Data Centers Drove PJM's Record Power Price. The Cost of Their Growth Is Spread Across Everyone's Peak, Not Billed to the Newcomers.

On 1 June 2026, the most expensive reliability insurance bill in PJM InterconnectionPJM InterconnectionThe largest US regional grid operator, coordinating the high-voltage electricity system for all or part of 13 states and Washington, DC.'s history started arriving. PJM is the largest US grid operator, the body that keeps the lights on for all or part of 13 states and Washington, DC. Once a year it runs an auction, the Base Residual Auction, that pays power plants in advance to promise they will show up on the hardest days of the year. Think of it as the grid buying insurance: customers pay a premium now so capacity is guaranteed later. For the 2026/27 delivery yeardelivery yearThe 12-month period (June to May) for which an auction's capacity is actually supplied and billed; the 2026/27 delivery year began 1 June 2026., that premium cleared at a record $329.17 per megawatt-day, and as of 1 June it began flowing onto electric bills.

Three weeks later, the federal regulator moved. On 18 June 2026 the Federal Energy Regulatory Commission (FERCFERCThe Federal Energy Regulatory Commission, the US agency that regulates wholesale electricity and interstate transmission.) issued Section 206 show-cause orders to all six US grid operators, a legal step that puts existing rules on trial and gives the operators 60 days to prove the rules are fair or be made to change them. The stated worry: that giant new electricity users, above all data centers, could shift costs and strain reliability for everyone else. The clock runs out on 17 August 2026.

It is now common knowledge that data centers are pushing up power prices. That broad claim is true and largely answered. So we asked something narrower: now that the jump is hitting June 2026 bills, where exactly is the design flaw? Is it that data centers underpay their share, or something subtler about how the extra cost their growth created gets divided? And would FERC's reform change it?

The distinction that matters is between two ideas that sound alike but point to opposite fixes. The first is that data centers free-ride, that they use the grid heavily but somehow escape the bill. The second is that they pay plenty in absolute terms, but the marginal cost their growth adds gets averaged across every customer's peak demand instead of charged to the loads that caused it. Those are not the same grievance, and only one of them is supported by how the rate mechanics actually work.

To answer it we did three things: traced where the record price came from and how much of it the market's own watchdog blames on data centers; followed the formula that splits capacity cost among customers, to see whether data centers under- or over-contribute; and read what FERC's 18 June orders can and cannot reach. The short version: the free-rider story is mostly wrong, the marginal-cost story is mostly right, and the fix is split across two regulators.

PJM's independent market monitorindependent market monitorAn outside watchdog (for PJM, the firm Monitoring Analytics) that audits the market and reports whether prices and rules are working as intended., the outside firm Monitoring Analytics that audits the market, is blunt about the leading cause. It calls forecast data-center load the primary reason for the recent and expected capacity-market conditions, the tight supply-demand balance, the shortfall in cleared capacity, and the high prices. In the December 2025 auction (for the 2027/28 year) it tied 40 percent of the $16.4 billion in costs, about $6.5 billion, directly to data centers. Across PJM's last two auctions, it estimated forecast data-center load drove roughly $23 billion in market costs.

But 'leading driver' is not 'only driver,' and honest reporting has to separate the market's reaction from its underlying mechanics. Three other forces pushed the same direction. Generators kept retiring faster than new plants arrived. PJM's interconnection queueinterconnection queueThe waiting line of new power plants seeking to connect to the grid; a long backlog means new supply arrives slowly even when demand is rising., the waiting line for new supply to connect, stayed slow, so even as demand climbed, fresh megawatts trickled in. And in January 2024 FERC approved a change to how PJM credits each plant for its real contribution on peak days, a switch from an average to a 'marginal' version of a measure called ELCC; first used at the July 2024 auction, it lowered the credited capacity of some resources and tightened the math. PJM itself has pointed more to its risk-modeling update than to accreditation as the bigger technical lever. The point for a reader: data centers are the largest single cause, but retirements, the queue, and accreditation changes are genuine co-causes. This is not a one-villain story.

One more piece of context keeps the dollars honest. The numbers above are the capacity line only. Capacity is roughly a fifth to a quarter of a typical retail electricity bill; the rest is energy, transmission, distribution, and various charges. So a record capacity price is a real increase, but it moves a slice of the bill, not the whole thing.

Here is the mechanism that decides who pays. PJM does not divide capacity cost by total electricity used. It divides it by how much each customer draws during the system's few hardest hours, its coincident peak, often summarized by five peak days (the '5CP' method). Picture a gym that bills you not for your total hours but for how much of the busiest hours you took up. The more you crowd the peak, the bigger your share.

Now apply that to a data center. A data center runs flat, 24 hours a day, seven days a week. In the jargon, it has a very high load factorload factorHow steadily a customer uses power. A data center running flat 24/7 has a high load factor; a home that spikes in the evening and idles at night has a low one. High load factor means a customer is drawing power during almost every peak hour.: it is drawing power during almost every peak hour, every year. A home is the opposite, it spikes in the evening, idles overnight, and is often modest during the worst summer afternoons. Under a peak-based formula, the steady, always-on user contributes heavily to peak and therefore picks up a large absolute share of capacity cost. So the simple charge that data centers 'don't pay' does not survive contact with the formula. In absolute dollars, they pay a lot.

