schedule, not fuel economics. May 11 covered batteries bought to steady gas turbines. June 16 argued that turbine slots, not chips, control the AI buildout.
• May–September: coal persistence, read correctly. May 26 separated coal capacity from generation: China builds coal as insurance, the US keeps plants the market had retired. July 13 covered TVA keeping coal to 2039 while expanding gas, which buys speed and leaves ratepayers carrying the term. August 9 covered New York's 1970s barge plants, ordered to run through 2029. September 19 covered the court ruling against DOE's retirement overrides.
• July–October: buyers adapt. July 7 covered Crusoe's bet on an uncertified turbine. August 27 traced transformers back to electrical steel. September 10 covered GE Vernova sold out to 2031. October 3 warned equipment suppliers about customer concentration. Diesel tightness (October 6) raises bridge-generation costs but doesn't change who gets equipment first.

Where it stands now

• Turbines are still sold out, and reservations are turning into orders. GE Vernova's Q3 release (SEC 8-K, Oct. 28, 2026) shows gas backlog plus slot reservations rising from 116 to 122 GW in the quarter. 13 GW of reservations converted to firm orders, and the company expects at least 130 GW by year-end.
• Transformers are still short, and policy made it tighter. Wood Mackenzie (Aug. 31, 2026) estimates a 2026 market shortage of about 15% for power transformers and 8% for substations. It expects the executive order restricting Chinese bulk-power equipment to make that worse. I could not confirm a current, sourced lead-time figure.
• Retirement overrides are legally weaker but still being issued. On Sept. 11, 2026 the D.C. Circuit vacated DOE's first Section 202(c) order for Michigan's J.H. Campbell coal plant; later extension orders remain in effect, the current one through Nov. 14 (Michigan AG; Utility Dive). DOE issued another 90-day order for a different coal plant the same day, according to Inside Climate News. I found no confirmation yet of a rehearing petition or appeal.
• TVA locked in its plan. TVA's board approved the final 2026 IRP and a new data-center rate class on Aug. 20, 2026 (Sierra Club; WFAE). The final plan's exact coal retirement dates were not clear in what I could find.
• China's coal is behaving like insurance. CREA's September 2026 snapshot has coal generation down 1.5% year-on-year, the third straight monthly decline. That follows a first-half rebound, when CREA and Global Energy Monitor found coal generation up 3.4%.
• Capital is building behind-the-meter platforms. Williams sold a 49% stake in a joint venture holding five behind-the-meter gas power projects to a Blackstone-led group for $5.34B (Business Wire, July 13, 2026). Total-capacity forecasts vary too widely to cite.

The pushback worth taking seriously

• "The shortage is narrower than you say." Commenters pushed back that only heavy-duty frame turbines from the top OEMs are sold out; aeroderivatives, engines, other brands and used units can still be bought. I accept that. Scarcity pushed buyers down the ladder, and the cost shows up in heat rate, O&M and certification risk rather than the sticker price.
• "Gas isn't fast either." People pressed that a two-to-three-year gas build assumes the turbine is already in hand. Right. The speed belongs to whoever holds the slot, not to gas.
• "Storage can match the date at some sites." Possibly. What I'd look at is whether lenders will write that credit case for a load that needs very high uptime. Most such deals still carry a gas backstop.
• "People are the constraint, not factories." On transformers, commenters argued that new plants still need trained workers who take years to develop, and that averages hide longer waits for high-voltage units. I agree. Easing will lag the plant opening dates.
• "Your cost numbers were too clean." My July 8 behind-the-meter cost comparison drew fair questions on sources and fuel-cell assumptions. Treat it as directional.
• On coal, people pressed that firm capacity has value at peak even as its generation share falls. Agreed. The open question is who pays that insurance premium.

What it means if you sell into this

• Sell certainty about dates. If you can contractually stand behind a delivery date, that's your margin. Price against what a late arrival costs the buyer, not against a competitor's quote.
• Sell into the gap between reservation and order. Slot holders still need controls, fuel trains, interconnection gear, storage and permits sequenced from day one. Work that protects a slot is easier to buy than work that competes with it.
• IPPs and behind-the-meter developers: your competitor for the slot is often a balance sheet, not another developer. What I'd look at is reservation-backed joint ventures and whether your financing structure can close as fast as the slot window requires.
• Equipment and supply-chain companies: watch terms before volume. Concentration usually shows up first as slower payment and harder negotiation, long before the backlog shrinks. Track how much of your backlog sits behind a signed end-user lease.
• Exposed to retained coal or old peakers? Extension orders are now legally contested. Model them as temporary.

The question I'd want answered first: If you lost your slot tomorrow, what would the next available date cost your buyer?

What would change my read

• GE Vernova or its peers report reservations falling because of cancellations or deferrals rather than conversions to orders.
• Interconnection or large-load reforms cut time-to-power below heavy-duty turbine lead times in major markets.
• Wood Mackenzie's transformer shortage estimate falls meaningfully in its next update, before the 2028 capacity additions arrive.

If you're already seeing one of these, I'd rather hear it now than find it in the next filing.

What to watch

• GE Vernova's Q4 2026 results (expected late January 2027): does gas backlog plus reservations reach the guided 130 GW, and do conversions keep outpacing new reservations?
• By March 31, 2027: does DOE win rehearing or Supreme Court review of the Campbell ruling, or keep issuing 90-day 202(c) orders regardless?