Who pays for a data center? The first answer is the investors and developers who are planning on pouring billions of dollars into building out power-hungry facilities to serve all sorts of internet services, especially artificial intelligence. And how much will it cost them? The numbers thrown around have a kind of casual gigantism that makes levelheaded evaluation difficult. $80 billion? $100 billion? $500 billion?
But while technology companies are paying for the chips and the systems that do the work of artificial intelligence, it may be normal people and businesses — homeowners, barbershops, schools — that end up paying for at least some of the electricity and system upgrades necessary to bring these facilities online.
That’s the argument made by Harvard Law School lecturer Ari Peskoe and Eliza Martin, a fellow at the school’s Environmental and Energy Law Program, of which Peskoe is a part. Their paper, published Thursday, is titled, “Extracting Profits from the Public: How Utility Ratepayers Are Paying for Big Tech’s Power.”
The core argument is this: The cost of maintaining and expanding the electricity distribution system is shared by all ratepayers — retail, business, and industrial — through a process governed by state public utility commissions. Utilities, meanwhile, have a legal mandate to serve everyone in their territory and a captive customer base of ratepayers, but they also compete among themselves for the business of energy-hungry customers, who can pick and choose where they set up shop. These customers often require new investment in grid infrastructure, which utilities pay for by asking state regulators to approve higher electricity rates — for everyone.
From there the conflict is clear: Utilities will want to attract big customers, and may sacrifice their retail customers in order to do so. And lately, with the AI boom booming, there are more of these big customers than at any other time in recent memory.
“Utilities’ narrow focus on expanding to serve a handful of big tech companies … breaks the mold of traditional utility rates that are premised on spreading the costs of beneficial system expansion to all ratepayers,” Peskoe and Martin write.
The traditional model of utility regulation is built on the premise that all ratepayers should pay for grid improvements, such as new transmission lines or substations, because all will benefit from them. This dynamic is disrupted, however, when it comes to customers demanding a gigawatt or more of power, the authors write. “The very same rate structures that have socialized the costs of reliable power delivery are now forcing the public to pay for infrastructure designed to supply a handful of exceedingly wealthy corporations,” the paper says.
“The assumption behind all this is that these are broadly beneficial projects that are going to benefit energy users generally,” Peskoe told me. “But I think that assumption is a bit out of date,” pointing to an example in Virginia of a $23 million grid infrastructure project retail customers paid for half of despite it being solely necessitated by the data center.
Peskoe and Martin set out an “alternative approach,” whereby data centers will power themselves — that is, outside of the utility system — and become a “formidable counterweight to utilities’ monopoly power.” In addition to being a more fair structure for the average customer, the authors also hope it will mark a “return to the pro-market advocacy that characterized the Big Tech’s power-sector lobbying efforts prior to the ChatGPT-inspired AI boom.”
While this approach would be a major challenge to almost a century of utility regulation, Peskoe and Martin also set out some more modest options, such as having state regulators “condition service to new data centers on a commitment to flexible operations.” That proposal cites research from Duke University — and featured previously in Heatmap — showing that a commitment by data centers to power down for a small portion of every year could allow utilities to avoid having to build billions of dollars worth of new infrastructure to serve the peak demand of the system.
The barrier to this approach is that utilities “have historically been hostile to regulatory attempts to require measures that would defer or avoid the need for costly infrastructure upgrades that drive utilities’ profits,” Peskoe and Martin argue. While the enormous investment in data centers is novel, Peskoe told me that the core issue of utilities using their captive ratepayers as a checkbook in order to pursue big fish customers is right at the heart of the utility playbook.
“A lot of this is baked into the utility business model,” Peskoe said. “The incentives to deploy capital and the ability to shift costs among consumer groups are unique to utilities.”
But just as utilities have a unique business model whereby investor-owned businesses are granted monopolies, they also have a unique regulatory structure. (Apple doesn’t have to go to a board appointed by a governor to get approval to hike the price of the iPhone.) This setup gives regulators unique powers — and unique responsibilities — to patrol and restrict utilities taking advantage of ratepayers, Peskoe said.
“Regulators can try to police this stuff. It's hard. But that's one of the goals of utility regulation, is to try to police these poorly designed incentives,” Peskoe said.
“None of the consequences are baked in, but some of the basic mechanisms and incentives are just inherent and not unique to data centers.” What is unique to data centers in this moment, Peskoe added, “is just the scale of this growth, and therefore the potential scale of these cost shifts.”
