Take a household in northern New Jersey that uses 650 kilowatt-hours of electricity a month. That is the figure New Jersey's Board of Public Utilities uses when it wants to show residents what is about to happen to their bill. In February 2025 the board did exactly that. For a PSE&G customer of that size, the typical monthly bill would go from $155.84 to $182.71 on the first of June. The household did not have to change anything to pay the extra $26.87. Customers of Jersey Central Power & Light were told to expect 20.2 percent more, customers of Atlantic City Electric 17.2 percent. The board's president, Christine Guhl-Sadovy, named the cause in one sentence: "PJM's recent capacity auction results are the main driver of these increases."

Most people who search for why their electric bill is so high never meet those three letters. PJM Interconnection runs the high-voltage grid and the wholesale power market for 67 million people in thirteen states and the District of Columbia, from the suburbs of Chicago to the Jersey Shore and south through Virginia. Once a year, give or take, it holds an auction that does not buy electricity. It buys a promise: that a power plant will exist, and be ready to run, in a delivery year two or three years away. Utilities pass the price of that promise through to customers inside the supply charge, a line almost nobody reads. It is one part of that charge, next to the cost of the energy itself. In 2025 it was the part that moved most.

The auction has one job. When the region needs more power plants than it has, the price of the promise is supposed to rise until building one is worth someone's money. PJM's forecasts, increasingly full of data centers, say the region needs more. So the price rose, to the highest level the rules allow, three auctions in a row. This piece follows that money from the June bill back to the auction, and from the auction to the plants it was meant to call into being. The short version is a matter of speed. In this market the money runs on a faster clock than the steel.

Why Your Electric Bill Rose: PJM Capacity Auction Prices, $28.92 to $269.92

For the delivery year that began in June 2024, the auction cleared at $28.92 per megawatt per day. PJM put the total cost to customers at about $2.2 billion. For the year that began in June 2025, it cleared at $269.92, and the cost rose to $14.7 billion. That is an increase of 833 percent in a single auction, and it is the one behind the June 2025 bill.

Much of the pressure came from demand that did not yet exist. The auction sets its target from PJM's load forecast, and that forecast had begun to fill with data centers still on the drawing board. Households were paying in 2025 for a shortage expected in 2027.

The price is paid to every megawatt that clears, old and new alike, and that shows where the money goes. Constellation, the country's largest owner of nuclear plants, reported in its 2025 annual filing that its average capacity price in the Mid-Atlantic had gone from $51.89 per megawatt-day in 2024 to $179.79. Around Chicago it went from $31.09 to $169.50. In the company's Midwest business, capacity revenue improved by $195 million, "primarily due to higher prices." The same filing shows the higher prices also reduced a federal nuclear tax credit, which offset part of the gain. And revenue is not profit: capacity payments also keep plants staffed, maintained and ready. None of it required a new plant.

This is not unusual. Paying every supplier the same price is how most commodity auctions work, and there is a good reason for it. An old plant that would otherwise close is also capacity, and keeping it open is cheaper than replacing it. The design accepts that existing owners will collect the scarcity price as the cost of telling new builders the region is short. That bargain only works if the new builders come. The rest of this piece asks whether they did.

A Letter About a Wealth Transfer, January 2025

By the end of 2024 the auction had an opponent who was not a consumer group. Pennsylvania's governor, Josh Shapiro, filed a complaint against PJM at the Federal Energy Regulatory Commission in December. On 13 January 2025 he wrote to the chair of PJM's board, Mark Takahashi, about the next auction, which under PJM's rules could clear far higher than the last. One line carried the whole argument: "I find the specter of such a vast wealth transfer to be unacceptable."

The phrase is worth stopping on. It came from the governor of the state that produces more electricity for the PJM market than any other, and it described the auction not as a way to buy supply but as a way to move money. Two weeks later, on 28 January, PJM and the governor announced a settlement. The next two auctions would have a ceiling of about $325 per megawatt-day and a floor of about $175. FERC approved it in April 2025, and on 28 April 2026 extended it through the delivery year that begins in June 2029.

The ceiling mattered. Without it, PJM's own simulation says, the most recent auction would have cleared at $554.72 and cost $29.7 billion instead of $16.4 billion. In April 2026 Shapiro's office put the savings from the first two capped auctions at $18.2 billion; that is the governor's count, and it depends on PJM's simulations being right. What the ceiling did not change was the signal. A capped price is still the highest price the rules allow, and from mid-2025 on, that is where every auction cleared.

Three Auctions at the Cap, July 2025 to July 2026

The ceiling was written once, in the units PJM uses for installed capacity. Because PJM credits plants a little differently each year, it shows up as $329.17 in the first capped auction, $333.44 in the second and $325 in the third. They are the same ceiling, and all three auctions hit it.

