The House Just Decided Who Pays for the AI Grid — and It Isn't You

This week the House is supposed to vote on a bill with an audience-friendly name — the Ratepayer Protection Act — before it breaks for the midterm campaign trail. It is billed as a way to keep AI data centers from driving up your electric bill, and the politics write themselves: families out tonight, Big Tech picking their pockets with a server farm. But read the plumbing and the bill is doing something narrower and more interesting than protecting you. It is re-deciding who on the grid has to act first when the numbers don't add up.

That matters because the whole reason data centers can raise your bill is not that they buy your power. They buy it by the megawatt, in bulk, at negotiated prices — usually cheaper than you pay. The problem is what has to exist before the first chip powers on. To serve a data center, a utility commits to building substations, high-voltage lines, and often whole gas plants, years before the facility opens and long before anyone knows whether it ever will. That is the accounting entry the whole story turns on: under the regulated utility model, a utility books that spending into its "rate base," the asset base on which regulators guarantee a return. Then it recovers the cost — including a profit — through rates spread across everyone on the system.

So when a data center is delayed, shrunk, or cancelled outright, the sunk infrastructure does not vanish. It becomes what the industry calls a stranded cost, and regulators allow the utility to sweep it onto the only customers who cannot leave: households. The geopolitical drama is theater; the resident of a rate district is the real forced actor, captive, billed for infrastructure whose revenue never showed up. That is the hidden subsidy that lets hyperscalers pay a discount while your line item grows.

The numbers behind it are big enough to be political. The Energy Information Administration data shows average residential prices up about a fifth since 2021; PJM, the grid operator serving the mid-Atlantic data-center belt, saw utilities pass $4.3 billion in transmission costs and $7.3 billion in generation costs to ratepayers in 2024 alone. Data centers are projected to triple from about 3% of U.S. electricity in 2024 to roughly 10% by 2030, and one academic model has individual regions' wholesale prices rising as much as 57% by then depending on how the load lands.

Here is what the Ratepayer Protection Act actually does, and it is more precise than the slogan. It tells state utility commissions that any "large load" customer — defined at 100 megawatts or more, which is the shape of a hyperscale data center — must pay the full incremental cost of the new generation and transmission built to serve it, must post financial assurances covering that infrastructure, and must carry the fixed cost of its reserved capacity even if it never uses it. That is the "used or not" contract. Take the promise, regulators say, and we will hold a seat in the queue for you; back out, and it is your balance sheet, not your neighbor's.

The fight over scope is telling. As originally written the bill swept in any big industrial user. During markup it was narrowed to single out data centers — and the data center trade group, which had supported the original, reversed and complained. The domestic steel industry, by contrast, endorsed the change precisely because it no longer had to carry the same exposure. Name the beneficiaries and you see the mechanism: this is not a cap on AI or a ban on anything. It is a rule about which class of buyer books the risk of a speculative buildout.

The market has already run the experiment and shown what cost allocation does to demand. Ohio, acting alone in 2025, began putting big loads through a tariff requiring them to pay for most reserved capacity even if unused, commit to eight years, and post collateral. Roughly 30 gigawatts of speculative interest collapsed to 5.6 gigawatts that actually signed. Virginia, the data-center capital, is moving toward requiring qualifying customers to commit for fourteen years and pay for at least 85% of contracted transmission and distribution and 60% of generation regardless of use. The load does not disappear when you stop offering it a free option that households underwrite; the unfunded wishful part of it does.

Now the investment read, which is where this gets useful rather than just angry. The utilities rally is one of the biggest compensated trades of the past two years — the sector is up roughly a third in a year and now trades near 19 times earnings, refashioned from a bond proxy into a growth story on the back of AI power demand. But remember the fixture underneath: a regulated utility earns a guaranteed return on a growing rate base, so more grid buildout is more guaranteed earnings no matter which customer ultimately foots the bill. The strategic issue for a utility investor has never been the buildout; it is regulatory and political durability — whether commissions keep allowing cost recovery, or affordability anger slows it into "regulatory lag."

Seen that way, a law forcing data centers to sign used-or-not agreements with financial assurances actually de-risks the rate base. It converts a soft, partially household-backed promise into a hard, corporate-backed one and removes the stranded-cost tail that is the real threat to a utility's equity. The environments to worry about are the ones where the bills do nothing and the strain builds until a commission slams the brakes. What the House vote is really telling you is that the cross-subsidy that quietly fattened the AI buildout — cheap power carried on household credit — is being clawed back at the political level. That is a modest headwind to hyperscaler landed power costs and a mild tailwind to the utility that secured the hard contract.

So read the theater, then read the entry. The forced actor is being moved from the family in the rate district to the company with the servers. For the retail investor, that is not the apocalypse the slogan implies and not the free lunch the lobby once enjoyed. It is a repricing of who shoulders the risk of the largest grid buildout in a generation — and a reminder that in a regulated market, the fight over the ledger is the trade.