PJM Now Costs More Than It Did in the Gas Crisis
July 23, 2026. FERC gave PJM until the end of September to reform its own governance or have reforms imposed on it. The argument in the hearing room was about board seats and voting rules. Nine years of PJM's market data show what is actually being argued over.
Open docket. FERC Docket AD26-7-000 is live. The Commission has said it will convene a dispute-resolution forum in September and impose reforms if PJM's stakeholders have not agreed a package by the end of that month. Every figure below is a snapshot of PJM's published market data taken on July 25, 2026, and will move as the year finishes. Year-over-year comparisons are aligned to January 1 through July 20, the window 2026 has completed.
On July 23, the Federal Energy Regulatory Commission spent a day in its Washington hearing room asking whether the largest power market in the United States can still govern itself. The subject was PJM Interconnection, which runs the grid for 67 million people across 13 states and the District of Columbia. The subject matter was procedural: how many years a director serves, which committee holds a filing right, whether a vote is recorded.
FERC Chairman Laura Swett did not treat it as a procedural matter. "PJM is facing a grave legitimacy crisis," she said. "Some transmission owners are openly discussing leaving the RTO altogether."
Underneath the procedure is a number, and the number is the reason the conference happened. PJM's day-ahead price in 2026 is running higher than it did during the global gas crisis of 2022 — at a time when gas costs a little more than half what it did then.
The most expensive year in the record#
Ask the Grid holds every hourly PJM day-ahead price back to January 1, 2018 (pjm.da_hrl_lmps). Comparing years mid-year takes some care: 2026 is only two-thirds finished, and PJM's expensive months are not evenly spread. So every year-over-year figure in this story is aligned to the same window, January 1 through July 20, the stretch 2026 has completed.
On that basis the PJM-wide reference price averaged $72.26/MWh in 2026, against $66.78 in 2022 and $32.24 in 2024. The system price has more than doubled in two years, and it has passed the high-water mark set when European gas buyers were bidding against American ones.
One month carries much of it. January 2026 averaged $142.49/MWh across the whole of PJM — the most expensive month anywhere in the nine-year record, and more than 40 percent above the worst month of 2022.
This is not the gas crisis repeating#
The 2022 spike had a simple explanation. Gas set the marginal price in PJM almost all the time, and gas was expensive: Henry Hub averaged $6.13/MMBtu over that January-to-July window. When fuel costs double, power costs double, and no governance reform would have changed it.
That explanation is not available in 2026. Over the same window this year, Henry Hub averaged $3.73/MMBtu — 39 percent below the 2022 level. Fuel got cheaper. Power got more expensive.
Dividing the power price by the fuel price gives the implied market heat rate — in effect, how much gas the market is pricing into every megawatt-hour. It is the cleanest available separation of a fuel problem from a scarcity problem, and it moves in one direction.
In 2021 the ratio was 9.57. Through the gas crisis it barely moved, reaching 10.89 in 2022 — confirming that year was about fuel, not scarcity. In 2026 it is 19.39, nearly double the crisis-year figure and the highest in the record.
Put the other way round: at 2021's ratio, this year's gas price would imply a PJM power price around $36/MWh. The market is clearing at twice that. The gap is what the system charges for being short.
Two honest caveats. This is a market-wide ratio, not a physical measurement of any plant: it rises when scarcity and capacity adders enter the price, when the marginal unit is a less efficient one, and when non-fuel costs such as emissions allowances rise. And Henry Hub is a national benchmark, while PJM's generators buy gas delivered at regional hubs whose prices can separate sharply from it in winter. Neither caveat reaches far enough to explain a ratio that has doubled. The direction is not ambiguous.
Scarcity stopped being rare#
Averages understate what changed, because the change is concentrated in the hours when the system runs out of room. Counting those hours is the sharpest single measurement in this story.
Across the seven full years from 2018 through 2024, the PJM-wide day-ahead price never once cleared $500/MWh. Not during the 2022 gas crisis, not in the December 2022 winter storm, not on any summer peak. The highest hour in all of 2022 was $470.19.
It cleared $500 three times in 2025. It has done so 77 times by July 20 of this year. Hours above $250 tell the same story from a lower shelf: 50 in the whole of 2022, 280 in the same partial year.
The worst week was in January, and it was not the one that made the news#
Ask the Grid covered PJM's July 2 demand record, when a heat wave pushed preliminary unrestricted demand to 168,158 MW and PJM warned large loads to prepare for backup generation. That day produced six hours above $500/MWh.
Five months earlier, Winter Storm Fern produced twenty-four. Every hour of January 27 cleared above $500 PJM-wide. It drew a fraction of the coverage.
