The artificial intelligence revolution is hitting a critical bottleneck: electricity. Across the United States, utilities are projecting eye-popping increases in energy demand—two to three times current levels within just a few years—all to power the massive data centers feeding the insatiable AI economy. But as these forecasts translate into concrete plans for new power plants and grid infrastructure, a fundamental question emerges: Are these projections based on real projects, or are they speculative bets that could leave ordinary ratepayers holding the bill?

Article illustration 1

The tension between the AI industry's voracious appetite for power and the electrical grid's capacity to deliver it has created what experts are calling an "energy crisis in slow motion." At the heart of the issue is a fundamental lack of transparency and verification in how utilities determine whether proposed data center projects will actually materialize.

"There's speculation in there," said Joe Bowring, who heads Monitoring Analytics, the independent market watchdog in the mid-Atlantic grid territory. "Nobody really knows. Nobody has been looking carefully enough at the forecast to know what's speculative, what's double-counting, what's real, what's not."

The Speculation Game

The problem stems from how data center developers engage with utilities and grid operators. Many companies seek preliminary grid connections to assess feasibility, but these requests often lack the final commitments—clients, financing, or otherwise—that would guarantee the project will be built.

Article illustration 2

"Utilities are in a 'fire drill' as they try to vet a deluge of data center projects all seeking electricity," said Igal Feibush, CEO of Pennsylvania Data Center Partners. "The vast majority, he said, will fall off because many project backers are new to the concept and don't know what it takes to get a data center built."

Compounding the issue is the practice of "double-counting," where developers submit grid connection requests across multiple utility territories without disclosing these parallel applications. PJM Interconnection, which operates the mid-Atlantic grid spanning 13 states, has identified this as a significant problem that artificially inflates demand forecasts.

"Often, developers, for competitive reasons, won't tell utilities if or where they've submitted other requests for electricity," PJM reported. "That means a single project could inflate the energy forecasts of multiple utilities."

Regulatory Response

The uncertainty has prompted unprecedented scrutiny from lawmakers and regulators. In September, Federal Energy Regulatory Commission member David Rosner called on grid operators to improve their verification processes.

"Better data, better decision-making, better and faster decisions mean we can get all these projects, all this infrastructure built," Rosner stated.

Article illustration 3

States are taking matters into their own hands. Texas, still reeling from the deadly 2021 winter storm that exposed vulnerabilities in its grid, was shocked when the Electric Reliability Council of Texas projected peak demand could nearly double by 2030. Lawmakers discovered that state utility regulators lacked tools to assess the viability of these projections.

"Lawmakers weren't sure if the power requests are real or just speculative or somewhere in between," said Texas state Sen. Phil King, who sponsored legislation now law that requires data center developers to disclose their electricity requests elsewhere in Texas and demonstrate substantial financial commitments.

In Pennsylvania, where PPL Electric Utilities projects data centers will more than triple its peak electricity demand by 2030, state Rep. Danilo Burgos introduced a bill to strengthen regulators' authority over utility forecasts. "Ratepayers in Burgos' Philadelphia district just absorbed an increase in their electricity bills—attributed by the utility, PECO, to the rising cost of wholesale electricity in the mid-Atlantic grid driven primarily by data center demand."

The Human Cost

Beyond the technical and regulatory challenges, the energy boom is translating directly into higher electricity bills for consumers. In the mid-Atlantic grid, ratepayers are already underwriting the cost to supply power to data centers, some built, some not.

"Once they make their buck, whatever company," Burgos said, "you don't see no empathy towards the ratepayers."

The Data Center Coalition, representing tech giants like Google and Meta, acknowledges the problem. "Improving the accuracy and transparency of forecasts is a fundamental first step of really meeting this moment of energy growth," said Aaron Tinjum, the coalition's vice president of energy.

Balancing Innovation with Responsibility

The AI industry's energy dilemma represents a critical inflection point. On one hand, the computational power driving AI breakthroughs requires unprecedented energy resources. On the other, the grid's capacity to deliver this power is finite, and the transition must be managed responsibly to avoid economic hardship for consumers.

As the industry continues its rapid expansion, the challenge will be creating a framework that allows for innovation while ensuring that the costs—and benefits—are distributed equitably. The alternative is a future where the AI revolution is powered by consumers who can no longer afford the electricity that makes it possible.