555 Was the Real Price. The Cap Said 325.

Two numbers landed on 14 July, and only one was chosen

On 14 July 2026 PJM announced the results of its Base Residual Auction for the 2028/2029 delivery year, and the headline read like relief. The clearing price came in at 325 dollars per megawatt-day across the entire PJM footprint, down from 333.44 the year before. PJM's chief executive David Mills did not present it as relief. His line, on the record, was that "demand for electricity continues to grow faster than electricity supply."

The arithmetic behind that sentence reconciles exactly. PJM procured 138,318 megawatts of unforced capacity from generation resources. Add the 10,864 megawatts carried by fixed resource requirement entities and the total reaches 149,182. The reliability requirement was 156,012.9. The region therefore finished 6,831 megawatts short of what its own planning standard says it needs, at a reserve margin of 14.7 percent, for a total auction cost of 16.4 billion dollars. This is the second consecutive auction in which the entire PJM region has fallen short. The previous one, for 2027/28, was the first ever to do so, by roughly 6.5 gigawatts.

The number to interrogate is the 325. It is the ceiling FERC approved under docket ER26-1556, set in agreement with 13 governors, and this is the third consecutive auction to run under that collar. PJM published its own counterfactual alongside the result. Without a price collar, the auction would have cleared at nearly 555 dollars per megawatt-day across the footprint and 777 dollars in the Commonwealth Edison zone in northern Illinois, at a total cost of 29.7 billion dollars against the 16.4 billion actually charged. The market wanted roughly 555. The rule said 325, and the auction cleared short anyway.

A 2.5 percent fall that contains no information

The price went down and the market did not. Last year's cap was 333.44 dollars per megawatt-day. This year's is 325. Both auctions cleared at their ceiling, which means the only thing that moved between them was the ceiling itself, recalculated by formula. Reading minus 2.5 percent as demand cooling is the easiest mistake available with this result, and it is the one most likely to end up inside a budget.

The 16.4 billion dollar cost carries the same warning. PJM says the identical auction without a collar would have cost 29.7 billion. The nearly 555 dollars the market would have reached, on PJM's own reckoning, sits roughly 70 percent above the 325 the rule permitted. A number produced by a binding rule describes the rule. It has stopped describing the balance of supply and demand underneath it.

The collar is not new either. This is the third consecutive auction to run with one, set in agreement with 13 governors and approved by FERC. A ceiling that binds once is a circuit breaker. A ceiling that binds three years running is the operating regime, and every figure that comes out of it, including the reassuring ones, is an output of the rule.

What 325 a day summoned: about 317 megawatts

The supply side answered the capped price in a single line of PJM's release. The auction, in PJM's exact words, cleared 525 megawatts of new generation and generation uprates. Utility Dive's breakdown is the part to keep. Of that 525, about 208 megawatts are uprates to plant that already exists, which leaves genuinely new build somewhere near 317 megawatts against a reliability requirement of 156,012.9. The previous auction produced 774 megawatts of new resources. The number went down.

Aurora Energy Research put the reason on the record. Julia Hoos says new generation needs "significantly more" than 325 dollars per megawatt-day to be financially viable. That is a statement about a capital decision. A developer weighing years of construction and hundreds of millions of dollars looks at the maximum revenue the market is permitted to pay. When the permitted maximum sits below the hurdle, the plant does not get built, and demand growth behind the ceiling does nothing to change that.

This is where a buyer's model has to change. Three auctions of signalling under a collar produced roughly 317 megawatts of genuinely new steel in the largest competitive electricity market in the world. On that evidence, money paid into the construct buys a share of the plant that already stands. It does not summon more of it.

Read it this way: a cap that binds becomes a rationing rule

Our read, and PJM has not said this. Once a price ceiling binds and the auction still clears short, the ceiling has stopped working as a discount and started working as a rationing mechanism. In an uncapped auction the shortfall clears at some price, and a buyer who needs certainty can decide to pay it. At 325, the shortfall does not clear at any price. It gets settled administratively instead, and none of the instruments that settle it respond to your budget.

Look at what PJM has proposed and the shape of the thing is hard to miss. A special Backstop Procurement, which PJM will take to FERC for approval in September 2026. An Expedited Interconnection Track carrying 10 state-sponsored projects. Connect and Manage, under which data centres connect but operate flexibly and curtail when needed. Google's Tapestry AI applied to the interconnection queue. A bilateral matching RFP. Each of those allocates scarce capacity by rule, by queue position or by curtailment terms, and none of them is a price. That is what a market looks like once its price has been taken out of service.

The honest caveat. PJM frames none of this that way and hedges carefully. It attributes the shortfall to demand outpacing supply, says the result was not unexpected, and says the shortfall "does not necessarily mean that the PJM system will be unable to serve load reliably." The collar is also not the only thing holding offers down, because the Three-Pivotal Supplier Test limits offers independently of any cap. All of that is true and should be carried. What keeps the cap in first place is Hoos again, on the political economy of it: "Lowering prices [with the price collar] was definitely politically attractive in the short term, but now we're well on our way to facing an intervention doom loop."

What to ask before you sign for 2027 to 2029 load

PJM is a US market and it is also the case that Europe's capacity-market debate keeps citing, because it is the largest competitive electricity market in the world and it has now run the capped auction three times with published numbers at the end. The UK and several EU member states operate capacity markets of their own, each with its own caps and de-rating rules. The transferable lesson is about what any administered ceiling does to a buyer of firm capacity once it binds. Budget certainty gets replaced by availability risk, and availability risk is far harder to hedge.

Change the question first. In a capped market the clearing price is the least informative figure on the page, because a rule wrote it. The figures that carry information are the reserve margin, the volume of genuinely new build as against uprates, and the gap between what the operator says it needs and what turned up. For this auction those read 14.7 percent, roughly 317 megawatts and 6,831 megawatts. Those three describe what a region can actually host. A tariff comparison describes nothing.

Then negotiate the right variable. If a bound cap rations, price is no longer the lever you control. Position and firmness are. Ask where in the interconnection queue your site sits and what would move it. Ask which curtailment terms your contract accepts, in writing, and what operating flexibly means in hours per year. Ask who carries the cost if a backstop procurement is invoked. And ask what happens to your supply when the ceiling is raised, because a ceiling that was set politically can be reset politically.

Read next: PJM Chose to Build Power for Data Centers Rather Than Curtail Them  |  Europe Already Ran New York's Data Centre Freeze