On July 14, PJM Interconnection, the regional grid operator for 67 million people across 13 states and the District of Columbia, announced the results of its capacity auction for the year starting June 2028. The price came in at $325 per megawatt-day, the maximum this auction was legally allowed to produce, making it the third consecutive auction to clear at its government-brokered cap.
Buried in the appendix of PJM’s own auction report is the number that should be on the front page: without the cap, the auction would have cleared at $554.72, about 71% higher. That gap is not a rounding error. By PJM’s own simulation, the cap knocked $13.3 billion off the bill for a single delivery year.
Here is the part almost nobody has reported: the cap is temporary, and it is almost out of runway. It covers four auctions. The third just cleared. The fourth runs in December. After that, per Pennsylvania’s governor’s office, the first uncapped auction arrives in May 2027, and the $554.72 world stops being a simulation. PJM has, in fact, already put the bigger number in writing: the emergency capacity auction it asked regulators to approve on July 31 carries a proposed price cap of $555.
What Is the PJM Capacity Auction?
PJM is the nation’s largest grid operator , coordinating the grid across all or parts of Delaware, Illinois, Indiana, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, West Virginia and the District of Columbia.
The capacity auction is not where electricity itself is bought. It is an insurance market: power plants bid for the right to be paid a daily retainer, in dollars per megawatt-day, in exchange for promising to be available three years from now when the grid hits its highest-risk hours. Those retainer payments get folded into the wholesale costs that your utility passes through to your monthly bill. PJM itself is careful to note that capacity is only a fraction of wholesale power costs, which are in turn a portion of retail bills. But when the retainer for the entire fleet jumps, tens of millions of households pay the increase at once.
For most of the last decade this market was boring. Then data centers started landing on the forecast. PJM’s July release notes, in its own dry phrasing, “the continued trend of the addition of large data center loads to the load forecast that forms the basis of the reliability requirement.”
Three Auctions at the Cap, and $16.4 Billion
The 2028/29 auction procured 138,318 megawatts (MW) of unforced capacity (UCAP), which is a plant’s output discounted for how reliably it actually shows up under stress. The arithmetic on what that costs is straightforward:
That $16.4 billion figure is PJM’s own total for cleared supply times the clearing price.
The cap itself was negotiated, not organic. In December 2024, Pennsylvania Governor Josh Shapiro sued PJM, arguing its auction design would saddle consumers with billions in unnecessary costs, and the resulting settlement, approved by the Federal Energy Regulatory Commission (FERC), capped the next two auctions at about $333 per megawatt-day. On April 28, 2026, FERC approved an extension of the price collar, a cap of roughly $325 per megawatt-day plus a floor of $175, covering the auction that just cleared and the next one, for the 2029/30 delivery year, which closes December 15. PJM says the collar was established “in coordination with the governors of all 13 PJM states” and FERC, and covers four auctions in total. The prior auction had cleared at its own, slightly higher cap of $333.44.
Three straight auctions pinned at a legal maximum tell you the cap is not a ceiling the market occasionally touches. It is the binding price, every time. The auction is trying to say a bigger number and is not allowed to.
The Real Price Is $554.72
PJM’s auction report includes a simulation of what the 2028/29 auction would have done with no cap and no floor: every zone clears at $554.72 per megawatt-day, except the Chicago-area ComEd zone, which separates and clears at $776.69. Total cost rises from $16.4 billion to $29.7 billion.
The trend of those simulations is the alarming part. PJM ran the same exercise for the two prior collared auctions: the uncapped price would have been $388.57 for 2026/27, then $529.80 for 2027/28, now $554.72. Across the three collared auctions, the suppressed “real” price climbed 43% while the visible price sat politely at the cap.
One honest caveat, which PJM itself flags: the simulation reuses the offers sellers actually submitted, and nobody knows how bidding behavior would change in a genuinely uncapped auction. The counterfactual is an estimate, not a parallel universe. But it is the grid operator’s own estimate, published in its own report.
Why Are Electric Bills Going Up?
Because demand is growing faster than supply, and the auction just put numbers on the mismatch. The peak-load forecast behind this auction is roughly 2,000 MW higher than the previous one. Against that, the auction cleared just 525 MW of new generation and uprates. For every megawatt of new supply that showed up, roughly four megawatts of new expected demand did.
“These auction results show that demand for electricity continues to grow faster than electricity supply,” PJM President and CEO David Mills said in the release.
The cap holds the price of that mismatch down; it does not make the mismatch smaller. PJM says as much: the collar “may reduce volatility” but does “not solve the underlying supply-demand imbalance.” Shapiro’s office, which negotiated the cap, claims the collar has saved $18.2 billion to date, roughly $207 per Pennsylvania household in the 2028/29 delivery year alone, and projects $45 billion in total savings across the region. Those are a politician’s own accounting of his own win, so treat the precision loosely, but the direction matches PJM’s independent $13.3 billion simulation delta for this auction.
This is the same demand wave behind New York’s data-center moratorium fight and the economics of AI’s compute appetite. A frozen unit price does not freeze the bill when the volume of the problem keeps growing.