The record gives a rough picture of how new supply answered the price. At $28.92, in the auction for 2024/25, 502 megawatts of new plants and upgrades cleared. At $269.92, for 2025/26, it was 864, most of it upgrades to existing plants. Then came the ceiling, and with it the question of what the highest allowed price would buy.

In July 2025, for 2026/27, it bought 2,669 megawatts. The region met its reliability target with 139 megawatts to spare, and PJM's press release said supply was responding to the price signal.

In December 2025, for 2027/28, it bought 774 megawatts. For the first time the auction fell short of the reliability target, by about 6,500 megawatts.

In July 2026, for 2028/29, it bought 317 megawatts of new plants and 208 of upgrades, 524.7 in all. The shortfall grew to 6,831 megawatts.

Three auctions in a row paid the most the rules allowed, and each one bought less new power than the one before.

The latest auction cleared 138,318 megawatts at $325. Only 524.7 of them were new or upgraded. For every new megawatt, the auction paid for about 260 that already existed. (Utilities supplied another 10,864 megawatts on their own outside the auction, which is why PJM's total for the year is 149,181.6.) PJM's chief executive, David Mills, gave the reason in one sentence: "demand for electricity continues to grow faster than electricity supply."

Three cautions belong here. These are megawatts committed for future years, not plants switched on. A shortfall means a thinner safety cushion, not a forecast of blackouts. And the reserve targets cannot be compared across years, because PJM changed how it credits each type of plant from 2025/26 on. None of the three changes the central fact: the loudest signal the market can send went out three times, and the new supply answering it fell from 2,669 megawatts to 774 to 525.

Two Testimonies, Philadelphia, 22 April 2026

At ten in the morning on 22 April 2026, the Pennsylvania House Democratic Policy Committee opened a hearing on utility price increases in the Fox-Gittis Room of Temple University's Liacouras Center, on North Broad Street. The witness list put the two halves of this story in one room. Brendan Taylor, a vice president of PECO, the utility that bills Philadelphia, submitted testimony that said: "Since 2024, the average residential customer bill has increased by $28, or 21%," an increase "driven primarily by higher capacity payments to PJM." Asim Haque, an executive vice president of PJM, brought slides. One said that "consumers are deeply exposed to this high-capacity market price." Another said: "We need more energy infrastructure, both generation and transmission."

Neither witness contradicted the other. The utility described what the auction had already done to the bill. The grid operator described what the auction had not yet been able to produce. Read side by side, the two documents describe the whole mechanism, and nobody in them is wrong.

53 Gigawatts With Contracts and No Electricity

If the price is this loud and new plants are not coming, the obvious suspect is that nobody wants to build. PJM's own figures point somewhere else. In a fact sheet updated in May 2026, it reported that about 53 gigawatts of generation had signed interconnection agreements, the contract that lets a plant connect to the grid, and were still not running. A signed agreement is not a shovel in the ground; financing, permits, land and equipment can still be missing. PJM blamed "permitting, siting, and supply chain challenges," and called them beyond its control. For comparison, about 25 gigawatts had connected since 2020.

Fifty-three gigawatts on paper is not 53 gigawatts in the auction. PJM credits each type of plant with only part of its size: in the 2028/29 auction, fixed solar counted for 7 percent, tracking solar 10, onshore wind 34, offshore wind 60, four-hour batteries 59, gas turbines 67 and combined-cycle gas plants 78. PJM does not publish the current mix of the 53 gigawatts. Its last public breakdown, from September 2024, covered about 36 gigawatts of signed projects: 23.4 of solar, 5.3 of gas, 3 of offshore wind, 2.7 of onshore wind and 1.9 of storage. Weighted by the 2028/29 ratings, that mix counts for roughly a quarter to 28 percent of its size. If today's 53 gigawatts look similar, they would be worth something like 13 to 15 gigawatts in the auction's own units. That is my estimate, not PJM's, and the mix has probably shifted toward gas and batteries since 2024. Even on the cautious end, it is about twice the latest shortfall. One more caution pulls it down: in 2024 about a third of the signed capacity was suspended, with the local transmission owner no longer working on the connection. On the active part alone, the margin over the shortfall is closer to one and a half times than two.

So the honest reading is narrower than "the plants are all there." There is a large pipeline of contracted projects that are not yet running, worth more than the shortfall in auction terms, and what holds them back is something a price moves only slowly. Behind them stands a longer line still: more than 170 gigawatts of applications processed in PJM's reformed transition between mid-2023 and the end of 2025, and about 220 gigawatts in the first new cycle.

