Free household electricity from a data-center power fund

Our first Lab piece modeled a half-cent surcharge on data-center power, rebated to households, and showed it was small to the operator and visible to the household. This piece is the same instrument inside the Universal Income System, where the question is different: not "how do we stop the cost shift" but "what should the machine economy pay for." The answer the design gives is the electric bill.

The power fund

Data centers pay 12 cents per kilowatt-hour on their consumption into a power fund. The fund pays household electric bills, through the banks like everything else, starting with the lowest-income households and working up until every home's bill is paid. At the model's load path (300 TWh in 2027, 650 TWh in 2030, then 8 percent a year) the fund raises $36 billion in 2027 and $248 billion in 2045, against household bills of about $270 billion a year³ growing 1 percent. Coverage goes from 13 percent of homes in 2027 to 55 percent in 2040, 77 percent in 2045, and all of them by about 2050. The surcharge is modeled as flowing through state utility commissions, which is where it can be enacted first (the rollout lists regional surcharges among the earliest steps), and the model treats the fund as national for simplicity.

Why 12 cents and not half a cent

At 12 cents the surcharge roughly doubles a data center's power cost. That is a deliberate choice and the design should defend it rather than hide it. The argument is that in a strong-AI economy the data center is the factory, its output is what the universal income is a claim on, and its power bill is the most direct place to attach that claim: it scales with the machines' work, it cannot be moved offshore without moving the latency with it, and it is metered. A surcharge that high will shape siting, and the design wants it to: toward regions with cheap, clean power and away from grids where households are already paying for upgrades. West Virginia, with its generation and its microgrid-district law, is a place that choice favors.

The honest counterargument is that 12 cents might slow the build-out the whole system depends on. The model does not include that elasticity. If the strong-AI path needs cheap power to happen, the rate should start lower and ratchet up with the load, and the slider below lets you see what a lower rate covers.

The risk the model names

Data-center load growing 8 percent a year forever is a lot of electricity. By 2045 the model has it near 2,000 TWh, about half of today's total U.S. consumption, and the fund's revenue depends on that load existing. If the machines get efficient faster than they multiply (and they will try, because at 12 cents they have every reason to), the fund shrinks and coverage stalls. That is not a failure of the design; it is the surcharge doing its job of pricing power into the machine economy's decisions. It just means "free electricity for every home by 2050" is a projection of a load path, not a promise.