On March 4, 2026, the CEOs of Amazon, Google, Meta, Microsoft, Oracle, OpenAI, and xAI stood in the Eisenhower Executive Office Building and signed a document called the Ratepayer Protection Pledge. President Trump held it up for the cameras and promised it would bring household utility bills down "very substantially." The companies committed to building, buying, or bringing their own power for the AI data centers they were racing to construct, covering the cost of grid upgrades themselves, and — the specific line that mattered — not passing those costs on to ordinary electricity customers.
Four months later, ratepayers across the PJM grid region are absorbing an estimated $23 billion in data-center-driven cost increases through 2028, according to the market's own independent monitor. A Manassas, Virginia homeowner named John Steinbach opened his January electricity bill to find a $281 charge, roughly triple what he normally pays. Virginia just became the first state in the country to impose a direct tax on data-center electricity consumption, because the federal pledge didn't stop the bills from rising. New York banned new large-scale data center construction outright. And on July 13, Reuters reported that the White House is quietly planning a second, expanded version of the pledge — this time bringing in the utilities themselves, an admission that the first version left out the parties who actually set the rates.
The easy read is that Big Tech made a promise and broke it. The more useful read is that the promise was never capable of being kept, because the mechanism that determines who pays for grid infrastructure was never part of the deal in the first place. The pledge was a press conference. The rules that govern electricity pricing are federal tariffs. Those are two entirely different documents, and only one of them controls what shows up on a monthly bill.
A pledge with nothing underneath it
The Ratepayer Protection Pledge is, on paper, comprehensive. Signatories agreed to finance new power generation, cover the full cost of transmission and delivery upgrades tied to their facilities, negotiate separate rate structures with utilities and state regulators, and pay for contracted power capacity whether they use it or not. It reads like a company voluntarily agreeing to internalize every cost its data centers create. The problem is that voluntary internalization isn't how electricity pricing works. Grid costs in most of the country are allocated through tariffs filed with and approved by the Federal Energy Regulatory Commission — dense, technical documents that predate the AI buildout by decades and were written for a world of relatively predictable, evenly distributed demand growth, not gigawatt-scale data campuses clustering in single counties.
FirstEnergy, a utility operating across the PJM footprint, told FERC directly that the existing rules require grid costs to be socialized across all customers regardless of what any individual company wants to pay. In other words: even a data-center operator that genuinely intends to cover 100% of its own infrastructure costs cannot currently do so cleanly, because the tariff structure spreads those costs across everyone connected to the grid unless a specific rule says otherwise. Seven companies could sign a pledge in front of the president promising to pay their own way, and the tariffs that actually move money from data centers to ratepayers — or fail to — would not change by a single word.
That gap is exactly why FERC's most significant action on this issue didn't come from the White House ceremony at all. On June 18, 2026, more than three months after the pledge was signed, FERC issued formal show-cause orders to all six major regional grid operators — PJM, MISO, SPP, CAISO, ISO-NE, and NYISO — instructing each to justify why its existing large-load interconnection rules are "just and reasonable" or propose reforms within 60 days, a deadline landing around August 17. That same day, a bipartisan pair of House members introduced the Ratepayer Protection Act, a legislative attempt to write into law what the March pledge only asked companies to volunteer for. Both of those are structural interventions aimed at the tariff layer. The pledge was not.
What the gap looks like on a bill
The distance between voluntary commitment and binding rule shows up first in the regions where data centers are most concentrated. Virginia's data-center corridor now accounts for roughly 40% of the state's total electricity consumption. In the PJM capacity market, which covers 13 states including Virginia, prices rose 174% for the 2025–26 delivery year, and Baltimore-area residents saw their average monthly bills climb by more than $17 after that auction, with the 2026 auction adding several dollars more. Steinbach's $281 bill wasn't an outlier so much as an early, visible data point in a pattern regulators were already tracking. A January 2026 survey found that nearly three-quarters of Virginia voters blamed nearby data centers for their rising costs, and by May, national polling from Consumer Reports found 75% of Americans doubted tech companies would actually follow through on pledges like this one.
States didn't wait to find out. Virginia's legislature and Governor Abigail Spanberger settled a months-long budget fight by enacting a $0.011-per-kilowatt-hour tax on data-center electricity consumption, effective July 1 — the first tax of its kind in the country, applied whether the power comes from the grid, a competitive retailer, or on-site generation. It's a blunt instrument, and it doesn't directly lower anyone's household bill, but it does something the March pledge never did: it creates a mandatory, enforceable charge tied specifically to data-center power use. New York went further still. Governor Kathy Hochul signed a statewide moratorium on permitting new data centers drawing 50 megawatts or more, the first freeze of its kind in the nation — though it can't touch the 48 projects already sitting in the interconnection queue, representing roughly 12 gigawatts of future demand that will accumulate grid-upgrade costs under the same socialized-cost rules FirstEnergy described to FERC.
The sequel confirms the diagnosis
The clearest evidence that the original pledge was structurally incomplete came from the White House itself. Reuters' July 13 report on a planned second summit describes an expanded framework that would include not just hyperscalers but the utilities that supply them power and the colocation developers who build data centers on their behalf. That's a tacit acknowledgment that the first pledge left out exactly the parties with the legal authority to set rates. A commitment signed only by the companies buying electricity, without the utilities and regulators who determine how its cost gets allocated, was never going to be enforceable — it was a promise from the demand side of the market about pricing decisions made on the supply side.
None of this requires assuming bad faith from the signatories. Several of the companies involved likely do intend to cover their own infrastructure costs where the rules currently allow it. But intention was never the binding constraint. The constraint is a set of federal tariffs that predate the AI boom, that spread new grid costs across existing ratepayers by default, and that took FERC until mid-June — more than three months after the cameras left the Eisenhower building — to even begin formally revisiting. A voluntary pledge photographed at a podium was always going to move slower than a wholesale capacity auction that resets every year. The tell isn't that ratepayers are still absorbing costs in July. It's that the administration is now trying to rebuild the pledge with the one set of parties, the utilities, whose absence from the original signing made the promise unenforceable from the start.