The Auction Bought Almost Everything and Still Fell Short
The stat that should worry you more than any price: PJM needed more capacity than the region had to sell. Committed supply came in 6,831 MW short of the reliability requirement, the amount needed to meet the standard of no more than one major shortfall event in ten years. The previous auction was short by about 6,500 MW, and PJM says these two auctions are the first in its history in which the entire regional footprint missed the requirement.
Falling short does not mean blackouts are scheduled. PJM still carries a 14.7% reserve margin for 2028/29, and a shortfall against a probabilistic standard means thinner insurance, not empty wires. But the response tells you how seriously PJM takes it. On July 31, PJM filed a plan at FERC for a one-time backstop capacity auction, running September 30 to October 21 with results by December 2, to buy its way closer to the standard. And here is the detail that gives the whole game away: PJM proposed a $555 per megawatt-day price cap for the emergency auction, up from the main auction’s $325. Set that next to the simulation. The emergency price ceiling lands within a dollar of the $554.72 that PJM’s own model says the market would charge without the collar. The side channel is priced at the real price. The cap did not delete the number; it moved it into a filing that is three days old.
Will Electricity Prices Spike in May 2027?
The honest answer is conditional. The cliff is real but not guaranteed, and the conditions on both sides deserve naming.
The case for the spike is simple: three auctions pinned at the cap, an uncapped counterfactual that has climbed every year, and a supply pipeline that just delivered 525 MW of new generation against 2,000 MW of forecast growth. Moody’s called the auction results credit negative for regulated utilities in the region, pointing at the tension between affordability and reliability as data-center load growth, plant retirements and transmission constraints reshape the market.
The case against it has three legs. First, the cap has been extended once already; Shapiro sued to get it, then negotiated the extension, and nothing stops a third round if the politics demand it. Second, PJM is trying to open the supply valve: clearing its interconnection queue backlog, fast-tracking up to 10 state-sponsored, shovel-ready generation projects, and brokering long-term contracts that match large loads directly with new generation. If even part of that pipeline lands before mid-2027, the uncapped price falls. Third, the demand side is itself a forecast, and forecasts of speculative data-center projects can deflate; a canceled campus is 1,000 MW that never needed insuring.
Weigh those honestly and the cliff shrinks from a certainty to a coin flip with bad odds. If the supply pipeline and a softer load forecast both land in time, the uncapped auction clears closer to the old cap than to the emergency ceiling, and this warning ages into a footnote. The evidence arguing otherwise is a market pinned at its ceiling every time it has run under the collar, which is still a thin sample. What that sample agrees on is the direction: a capped market that keeps maxing out is storing pressure, not releasing it.
Oregon Wrote Down a Different Answer
There is another way to route the cost, and one utility on the other coast just demonstrated it. Following a 2025 Oregon law, House Bill 3546, regulators created a separate rate class for large data centers served by Portland General Electric. The result, effective July 8, 2026: an average rate increase of 29.7% for data-center customers, alongside average decreases of 1.3% for residential, 2.1% for commercial and 1.4% for other industrial customers.
Read that again: the customers driving the new demand absorbed the new cost, and everyone else’s rates went down. It is one utility, one state, and a rate structure rather than a capacity market, so the mechanics do not transplant directly into PJM’s 13-state auction. But it is proof that “data centers raise everyone’s bill” is a policy choice, not a law of physics. PJM is edging toward its own version with “Connect and Manage” rules that would let big loads join the grid on the condition that they power down flexibly during crunch hours , and FERC has already forced the issue for co-located data centers at power plants.
The gray truth of the cap is that both sides of the argument are right. The collar saved consumers real money; PJM’s own simulation says so. And the collar built nothing; PJM’s own shortfall says that. A price cap is a painkiller, and the region is on its third dose while the infection spreads.
Three Dates Decide What You Pay
Watch September 30, when PJM’s just-filed backstop auction is set to open, pending FERC approval, with its $555 emergency ceiling. Watch December 15, when the fourth and final collared auction closes for 2029/30. Then watch May 2027, when the first auction beyond the extended collar’s reach arrives and, unless someone negotiates a third round of caps, the training wheels come off a market that has hit its ceiling three times in a row.
Sources (9)
- pjm.com PJM News Release: 2028/2029 Base Residual Auction Results (July 14, 2026)
- pjm.com PJM 2028/2029 Base Residual Auction Report
- insidelines.pjm.com PJM Inside Lines: FERC Approves Capacity Auction Price Collar to Next Two Capacity Auctions
- pa.gov Pennsylvania Governor's Office: Legal Action Again Prevents Price Hike Across 13 States
- sec.gov Portland General Electric Form 10-Q, quarter ended June 30, 2026 (SEC EDGAR)
- utilitydive.com FERC approves PJM capacity auction price cap, floor
- reuters.com High PJM grid prices are credit negative for regulated electric utilities, Moody's says
- utilitydive.com PJM capacity prices hit price cap, reserve shortfall grows
- utilitydive.com PJM files backstop auction plan at FERC to meet capacity shortfall
🦋 Discussion on Bluesky
Discuss on Bluesky