AI boom could leave Americans with a $23B power bill

Electricity consumers from 14 states in the mid-Atlantic and midwestern United States will incur an additional cost of $23 billion through at least the end of 2028 as a result of the surge in power demand coming from data centers, according to an independent market monitor report for PJM Interconnection (PJM), a wholesale electric power market operator.  The report comes as regulators are still determining who should bear the financial burden of the necessary infrastructure to support the rapid growth of the AI sector.

The dilemma is not whether new power infrastructure is necessary, but rather how the costs will be divided.

Theodore Kury, who researches state programs affecting large electric consumers and is the author of a Fortune analysis on the PJM market monitor report, explained that regulators are first interested in establishing the total costs of a particular utility before they can decide how these costs will be divided among residential, commercial, and industrial customers.

Certain expenditures are quite simple. If a data center reportedly requires the installation of a new power line linking it with the existing substation, the responsibility for paying for this is on the data center operator.

However, when the utility companies are faced with the necessity to expand substations, strengthen transmission lines or introduce new generation, the provision of such services brings benefits to the whole grid, which makes it difficult to decide how much of the bill should be shouldered by data centers and the other customers.

A loophole in how peak demand is measured

One contested area of utility pricing is how demand charges are calculated.

Many rate structures are based on “coincident peak demand,” which is defined as the amount of power used when the entire utility system is experiencing peak demand. Large data centers are able to significantly reduce their energy consumption during peak hours, helping them to save money on their utility bill in ways not available to the common consumer.

Kury called attention to crypto-mining activities in Texas that have already implemented this approach to lower electricity expenses. In contrast, home customers have hardly any power over their peak demand costs.

This same inconsistency is also evident in regulatory processes. Utility providers, industrial customers, and data center managers use the services of expert witnesses to argue in favor of their preferred methods of cost distribution. However, residential customers are not similarly privileged.

According to Kury, most states, except for Georgia, Louisiana, and Idaho, have established consumer advocacy groups, and many of them have been legally mandated to represent all consumers equally. This restricts the advocacy group’s ability to argue that data centers should bear a larger share of new infrastructure costs.

Regulators are still writing the rules

Electric power demands spurred on by artificial intelligence have outpaced the ability of regulators to respond with formal electricity rate rules, forcing state commissions to devise their own cost-sharing policies for every situation, rather than following a unified national policy.

At the same time, the electrical consumption of data centers keeps growing.

According to the Department of Energy’s report published by the Lawrence Berkeley National Laboratory, the consumption of electricity by data centers was estimated to be 4.4% of the total electricity consumption in the United States in 2023. It is expected that this share will grow to 6.7% – 12% by the year 2028. It has also been reported that annual electricity consumption increased from 58 terawatt-hours in 2014 to 176 terawatt-hours in 2023.

The projection of the Electric Power Research Institute is even higher, with data centers accounting for 9% to 17% of electricity demand in the United States by 2030.

This has raised fears that households might have to shoulder a major part of the cost. According to a commentary by Brookings published on July 9, a forecast from ICF predicted that residential electricity rates will rise by 15% to 40% by 2030 if the tariffs are not modified, and some utility companies may double their rates from now until 2050.

Brookings authors David M. Klaus and Mark MacCarthy said a growing consensus has emerged that customers outside the data center industry will face higher electricity bills if cost-allocation rules remain unchanged. Moreover, they emphasized that, a year before, many industry advocates argued that the expansion of investment in the data center sector would lead to a decrease in electricity costs.

Overbuilding is another concern. Brookings referenced estimates from Sightline Climate that as many as half of the large data center projects set to commence in 2026 will never be built. Different data from Wood Mackenzie also reported that announcements of new, large-scale data centers halved between the third and the fourth quarter of 2025, highlighting the possibility of stranded infrastructure projects that utility consumers will have to bear the costs of.

The smartest crypto minds already read our newsletter. Want in? Join them.