PJM's capacity auction price has gone from $29/MW-day to $325 in four years. Data centers drove 46% of the cost — $29.4 billion — across those auctions, and the grid's independent watchdog says ratepayers are footing the bill for power plants that may never be needed. Here's the full timeline, who pays, and the three reform proposals on the table right now.
On July 14, 2026, PJM Interconnection — the grid operator for 67 million people across 13 states and Washington, D.C. — published the results of its fifth consecutive capacity auction shaped by data center demand. The numbers tell a story that every ratepayer in the mid-Atlantic should understand, because the cost lands on their bill whether they've heard of a "base residual auction" or not.
PJM's capacity market pays power plants to exist — to be available on the hottest summer afternoon even if they don't run. The clearing price determines how much consumers pay for that insurance. Here is what has happened to it:
| Delivery year | Clearing price ($/MW-day) | Change | Total cost |
|---|---|---|---|
| 2024/2025 | $28.92 | — | ~$5 B |
| 2025/2026 | $269.92 | +833% | ~$14.7 B |
| 2026/2027 | $329.17 | +22% | ~$16.4 B |
| 2027/2028 | $333.44 (price cap) | +1.3% | ~$16.4 B |
| 2028/2029 | $325.00 | −2.5% | ~$16.4 B |
Read that first jump again. In a single year, the price that sets roughly 20% of your electricity bill rose by a factor of nine. And it has stayed there for four straight auctions, pinned at or near the FERC-approved price cap.
The price cap is doing real work here. Without it, the 2028/2029 auction would have cleared at roughly $555/MW-day RTO-wide — and $777/MW-day in the ComEd zone around Chicago. The cap has saved consumers an estimated $45 billion over four auctions. But it hasn't solved the underlying problem — it's just hidden how bad the underlying problem is.
Two things happened in 2025 that had never happened before: PJM's load forecast added tens of thousands of megawatts of data center demand, and the clearing price hit the administrative ceiling. By the 2027/2028 auction it got worse — PJM fell 6,625 MW short of its own reliability target, the first RTO-wide capacity shortfall in the market's history.
PJM's Independent Market Monitor — the neutral referee Congress created to keep wholesale markets honest — has now quantified data centers' share of the bill. The numbers are stark.
| Metric | Value |
|---|---|
| Data center share of the latest auction | $6.3 B (38% of $16.4 B) |
| Data center share over four auctions | $29.4 B (46% of $63.6 B) |
| Share of projected load growth from data centers | 94% |
| Customers who pay these costs | 67 million across 13 states + D.C. |
Market Monitor president Joseph Bowring put it plainly: "This is not something the data centers are actually paying themselves. This is a cost being imposed on all customers in the PJM footprint."
That distinction is the whole story. The capacity market doesn't charge data centers directly for the demand they add. It charges everyone — residential, commercial, industrial — based on their share of system peak load. When a 500 MW data center raises the system peak, your household's fractional share of the new, larger peak becomes your new capacity bill — even though your usage hasn't changed.
Bowring's diagnosis: "PJM is continuing to act like it's business as usual" when what the grid is experiencing is "a paradigm shift."
The capacity charge is typically buried on page two of your bill, lumped into "supply charges" or "generation service." But it's real money:
Virginia — which hosts more data center capacity than any other state — and northern Illinois are ground zero. But capacity charges are allocated across the entire RTO, so customers in Ohio, Indiana, Kentucky, and the Carolinas are paying too.
You might think a price cap is good news. It isn't, for two reasons:
1. The cap isn't low. At $333.44/MW-day, a ratepayer's capacity bill is already 11 times what it was in 2024. The cap prevents further increases; it doesn't roll back the ones already baked in.
2. When supply can't clear above the cap, it doesn't show up. In the 2027/2028 auction, 809.6 MW of capacity offered to sell but was priced above the ceiling. It couldn't clear. Result: the grid fell short of its reliability margin for the first time. The cap kept the price from rising — and in exchange, it kept the power from arriving.
The 2028/2029 auction was even worse: 6,831 MW short, with a reserve margin of just 14.7% against a 20% target — the lowest ever recorded. Only 525 MW of new generation cleared, down from 774 MW the year before.
PJM has asked FERC to approve a temporary "reliability backstop" auction to fill the gap. But that doesn't change the structural problem: the capacity market was designed for a world where demand grew 1% a year. Data centers blew past that assumption so fast that the market's own safety valve is now working against reliability.
The debate has moved from "is this a problem?" to "which fix?" Here are the three live proposals:
1. Separate auction for large loads (Bowring / Market Monitor) Data centers that can't self-supply would bid in a segregated 15-year capacity auction. Their costs would be allocated to them, not to residential ratepayers. Bowring: "There's only one way to do what hyperscalers agree is the right thing to do, and that is to run a separate auction."
2. Illinois POWER Act (state legislation) Would require data centers to pay their own interconnection costs, bring enough new capacity to cover their own load, and supply new clean energy or reimburse the state. Failed to pass in spring 2026, but the Citizens Utility Board says the fight is "far from over" and will return.
3. Federal Power for the People Act Would direct FERC to ensure data centers fund the local transmission upgrades their load requires — rather than spreading those costs across the RTO.
All three share one principle: the entity that causes the cost should bear the cost. That principle already applies to transmission interconnection for new generators. It has never been applied to new load at this scale, because load at this scale didn't exist until AI.
If you live in PJM territory — Delaware, Illinois, Indiana, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, West Virginia, or D.C. — here's how to use this:
Find the capacity charge on your bill. It's usually inside "supply" or "generation service." Your utility's tariff schedule will break it out. Once you see it, you'll understand why your bill moved.
Ask your state PUC one question: "What is my utility's plan to ensure data center load growth doesn't raise residential capacity costs further?" File it in writing. It becomes part of the record in any future rate case.
Support cost-causation reform. Whether it's the POWER Act in Illinois, the Market Monitor's separate-auction proposal at FERC, or your state's own version — the policy ask is the same: large loads should pay for the capacity they require, not externalize it onto 67 million households.
Show up at PJM stakeholder meetings. PJM's Board is actively deliberating on data center demand management right now. Public pressure from ratepayer advocates — including the Citizens Utility Board, NRDC, state consumer advocates, and individual residents — is what moves the timeline from "under study" to "implemented."
The capacity market is the most expensive part of your bill that you've never heard of. Now you've heard of it — and you know who's driving the price.
Use the Your Utility Bill tab for a full breakdown of how capacity charges flow from the auction to your monthly statement, and the States & Officials tab to find your PUC's complaint portal.