Regulators to rule on who pays

• The Federal Energy Regulatory Commission is expected to rule this month on whether AI data centers or household ratepayers fund new grid infrastructure.
• PJM Interconnection’s latest capacity auction saw 40% of $16.4 billion in costs attributed to data center demand.
• PJM projects summer peak demand will rise 58% by 2046, from 160 to 253 gigawatts, driven primarily by AI servers.
• Two competing regulatory approaches: data centers connect through regular grid (costs spread to ratepayers) or connect directly to power plants (companies bear costs).
• Customers in seven PJM states already paying $4.4 billion for data center transmission lines through electricity bills due to regulatory gaps.

The collision course: Historic demand meets failing infrastructure

The nation’s largest electric grid operator is buckling under the weight of artificial intelligence, and federal regulators are poised to decide this month whether ordinary Americans or trillion-dollar technology companies will pay to keep the lights on.

The Federal Energy Regulatory Commission is expected to issue a ruling in Docket No. RM26-4-000 by the end of June 2026 that will determine how the U.S. power system accommodates AI-driven data centers—facilities that now consume twice the electricity of older server farms and are being built at an unprecedented pace across the United States.

At stake is not merely who writes the checks for new transmission lines and power plants, but whether the AI expansion proceeds as a publicly subsidized enterprise or as one in which the corporations generating the demand bear the costs of their own infrastructure.

The crisis is most acute in PJM Interconnection, the grid operator serving 65 million people across 13 states and the District of Columbia. PJM’s independent market monitor reported that existing and projected data centers accounted for 40% of the $16.4 billion in costs from the grid operator’s most recent capacity auction—a mechanism designed to ensure enough power generation exists to meet future demand.

$14.7B auction signals breaking point

The numbers tell a story of accelerating strain on a system built for a different era. PJM’s capacity auction costs surged to $14.7 billion, driven almost entirely by data center electricity demand. The grid operator’s 2026 long-term forecast projects summer peak demand will climb from 160 gigawatts in 2025 to 253 gigawatts by 2046—a 58% increase.

PJM’s market monitor offered a blunt assessment: “The current conditions in the capacity market are almost entirely the result of large load additions from data centers, both actual historical and forecast.”

The problem extends beyond PJM. In Texas, the Electric Reliability Council of Texas saw 300% growth in its large load queue last year, with more than 70% of requests coming from data centers. ERCOT planners are now assessing more than 233 gigawatts of large load interconnection requests—a volume that overwhelmed existing review processes.

Two paths forward: Public subsidy or corporate responsibility

FERC is weighing competing approaches that will shape the nation’s energy future for decades.

One option would connect hyperscale data centers through the regular electric grid, leaving utilities, state regulators and household ratepayers to manage the cost of additional transmission lines and power generation. Under this model, the $4.4 billion in data center transmission lines already scheduled for construction in 2024 across seven PJM states—costs passed to customers through electricity bills—becomes the template for future expansion.

The alternative would require data centers to connect directly to power plants or develop their own on-site generation, keeping those massive loads outside the broader grid. This approach would assign costs directly to the technology companies driving demand, forcing Amazon, Google, Microsoft and other AI developers to internalize infrastructure expenses rather than socializing them.

FERC has already signaled direction. In December 2025, the commission ordered PJM to create transparent rules for large loads co-located with power generation. In January 2026, FERC approved the Southwest Power Pool’s High Impact Large Load initiative, establishing a process for accelerating major customer interconnections alongside the generation needed to serve them.

Historical context: A system designed for a different century

America’s electric grid was built over a century through a regulatory compact: utilities received monopoly service territories and guaranteed returns on investment in exchange for providing reliable power to all customers at reasonable rates. This system assumed demand would grow at predictable rates of 1% to 2% annually.

AI has shattered that assumption. Data centers now seek amounts of electricity once associated with entire cities. The facilities being built in Virginia’s data center alley—the world’s largest concentration of server farms — consume power equivalent to medium-sized metropolitan areas.

The current regulatory framework never contemplated a scenario in which a handful of technology companies would need to double the nation’s generating capacity within two decades. Nor did it anticipate that those same companies would use regulatory gaps to shift transmission costs onto household ratepayers who have no say in the expansion.

The ratepayer burden grows while tech profits soar

Federal regulators have received more than 3,500 pages of public comments as they weigh how to manage the demand surge. The central question remains unresolved: should the costs of connecting AI data centers be spread across all electricity customers, or assigned directly to the companies creating the demand, some of which are the most affluent corporations to ever exist?

Chairman Laura V. Swett stated FERC is “addressing this challenge head-on.” But for customers in PJM’s territory, the outcome carries immediate financial consequences. If data center costs are absorbed through the broader grid, existing customers face higher charges as utilities expand infrastructure. If the companies driving demand must bear direct responsibility, hyperscale developers will face higher upfront costs and additional requirements before connecting.

The disparity is stark. The technology companies building these data centers reported combined profits exceeding $100 billion in 2025. Meanwhile, households in PJM states are already paying for transmission lines that serve data centers exclusively, with no mechanism to recoup those costs from the corporations using them, all while American families are struggling with hyperinflation.

A defining moment for American energy policy

The FERC ruling expected this month will establish whether the AI buildout proceeds as a public obligation or a private investment. The decision carries implications beyond electricity prices. It will determine whether the nation’s grid expands in an orderly, equitable manner—or whether America creates a two-tiered power system in which technology companies build their own infrastructure while ratepayers are left with rising bills and aging equipment.

PJM’s 253-gigawatt demand projection for 2046 is not an abstraction. It represents real power plants, real transmission lines and real costs that must be paid by someone. The question before regulators is whether those costs fall on the trillion-dollar corporations generating the demand or on the families and small businesses already struggling with rising utility bills.

The decision will also shape how other grid operators treat large-load customers. As of mid-June 2026, FERC had not issued its final order. But the ruling, when it comes, is likely to become the national model for how the power sector responds to the AI-driven electricity buildout—for better or worse.

Sources for this article include:

YourNews.com

UtilityDive.com

Axios.com