Large customer electricity payments during sustained underuse

SMUD, Sacramento's public utility, is writing new rules for data centers and other customers who use more than 50 megawatts. Its Board will take them up in November 2026. My paper asks one question: when a big customer reserves a lot of grid capacity and then uses far less than it asked for, who pays for the difference? SMUD currently bills these customers through its Site Infrastructure Charge, which is based on their highest demand over the past twelve months. I compare that to a few other options: billing on the capacity the customer signed up for, requiring minimum contract lengths, and setting a floor on what customers must pay. I test each option in four situations: the customer's use falls, the customer ramps up slowly, the customer leaves early, or the customer never connects. Over ten years, billing only on past peak use gets SMUD about 65% of what it would collect if the customer paid for its full requested capacity after a drop in use. If the customer only ever reaches half of what it requested, SMUD gets about 45%. If the customer leaves at bill 36, SMUD gets about 30%. If the project never connects, SMUD gets nothing. I also coded the published billing rules of six other California public utilities. They don't do any better, except in the case where use falls. California's new law, AB 2383 (Chapter 435, Statutes of 2026), requires minimum payments, exit fees, and collateral for data centers served by investor-owned utilities. But the signed text doesn't cover public utilities like SMUD. The paper ends with seven questions SMUD should answer as it sets its rules. It is important to note that the scenarios are illustrations and not real contract data because those contracts aren't public. I have included all the code and data so anyone can check the results. Version 1.3 adds answers from SMUD’s pricing team (October 21, 2026) confirming the twelve-month billing window and that contract capacity is set in customers’ tailored agreements.