The Turbine Line

When PJM wanted firm power quickly, it did not wait for the auction. In May 2025 it ran a one-time fast track, the Reliability Resource Initiative, and picked 51 projects: about 11,800 megawatts of installed capacity, or about 9,400 in the units the auction counts. Most are natural gas, and 39 of the 51 are upgrades to plants that already exist. PJM expects about 90 percent of it online by 2030 and all of it by 2031.

Those dates are set by the physical pipeline: permits, construction crews, fuel connections, and the factories that make the machine at the center of a gas plant. GE Vernova, the largest maker of heavy-duty gas turbines, reported in April 2026 about 100 gigawatts of gas power already spoken for, in orders and reservations, with roughly 10 gigawatts left across 2029 and 2030. Its chief executive, Scott Strazik, put the wait for a new heavy-duty turbine at about three years. On the same call he said the turbine is often not even the slowest part; the construction capacity, the permits and the fuel supply are. Wood Mackenzie expects turbine prices near $600 per kilowatt by the end of 2027, nearly three times the 2019 level.

Put the clocks side by side. The auction runs once a year and buys capacity two to three years ahead. A new gas plant, from turbine order to first power, takes longer than that. The bill resets every June. A plant ordered the day the auction results come out will often arrive after the year the auction was paying for. The money moves on the fastest of the three clocks, and the steel on the slowest. Put generally: a scarcity price turns into a transfer whenever new supply takes longer to build than the period the price is paying for.

The same mismatch shows up wherever money can be committed faster than metal can be formed. After the war in Ukraine drained American artillery stocks, the Army set out to raise production of 155-millimeter shells from about 14,000 a month to about 100,000 by October 2025. By March 2026 it was making about 36,000. A new $469 million plant in Mesquite, Texas, planned for 30,000 a month, had produced no usable projectile metal parts at all, according to the Pentagon's inspector general as reported by Breaking Defense. The contracts were signed and the money appropriated, and the shells still waited on a forging line that did not yet work. Wherever money can be committed faster than metal can be formed, the price arrives first and the supply last, and in between the money goes to whatever already exists.

$1.2 Billion in One Filing, 14 July 2026

On the day PJM published its 524.7 megawatts, Talen Energy filed a report with the Securities and Exchange Commission. Talen owns plants in Pennsylvania and neighboring states, including most of the Susquehanna nuclear station. The filing said Talen had cleared 10,180 megawatts in the auction at $325, worth approximately $1,208 million for the year. A year earlier it had cleared 6,702 megawatts, worth about $805 million. Vistra reported in December 2025 that it had cleared about 10,566 megawatts at $333.44; by my arithmetic, not Vistra's, that is roughly $1.29 billion for a year.

None of this is misconduct. It is the auction working as designed: existing capacity earns the clearing price, and the clearing price was the ceiling. It shows where the signal lands while the new plants wait. On 14 July 2026, one company's share of a single auction was worth more than $1.2 billion. The whole region's new generation and upgrades in that same auction came to 524.7 megawatts.

One filing, $1.2 billion; one region, 524.7 new megawatts.

The Strongest Case That the Price Is Working

The defense of the auction is serious, and parts of it are right.

Keeping a plant open is the fastest way to add capacity, and the high prices did that. After the auction of July 2024, PJM reported, owners of 17 generating units, about 1,100 megawatts, withdrew their notices to retire them. That capacity never shows up as "new," yet every megawatt of it is one the region would otherwise have lost. The 2026/27 auction drew 2,669 megawatts of new supply and upgrades, more than the auctions before it. The fast-track projects will add about 9.4 gigawatts, in auction terms, by 2031. A new plant takes four to six years from decision to power, so judging a 2025 price by what is running in 2028 tests it before it could answer. And the ceiling, introduced to protect customers, also muffled the signal it carried. A critic could fairly say the auction has not failed to call new supply; it has simply not finished calling.

Most of that is true, and it narrows what this piece can claim. The price is not doing nothing. What it cannot do is buy new megawatts faster than the physical pipeline can deliver them. In the meantime, and the meantime is now several years long, the full scarcity price is paid on roughly 138,000 megawatts while the new megawatts it is meant to summon are counted in hundreds. That is my reading of the record, not a finding PJM has published.

What I cannot establish is how much of the kept and added capacity would have come forward at half the price. The auction publishes what cleared, not what would have cleared, and PJM's simulation without the ceiling only runs the price upward.

Who Pays for the Data Centers

The reason the region is short is no mystery. PJM's 2025 load forecast expected peak demand to grow by about 32 gigawatts between 2024 and 2030, and all but about 2 gigawatts of that growth to come from data centers. The auction prices that demand in before it arrives, because the target is set by the forecast.