At 7 a.m. Eastern on January 27, the day-ahead price in the BGE zone around Baltimore cleared $4,150.32/MWh, and the PJM-wide reference cleared $2,314.58 — the highest day-ahead hour in the nine-year record, roughly five times the worst hour of 2022. PJM projected an all-time winter peak near 147 GW, and on January 24 it applied to the Department of Energy for emergency authority under Section 202(c) of the Federal Power Act — five months before the two June orders that carried it through the July heat event.
One limit worth stating plainly: PJM's hourly load feed in our warehouse begins on May 14, 2026, so this story can show what January cost but not what January drew. The peak-demand figure above is PJM's own projection, not our measurement.
The increase has a geography, and it is the geography of the fight#
A system average hides the part that matters politically. PJM prices power by location, and the locations have separated.
| Zone | 2024 average | 2026 average | Change |
|---|---|---|---|
| BGE (Baltimore) | $39.43 | $101.05 | +156% |
| PEPCO (Washington DC) | $37.64 | $98.30 | +161% |
| DOM (Dominion, Virginia) | $36.89 | $94.49 | +156% |
| APS (Allegheny) | $31.75 | $76.71 | +142% |
| PJM-RTO (system reference) | $32.24 | $72.26 | +124% |
| PSEG (New Jersey) | $27.52 | $61.41 | +123% |
| AEP (Ohio, West Virginia) | $30.36 | $61.78 | +103% |
| COMED (Chicago) | $25.00 | $44.24 | +77% |
The Baltimore, Washington and Northern Virginia zones have roughly two and a half times their 2024 price. Chicago has gone up 77 percent — painful, but a different order of magnitude. In 2024 the BGE zone cost about 1.6 times the ComEd zone; today it costs 2.3 times.
That is not a coincidental map. It is the corridor where PJM forecasts the overwhelming majority of its data-center growth, where the July emergency actions were targeted, and whose state regulators arrived at FERC asking for a formal role in decisions their own retail customers pay for. The states pressing hardest for filing rights are the states whose zonal prices have moved most.
A necessary caution: these are wholesale prices at the zone, not household bills. What a customer pays depends on the utility's procurement, its hedges, and state regulation, and wholesale increases reach retail slowly and unevenly. The table shows where the wholesale burden landed, not what any family paid.
The retail consequence is nonetheless what drew the governors in. PJM's capacity auctions cleared at successive records roughly two years ago, and the resulting rate increases exceeded 20 percent for some utilities' customers. That is the political fuse on the docket: the capacity market produced the bills, the stakeholder process sets the capacity-market rules, and the states that pay have no vote in it.
The supply side did not hold up its end#
Prices rise when demand grows into supply that is not there. Demand growth is well documented. The other half is less discussed, and PJM has published it since 2015.
Forced outages — plants that failed, as distinct from plants taken off for scheduled work — averaged 9,328 MW across June and July 2026, the highest summer figure in the twelve-year record and more than double the 4,431 MW of summer 2020. The 2026 fleet is both more heavily loaded and less reliably available than the fleet of five years ago.
Then, the morning before the conference, three gigawatts walked off#
Everything above is a price record: the cost of a system that cannot build supply as fast as demand arrives. On the Wednesday morning before the conference, PJM got a different kind of warning.
At about 7:56 a.m. Eastern on July 22, a transmission line in northern Virginia went out of service. The data centers behind it did what their own protection schemes are designed to do: they disconnected from the grid and transferred to on-site backup generation. More than 3 GW of demand, roughly three percent of what PJM was serving, vanished from the system at once. The resulting voltage disturbance was observed from Washington to Chicago.
That event is visible in PJM's own published telemetry, which is worth showing because most accounts of it are not quantified.
Dominion's metered load falls from 17,683 MW to 14,775 MW between two consecutive five-minute samples: 2,908 MW gone, and because the feed only reports every five minutes, that is a floor on the true instantaneous loss rather than its depth. Load claws back over the following twenty minutes as the sites resynchronise.
The more revealing measurement is what the loss did to the balance of the system. Area control error is the operator's running score of how far the balancing area sits from equilibrium; through the preceding hour it wanders inside roughly plus or minus 400 MW.
At 7:57 the minute average hit 2,964 MW, with individual samples reaching 3,487 MW — close to nine times the normal band. It took until 8:08, about twelve minutes, for the system to settle back inside it. Dominion said its operators stabilised conditions within minutes, and PJM said there were no impacts to PJM's reliability. Both are consistent with the data. A grid that absorbs a 3 GW step change in twelve minutes without shedding load is a grid that worked.
The unsettling part is the mechanism. This was not a heat wave or a freeze. A single line fault caused three percent of demand to remove itself from the system in seconds, on the decision of private control schemes that PJM does not dispatch and cannot see in advance. The same concentration of load that shows up in the price table above as a 156 percent increase in the Dominion zone also turns out to be capable of leaving all at once.
