Opinion: utilities count data centre costs in rate cases but leave out the extra sales
Edward Burgess, a partner at the consultancy Current Energy Group, argues in an opinion piece published by Utility Dive on 12 November that some US utilities are setting rates in a way that keeps the benefit of data centre growth from reaching existing customers. It is an argument from an expert witness who has testified against one of the utilities he names, and it should be read as such, but the mechanism he describes is simple to check.
Rates are roughly costs divided by sales. If a large new load raises sales faster than it raises costs, average rates should fall. Burgess's complaint is that in several recent cases utilities included the costs that new data centres trigger while keeping the sales figure in the formula at an old level.
His main example is Duke Energy in North Carolina, which proposed raising customer rates by more than 15% over two years. According to Burgess, the calculation held kilowatt-hour sales essentially at 2024 levels even though Duke had contracted hundreds of megawatts of new data centre load and counted the related costs. Current Energy Group testified for the state attorney general's office and found that 54% of the requested increase, about one billion dollars a year, would disappear if sales were set to Duke's own planning assumptions. Correcting sales and the return on equity together, he writes, turned the 15% increase into a 4% decrease even with large new grid spending.
He cites two more cases. In Missouri, Evergy Missouri Metro included the cost of new gas plants for data centres without the additional sales to them. In Louisiana, higher sales from a new Meta data centre could have produced a refund over three years, while Entergy proposed spreading that benefit over 20 years.
Burgess calls the result reverse regulatory lag: rates are not cut fast enough to reflect rising revenue, while accelerated recovery mechanisms such as Construction Work in Progress let utilities collect for new plants years before they run. In Duke's original proposal, customers would pay for new gas plants four years before they operate and six years before the data centre load enters the sales figure. He asks regulators to require current sales forecasts in every rate case and to reconsider accelerated recovery.

Why it matters
The debate about whether data centres raise or lower bills often stays abstract. Burgess turns it into a question about two numbers in a rate filing. Utilities will contest his figures, but the test he proposes, using the same forecast for rates that is used for building plants, is one regulators can apply case by case.