PJM's independent market monitor, Monitoring Analytics, has tried to put a number on it. In a presentation to PJM members on 27 July 2026, it said that of the $16.4 billion in charges from the latest auction, "$6.3B was a result of data center load," or 38.2 percent. Across the last four auctions it put the figure at $29.4 billion, 46.2 percent of the total. These are what-if estimates, the cost of the auctions compared with what they would have cost without the forecast data center load. They are not an amount on anyone's bill.

Who should pay for capacity bought to serve new data centers is the open question in every forum that matters, and none has answered it. On 29 September 2026 FERC accepted PJM's plan for a one-time emergency procurement, froze it for five months until 28 February 2027, and opened a hearing on how to split the cost. On 9 October it rejected most of PJM's proposal for handling new large loads, including a rule that would have let PJM cut off new data centers first in an emergency, because parts of it reached into retail rates, which belong to the states. It kept only a definition of a large load, 50 megawatts or more, and a registry. The day before, at a summit in Chicago hosted by Illinois governor JB Pritzker, governors from across the region signed a memorandum formally establishing the PJM Governors' Collaborative, first announced in September 2025, to press for a stronger state role in the market's rules.

While the question waits, the default answer is the one the auction already gives. Every customer pays the clearing price, in proportion to their share of peak demand, whether or not they caused the forecast to rise.

June 2026 and the Bill That Stayed

The ceiling did one thing households could see. On 1 June 2026 the typical PSE&G bill, at the same 650 kilowatt-hours, went from $183.46 to $180.23, before credits. The board said the ceiling had kept prices lower than they would otherwise have been. That is true, and it is the whole of the relief. The bill stopped rising and stayed where the 2025 auction had put it, about $24 a month above the $155.84 the same household paid in May 2025.

15 December 2026

The next test has a date. PJM's auction for 2029/30 closes on 15 December 2026, with a ceiling of $337.83. If it clears more than the 2,669 megawatts of new supply the 2026/27 auction drew, or meets the reliability target, the price is starting to buy what it was built to buy, and this reading weakens. If it clears at the ceiling again with less than 1,000 new megawatts, the gap between the auction's clock and the factory's clock is the story behind these bills for the rest of the decade.

The household in northern New Jersey will see the result in June 2029, folded into the supply charge, without using a single extra kilowatt-hour.

A price can only buy what the queue lets through. Until the queue moves, the price is mostly a transfer.

Evidence Map

Facts, interpretations, forecasts, and disconfirming signals.

Core claim. PJM's capacity price has been at its cap for three auctions while the new generation and uprates it cleared fell from 2,669 to 774 to 525 megawatts. The high price has kept existing plants open, which has real value, but because the binding constraint is the physical build pipeline, from interconnection and permits to construction and turbine slots, the high price mostly pays existing plants until new ones can be built.

Evidence level. Facts, high: auction prices, volumes and shortfalls from PJM's auction reports; bill changes from the New Jersey Board of Public Utilities; company revenue from SEC filings; FERC actions from the orders. Interpretation, medium: that the price acts mainly as a transfer during the build gap. Forecast, speculative: the 15 December 2026 auction result.

What would confirm this. The 2029/30 auction clears at or near the cap with less than 1,000 megawatts of new generation and uprates; the share of signed interconnection agreements not operating stays near 53 gigawatts.

What would weaken this. The 2029/30 auction clears more than 2,669 megawatts of new generation and uprates, or meets the reliability requirement. That would show the response improving, not that the earlier gap was harmless. What would contradict the mechanism. Evidence that new capacity was held back by the price itself rather than by the build pipeline, for example signed projects with permits and equipment in hand that declined to clear at the cap; or the fast-track projects closing the shortfall on schedule by 2030.

Watchlist. PJM auction results on and after 15 December 2026, scored in the archive's Track Record once PJM publishes them; the condition to watch is whether the build pipeline can answer the price inside the delivery year it pays for, the point at which a capacity price stops being mostly a transfer; FERC's cost-allocation hearing on the backstop procurement; New Jersey and Pennsylvania supply prices on 1 June 2027.

Sources. New Jersey Board of Public Utilities, press release of 12 February 2025 and of 12 February 2026; PJM's 2025/26 auction report and 2028/29 auction results report; Monitoring Analytics, market monitor presentation of 27 July 2026; PJM's interconnection reform fact sheet; PJM's ELCC class ratings for 2028/29; Talen Energy's filing of 14 July 2026; the written testimony to the Pennsylvania House hearing of 22 April 2026; and FERC's order of 29 September 2026.

Jerry van der Laan writes The Manifest Archive, Forensic Narrative Intelligence Writing on the systems beneath power, money, and history. He traces the structures beneath them.