It is a preview of the reliability question the governance fight has not caught up with. PJM's capacity market and its planning process both treat large loads as demand to be served. July 22 is a reminder that they are also equipment, with their own protection settings, their own operators, and their own reasons to disconnect. Federal and state regulators have begun studying how to manage sudden demand drops from data centers and crypto miners; none of that work was on the conference agenda the next morning.
What FERC is actually being asked to do#
Against that record, the July 23 conference in Docket AD26-7-000 examined PJM's decision-making machinery. PJM's rules are set through a layered stakeholder process in which member companies — utilities, generators, transmission owners, traders — vote by sector, with a Board of Managers deciding whether to accept the result. Critics argue the structure lets a well-organized bloc stall a change indefinitely, and that the board's deliberations and votes happen out of public view.
PSEG's regulatory officer, Jodi Moskowitz, described the defect from inside the process: the sector-weighted vote "doesn't sufficiently align voting interest with accountability." The companies that carry the least of the consequence can carry a decisive share of the vote.
| Proposal | What it would change | Raised by |
|---|---|---|
| Board independence | Move decisions from member votes to the Board of Managers, leaving stakeholders advisory | FERC commissioners, AEP, PSEG |
| Longer board terms | Extend three-year terms to six or nine years so directors cannot be removed for unpopular decisions | Discussed at the conference |
| State filing rights | Let states file proposals at FERC alongside PJM, rather than lobbying through the member process | State regulators, R Street |
| Open board proceedings | Public board meetings, roll-call votes in the minutes, and written orders explaining decisions | R Street testimony |
| Public interest mandate | Write an explicit affordability and reliability duty into how the board operates | PJM's independent market monitor |
FERC Commissioner David LaCerte framed the independence problem directly: "I don't know how to get around it without breaking the system up to where the board is empowered to get a backbone." Peter Lake, of the White House National Energy Dominance Council, said PJM needs "an independent, transparent board that can take action and make tough decisions without fear of being fired." Deputy Secretary of Energy James Danly put the transparency ask in one line: "We want the board to not be shrouded in mystery or secrecy."
The most commonly cited alternative is MISO's, where stakeholders advise and the board decides. AEP, which has previously said it would consider leaving PJM, backed an advisory structure and a larger state role through senior vice president Stacey Burbure. Kelsey Bagot, who chairs the Virginia State Corporation Commission, told the Commission that states currently meet PJM's board three times a year and that those meetings are not productive — a thin channel for the officials who answer to the retail customers now absorbing the increase.
Joseph Bowring, PJM's independent market monitor, put the sharpest version of the problem on the record. The staff's goals, he said, "should be specified as acting in the public interest, not acting in the interest of their members."
PJM did not resist. David Mills, who became chief executive in May, said the operator is "fully committed to rise to the challenge, including capacity market reform."
The Commission attached a deadline anyway. PJM's stakeholders have until the end of September to agree a governance package; FERC has said it will convene a dispute-resolution forum that month and impose reforms itself if no agreement emerges.
What we can measure, and where the record stops#
This publication can price any hour in PJM to the cent. It holds more than a billion day-ahead price rows and 825 million real-time rows across roughly 14,450 pricing nodes, back to 2018. We can say what the Baltimore zone cost at 7 a.m. on January 27, how many hours cleared $500 in each of nine years, and how much of today's price is not fuel.
We can also measure a three-gigawatt disconnection to the megawatt and the minute, three days after it happened, without asking anyone's permission.
We cannot see a single board vote. We also cannot see which data centers disconnected on July 22, what their protection settings were, or who decided them.
That asymmetry is the governance case stated in its most concrete form. The consequences of PJM's decisions are published in extraordinary detail — hourly, nodal, auditable by anyone. The decisions themselves are not. There is no public record of how the Board of Managers voted on a capacity-market rule, which directors dissented, or what reasoning was given. A market whose outputs are published to the cent, hour by hour, at 14,450 locations is governed by a process that publishes no roll call.
That is what makes the transparency proposals the most testable ones on the table. Board independence and term length are structural bets whose effects will take years to read. Open meetings, recorded votes and written decisional orders are different: either the record exists afterward or it does not. If PJM adopts them, the next version of this story can compare a decision to its consequences. Today it can only measure one side.
Our last PJM story ended at a data center's fence line, at the point where public reporting stopped. This one ends at a boardroom door. September will show whether it opens.
The charts above are frozen from the PJM data behind Ask the Grid at publication. Prices come from pjm.da_hrl_lmps, outages from pjm.gen_outages_by_type, the July 22 event from pjm.inst_load and pjm.area_control_error, and the gas benchmark from the EIA Henry Hub daily spot series. Open PJM on the main map to check any hour in this story yourself.