The data centers are not skipping the bill. They are paying a large absolute share, and still, the extra cost their arrival created is being smeared across everyone's peak.

So where is the grievance? It is marginal, not absolute. When a cluster of new data centers arrives, it lifts the whole region's forecast peak and tightens the supply-demand balance, which lifts the clearing price for everybody. That incremental cost, the part the new load created by showing up, is then recovered through the same peak-share formula applied to all customers. The newcomers pay their slice of it, but so does every existing household and business, in proportion to their own peak use. Regulators have a name for the principle this strains: cost causation, the idea that whoever causes a cost should pay it. The narrower version, marginal cost causationmarginal cost causationA narrower version of the principle: even if a customer already pays its share of total cost, the extra cost its growth adds should be charged to it, not averaged across everyone., says even a customer already paying its fair share of total cost should be charged the extra cost its growth adds, rather than having it averaged across the system. That is the design seam FERC is now poking at.

To make the abstraction concrete, look at one zone where an advocate has done the math. For the Washington, DC area served by Pepco, Synapse Energy Economics, working for the DC Office of People's Counsel, the office that represents utility customers, modeled the bill impact of the 2025/26 capacity surge. It found the capacity portion of the bill rising about 9 percent for the average residential customer, on the order of $10 a month, with the commercial class seeing a similar percentage increase.

Read this chart carefully, because it is easy to over-read. It is one advocate's estimate for one zone, tied to the 2025/26 jump, not a 13-state average, and PJM prices clear by sub-region, so a data-center-heavy zone like Dominion in Virginia can land very differently. The dollars also depend on usage: the same per-kilowatt-hour adder is a few dollars for a small apartment and thousands a year for a large commercial account. The useful takeaway is the rate, not the total: the increase arrives as a roughly flat charge per kilowatt-hour, hitting homes and businesses at about the same rate.

A flat per-kilowatt-hour increase is not neutral across incomes. Lower-income households spend a larger share of their income on energy, they have a higher energy burdenenergy burdenThe share of a household's income spent on energy bills; lower-income homes have a higher energy burden, so a flat per-kilowatt-hour increase hits them proportionally harder., so the same cents-per-kilowatt-hour adder takes a bigger bite from them than from a wealthier home. That is the distributional sting hiding inside a number that looks even-handed on a chart.

This is where the jurisdictional map matters, and where a lot of the public hope may be misplaced. FERC's 18 June orders are aimed at the wholesale, federal layer: how large loads connect to the grid, how their costs are studied and assigned, and how to prevent cost-shifting onto other transmission customers. In principle FERC could push the operators toward a wholesale rule that puts more of the marginal cost on the loads that cause it, a large-load charge of some kind. The orders even list 'preventing cost shifting and requiring transparency into transmission costs' as one of five reform categories, alongside faster interconnection and rules for co-located generation.

But the part of the bill a household sees, how cost is divided among residential, commercial, and industrial classes at the retail level, is set by state regulators, not FERC. So even a strong wholesale fix would change the size and assignment of the pie at the federal level and then hand the slicing back to the states. Translating any wholesale reform into lower household bills would still have to pass through separate state proceedings across PJM's 13-state footprint. That is a long, uneven road, and it is why a single federal order, however aggressive, does not automatically reach your bill.

It is tempting to assume the spike is temporary and will fall back, the way commodity shocks often do. Here the evidence genuinely does not say. Both the 2026/27 and 2027/28 prices cleared at a FERC-approved temporary cap, about $325 per megawatt-day, a ceiling that grew out of a 2025 settlement with Pennsylvania's governor and was approved 4-0. When a price sits exactly at its cap, the cap is not telling you the market has stabilized; it is telling you the true clearing price is somewhere above the line and hidden. PJM's own simulations make the point: without the cap, 2026/27 would have cleared around $389 and 2027/28 around $530.

So the honest answer to 'will it come back down?' is that we cannot see the real signal yet. The cap may be holding back still-rising pressure rather than marking a stable plateau. Whether this is a spike that round-trips or a step to a permanently higher level depends on supply catching up to load, new generation clearing the queue faster than data centers arrive, and that race is not yet decided.

Data centers are the largest single driver of PJM's record capacity price, but not the only one, and they are not free-riding: because they run flat around the clock, the peak-based formula already hands them a large absolute share of the cost. The real flaw is narrower and more technical. The marginal cost their growth creates, the price increase their arrival triggers for the whole region, is recovered across everyone's peak rather than charged to the loads that caused it. FERC's June orders can in principle tighten that at the wholesale level, but the part households feel is set by 13 state regulators, so relief, if it comes, is slow and indirect. And with the price pinned to a cap, even the question of whether the spike reverses is, for now, unanswerable.