Chatbots don’t run on magic. They run on your money.

Imagine someone walks into your town with a proposition: Rezone large swaths of residential and farmland. Hand out tax breaks. Let us build ugly, noisy facilities for chatbots — facilities that will devour nearly a quarter of the power supply.

Then, before you run him out of the room, he adds a final promise: Do not worry. We will pay our own way.

Argue about the projections if you want. Do not tell the public they will not pay more for data centers. They already do.

That is the rope-a-dope Americans are supposed to accept from the government-tech oligopoly, even as politicians insist that data centers will not cost the public a dime.

Sensing a growing backlash against the data-slop colonization of rural America, President Trump promised during the State of the Union that every data center company will pay its own way. Awareness of the problem helps. The president’s pledge does not.

Facts on the ground point in the opposite direction: consumers already pay for data centers, the economics make “paying their own way” implausible at scale, and the industry fights efforts to put that promise into law.

The scope of the problem

The hyperscale build-out being stacked on top of roughly 4,000 existing facilities is not a “burden” on the grid. It is an industrial-scale demand shock.

MIT Technology Review reports that AI alone could soon consume as much electricity as 22% of all U.S. households. Boston Consulting Group projects data center energy needs of up to 1,050 terawatt-hours annually by 2030 — about 120 gigawatts on average. That figure exceeds current U.S. nuclear capacity by roughly 23%.

To put it in plain terms, the United States has about 97 gigawatts of nuclear capacity across 94 reactors. If the high end of OpenAI’s hyperscale ambitions materializes, those facilities alone would require roughly 36% of total U.S. nuclear capacity.

Now scale it out. Clearview estimates that if the 680 planned data centers get built and become operational, they would require the energy equivalent of 186 large nuclear power plants.

That should end the fantasy that these companies can “pay their own way” while drowning in debt, burning cash, and chasing thin margins.

These are not last decade’s data centers, either. Bloomberg reports that only 10% of facilities today draw more than 50 megawatts. Over the next decade, the average new facility will draw well over 100 megawatts. Nearly a quarter will exceed 500 megawatts, and a few will top 1 gigawatt.

Electricity is only the first bill. This demand shock forces major grid upgrades: transmission lines, transformers, substations, and capacity expansions. Utilities do not eat those costs. They pass them on to taxpayers — that is, us.

Wood Mackenzie estimates that AI-driven build-outs will push transformer demand beyond supply by about 30% this year, driving costs up and delaying projects. Consumers will pay for that, too.