Trump Called AI Data Centers the New Oil. The Electricity Bill Says Otherwise.

The AI investment cycle has moved one layer down the stack. After two years of capital flooding into chips and frontier labs, the buildout is now a power, electrical-equipment, and data-center real-estate story, and the ratepayers funding the next layer are starting to push back.

At the All-In Summit, Donald Trump called data centers "the oil of the next 20-25 years," arguing they will make states and communities wealthy. Nvidia CEO Jensen Huang told CNBC the broader thesis was right and put a number on the capital flow: about $400 billion in venture financing has moved into AI companies over the past six months alone. The markets pricing that thesis are not the chip makers.

A single large AI data center can draw as much electricity as 100,000 homes, according to the AOL aggregation of the summit remarks. The World Resources Institute projects that, taken together, U.S. data centers could account for as much as 12% of national electricity consumption by 2028, up from a low-single-digit share today. That demand shows up as a queue at the substation, a transformer on order, and a rate case at the state public utilities commission.

The investors positioning for it are not just chip-adjacent. The next layer sits with independent power generators such as Vistra and Constellation Energy (utilities that own nuclear and gas plants), electrical-equipment manufacturers such as GE Vernova (turbines and grid hardware) and Eaton (switchgear and power management), and the real-estate investment trusts that own the buildings themselves. Huang's framing, in his Mad Money interview, was that AI is a "new industrial revolution" requiring chips, manufacturing, electricity, and physical infrastructure in roughly equal measure. The market is now starting to price the latter three.

Electricity prices rose 11.5% in 2025, and rate increases tied to data-center load totaled more than $60 billion last year, according to the Energy Efficiency and Sustainable Infrastructure group. The pattern is concrete enough that regulators are beginning to write new rules. PPL Electric's $275 million rate case settlement in Pennsylvania included a dedicated data-center tariff that bills the largest new customers for grid upgrades. Virginia's Dominion Energy has moved to a similar large-load tariff framework, documented in a June 2026 Columbia Climate Law analysis.

Local opposition is following the same logic, with 2026 ballot measures tracked by Ballotpedia as the next front. Once a data center is built, the long-term operations headcount is thin, often reported as fewer than 150 people and sometimes as few as 25, well below the construction-era peak. The headline revenue figure of $162 billion in government tax receipts by 2023, often cited to argue that data centers pay for themselves, shrinks sharply once state tax abatements are netted out, leaving counties to absorb the water and grid upgrades. Community organizing around siting is now tracking that gap.

The Trump-Huang thesis is not wrong. The capital is arriving, and the buildout is real. The question now is whether the rate base can carry the load without the bill showing up at a house that did not place the order, and whether the 2026 rate cases and ballot measures are the start of a tariff regime that prices large loads at their true grid cost, or another round of socializing it. The answer will not come from a chip announcement. It will come from a utilities commission order, a transformer delivery date, and a vote on a November ballot.