A data-center power surcharge, returned to households
Every forecast of U.S. electricity demand now bends upward because of data centers. Lawrence Berkeley National Laboratory's December 2024 report put data-center consumption at 176 TWh in 2023, about 4.4 percent of all U.S. electricity, and projected 325 to 580 TWh by 2028, which would be 6.7 to 12 percent of the national total1. Growth of that size needs new generation, new transmission and new substations. Under most utility rate designs, the cost of that capacity goes into the rate base and is recovered from every customer class, which means a household's bill rises to serve a load it will never use.
What we propose
A surcharge of a few tenths of a cent to a few cents per kilowatt-hour on customers above a large-load threshold (say 50 MW of contracted demand), collected by the utility and either rebated to residential customers as a line item or placed in a fund restricted to the capacity those loads require. The mechanism is deliberately dull: it uses the meter that already exists, it does not pick winners among operators, and it is visible on both sides of the ledger. Operators can plan around a known per-kWh number far more easily than around an open-ended rate case.
The model
Move any slider. Every default is sourced below; every number on the page recalculates. Nothing here is a forecast, it is arithmetic on stated assumptions.
Why the numbers land where they do
At the defaults (325 TWh, half a cent, everything rebated) the surcharge raises about $1.6 billion a year and returns about $12 to each household. That is modest, under one percent of an average bill. The point of the default is not the size of the check; it is that the direction of the money is reversed. The same half-cent is a 6 percent increase on an operator paying 8 cents, which is inside the year-to-year variance of wholesale power and well below the premium operators already pay for clean-energy contracts.
Push the load to the high end of LBNL's range and the surcharge to two cents and the numbers become large: about $13 billion a year, roughly $100 per household, a quarter of the operator's power price. We do not recommend that setting; we show it so the trade-off is on the page. A surcharge that high would move siting decisions, and siting is a legitimate thing for a state to want. That is a choice for a legislature or a commission, not for a model.
Assumptions, stated
| Assumption | Why it matters |
|---|---|
| All data-center load is above the large-load threshold. | Overstates revenue slightly. Smaller colocation sites would fall below a 50 MW line; LBNL's total includes them. |
| The surcharge is not passed through to residential customers. | Data centers sell compute nationally and globally, so there is no local retail channel to pass the charge to. This is the strongest argument for the design. |
| Rebate is a flat per-household amount. | A flat rebate is progressive relative to bills; a per-kWh rebate would favor larger homes. Either is implementable. |
| National figures, one price. | Real tariffs are set state by state and utility by utility. The PLACES platform carries county-level data; a state version of this model is the natural next piece. |
| Static load. | A surcharge might slow growth or shift it between states. We make no claim about elasticity. |
What this means for West Virginia
West Virginia is actively recruiting data centers and in 2025 created a framework for data-center microgrid districts that can generate their own power outside the utility's rate base. That is one answer to the cost-shift problem: if the load brings its own generation, it does not load the rate base. The surcharge is the other answer, for loads that connect to the grid. A state can do both. What it should not do is neither, which is the default under existing tariffs. The PLACES platform publishes county-level electricity, employment and site data for every West Virginia county, and we will add a state version of this model, with Mon Power and Appalachian Power prices, as the next piece.
Sources
- 2024 United States Data Center Energy Usage Report — Lawrence Berkeley National Laboratory, December 2024 · original · data-center load 176 TWh in 2023 (4.4% of U.S. electricity); 325–580 TWh by 2028 (6.7–12%). Not yet locked.
- Electric Power Monthly, Table 5.6.A (average retail price by sector) — U.S. Energy Information Administration · original · residential and industrial average prices; the defaults below round the 2025–26 figures. Not yet locked.
- Electric Sales, Revenue, and Average Price (residential sales and customers) — U.S. Energy Information Administration · original · about 1,500 TWh of residential sales a year and roughly 10,800 kWh per residential customer. Not yet locked.
- Households and Families, American Community Survey 1-year — U.S. Census Bureau · original · about 131 million households (2023). Not yet locked.
- Large loads and cost allocation: proceedings and tariffs — state public service commissions (various) · several states have opened large-load tariff dockets since 2024; the model is tariff-neutral and does not depend on any one of them.
Numbered references in the text point to this list. "Locked" means we hold a dated copy of the file on the PLACES platform so the figure can be checked later even if the publisher changes the page.