FERC Gave the Grid 60 Days to Stop the Data-Center Power Subsidy. Three Years of It Already Cleared.

On June 18, 2026, the Federal Energy Regulatory Commission (FERCFERCThe Federal Energy Regulatory Commission, the federal agency that regulates interstate electricity markets and approves the rules grid operators use to set prices.), the federal agency that polices interstate power markets, issued show-cause orders to all six regional grid operators, including PJM InterconnectionPJM InterconnectionThe largest U.S. regional power-grid operator, coordinating electricity for about 67 million people across 13 states and Washington, D.C., from the Mid-Atlantic into the Midwest., the system that keeps the lights on for more than 67 million people across 13 states and Washington, D.C., stretching from New Jersey to Illinois. The orders give the operators 60 days, until about August 17, to either defend their current rules or file new ones for how very large new electricity users, above all data centers, connect to the grid and how their costs are split among customers.

FERC framed the action as a push to speed the connection of large new loads. But a core piece of it is cost allocationcost allocationThe rules deciding which customers pay for which grid costs, the lever FERC's order actually pulls by aiming to shift more of the bill onto large new users.: writing rules so the households and businesses already on the grid do not end up paying for the upgrades needed to serve a new hyperscale data center. Consumer advocates have spent more than a year arguing that this is exactly what has been happening.

PJM runs a capacity marketcapacity marketA market where the grid pays power plants to promise they will be available on the hottest or coldest days, separate from paying them for the electricity they actually generate. Think of it as a retainer fee for standby capacity.: alongside paying power plants for the electricity they actually produce, it pays them a separate fee just to promise they will be available on the worst demand days of the year. Think of it as a retainer for standby power. The price of that retainer is set in a once-a-year auction called the Base Residual Auction, normally held about three years before the power is needed. Because the auction happens years ahead and locks in a fixed price, it is a forward auctionforward auctionAn auction held years before delivery that locks in a fixed price now for power supplied later, so the cost is contracted long before it appears on a bill.: the cost is contracted long before it ever shows up on a monthly bill.

That timing is the whole story. By the time a regulator acts, the auctions that set the next several years of bills have already cleared. So the question is not whether data centers are driving up capacity costs. The monitor's own numbers say they are. The question is narrower and more consequential: of the data-center costs already flowing into bills, how many can the June 18 order actually remove, and how many are locked?

A show-cause ordershow-cause orderA FERC directive telling a grid operator to either justify its current rules or propose new ones within a set deadline, here 60 days. It produces rules going forward, not refunds for the past. is a forward-looking instrument. It tells a grid operator to justify its existing rules or file new ones, and whatever rules result apply to future auctions. It does not order refunds for prices that have already cleared. That limit is reinforced by a legal principle called the filed-rate doctrinefiled-rate doctrineA legal principle that once a rate (here, an auction's cleared price) is on file and approved, it generally cannot be changed retroactively. It is why a cleared capacity price is hard to claw back.: once a rate is on file and approved by FERC, it generally cannot be changed after the fact. A cleared capacity price is, in effect, a filed rate, which is why clawing one back is so hard. The Third Circuit, a federal appeals court, applied that doctrine to block a retroactive change to a PJM auction rule.

Run the calendar forward and the consequence is stark. The 2025/26, 2026/27, and 2027/28 delivery years all cleared at auctions held between July 2024 and December 2025. Those prices are set. Through the show-cause route, the dollars the June 18 order can strip from them is zero.

The cleanest way to see this is a ledger. Each delivery yeardelivery yearThe 12-month period (June through May) for which a capacity auction's price applies. The 2026/27 delivery year runs June 2026 to May 2027. traces back to exactly one auction, on one date, at one cleared price. Once you line them up, the headline figures snap into place, and the famous 63% turns out to belong to the 2025/26 auction (which cleared at $269.92, a roughly nine-fold jump from $28.92 the year before), not to the $329.17 price that is making news now.

Three of these years are marked No: locked. The 2028/29 auction is clearing right around now, before any rule filed in August could be reviewed and made effective. The first year a new rule could plausibly touch is 2029/30, with bills starting in June 2029.

It is tempting to read the data-center shares as a falling trend: 63%, then 82%, then 40%. They are not a series. Each is the market monitor's modeled estimate, and each measures a different thing. The 63% is data centers' share of the price increase in the 2025/26 auction. The 82% is how much higher the 2026/27 auction's total cost was because forecast data-center load was included (the cost would have been about 82% lower without it). The 40% is data centers' share of the total cost of the 2027/28 auction. Different numerators, different denominators.

These are modeled estimates, not metered allocations. No meter splits a capacity bill into a 'data-center' half and a 'household' half.

Holding that distinction matters because it stops a reader from inferring a story the data does not tell, that the data-center share is shrinking, or growing, in a smooth line. What the numbers do agree on is direction: forecast data-center demand is the single biggest force pushing these auctions higher.

Suppose FERC's effort fully succeeds and future rules shift the data-center share of costs onto data centers. Would bills fall to where they were before the surge? Mostly no. The monitor built a counterfactualcounterfactualA modeled estimate of what would have happened under a different assumption, here, what capacity costs would be if new data-center demand were removed. It is a calculation, not a measured split. for the 2026/27 auction, a modeled estimate of what the cost would have been with the forecast new data-center load stripped out. Total procurement cost falls from about $16.1 billion to about $8.85 billion, a difference of roughly $7.3 billion. Real money. But $8.85 billion is still far above the pre-surge norm.

The reason is supply, not just demand. Power plants are retiring faster than new ones connect, so capacity is scarce. Scarcity sets a high floor under the price no matter who is bidding. Cost allocation, the lever FERC is actually pulling, decides who pays the bill. It does much less to shrink the bill itself.

Lay the regulatory clock over the auction clock and the mismatch is obvious. Every auction that sets a currently billed year cleared to the left of the June 18 order. The order's own 60-day deadline lands in August 2026; even on an optimistic path, the rules that follow have to be drafted, reviewed, and approved before they govern an auction. By then the 2028/29 auction has cleared too. The first delivery year a new large-load rule could plausibly reshape is 2029/30.

So the honest answer to the headline promise is that households and businesses keep paying the data-center share for at least three delivery years, whatever the operators file in the next 60 days.

There is a single path that could reach into an already-cleared auction, and it is not the show-cause order. Section 206 of the Federal Power Act lets FERC declare an existing rate unjust and unreasonable and replace it. Whether that power can reach a price that has already cleared is contested. In a January 13, 2026 decision involving PJM's 2024/25 auction re-run, the D.C. Circuit held that FERC had wrongly concluded a Third Circuit ruling foreclosed its Section 206 authority, and sent the question back to FERC to decide what relief, if any, that section allows.

That caveat aside, the central finding holds. The instrument that grabbed the headlines, the 60-day order, arrives after the costs it targets were already contracted. The subsidy, if that is the word for it, was locked in at auctions that closed months and years ago.