As data centres rapidly expand due to the artificial intelligence boom, concerns are growing in the United States that power facility costs could be passed on to ordinary households through electricity bills.

The issue is who bears the cost of generation, transmission and substation facilities added as data centres grow, and how those costs are allocated, the online media outlet Gigazine reported on July 18 local time.

Theodore J. Kury (시어도어 J. 퀴리), who leads energy research at the University of Florida, analysed that rising power demand from data centres is translating into higher electricity procurement costs. He cited a 2025 report by Monitoring Analytics, the independent market monitor for PJM, a grid operator in the eastern and midwestern United States.

The report estimated that without existing and planned data centre demand, power producers' market revenues would not have risen sharply in three capacity-market auctions from 2025-2026 to 2027-2028. It said the resulting cost increase from data centre demand would be borne by all electricity users in the PJM region.

The question is how much of that cost flows into residential rates. Electricity rate regulators allocate many cost items, from investment in power plants, transmission lines and substations to fuel and labour costs, across the residential, commercial and industrial sectors. If a specific user group consumes 20 percent of supplied power, it pays 20 percent of related costs. The number of users or usage at specific times can also be a basis.

Facilities dedicated to data centres have a relatively clear payer. For example, it is easy to conclude that data centres should pay for a power line linking a data centre to a nearby substation. But if a substation is expanded or new generation capacity is secured to supply data centres, other users will also use those facilities. In that case, the costs could be allocated across all users.

Kury in particular pointed to limits in the "coincident peak demand" standard. Under that method, costs are divided based on power usage at the moment when total grid usage is highest. Data centres can automatically adjust computing throughput to cut power use only at that moment, he said.

Long-term investment risk is also a variable. Power facilities are used for a long time, but not all planned data centres are built. Even after completion, if power use is lower than expected or rapid technological change makes facilities less used, costs a utility invested upfront based on data centre demand could be shifted to other users.

There is also an opposing view. Sean Regan (숀 리건), a senior fellow at the Manhattan Institute, argued that "it cannot be concluded that data centres raised household electricity bills." He cited a 2015-2024 analysis by Asa Warton and others at the Electric Power Research Institute. The study used the 1947 interstate highway plan, which is not directly linked to current electricity rates, to estimate each state's potential to host data centres.

The analysis estimated that if power capacity for data centres doubles, the average residential retail electricity rate falls by about 3.5 percent. Applying that data centre power capacity rose 160 percent in 2019-2024, it calculated that residential rates were about 6 percent lower than they would have been without the increase in data centres. The researchers said average cost per kilowatt hour could fall if fixed costs for power plants, transmission lines and distribution networks are spread over higher power usage on top of existing spare capacity.

Even so, the analysis is conditional. The researchers warned that "if constraints emerge in future power supply, the rate-lowering effect could disappear." Regan also argued that generation and transmission facilities must be expanded in line with rising data centres, and that new facility costs should be borne by data centres.

Actual policy adjustments have also begun. The Oregon Public Utility Commission approved Portland General Electric's rate overhaul plan in July 2026. Under it, rates for large power users such as data centres rise by an average 29.7 percent, while residential rates fall by an average 1.3 percent. The move lends weight to the argument that creating rate tiers dedicated to large users can prevent the cost of data centre expansion from being shifted to ordinary households.

Ultimately, the debate is increasingly focused on the idea that whether data centres raise electricity rates depends more on cost-allocation rules than on the rise in power demand itself.