Gartner forecasts global data center electricity consumption to reach 565 terawatt-hours in 2026 , up from 447 TWh last year. That's more power than the entire nation of France consumed in 2025. The surge is driven almost entirely by artificial intelligence—training models, running inference at scale, cooling the chips that make it all possible. Gartner's director analyst Linglan Wang calls power availability "the new battle ground for scaling and protecting margins in the global AI race."
The problem is that the grid was not built for this. The U.S. power grid was designed for 20th-century load patterns, not the surge demands of hyperscale data center campuses, and transmission upgrades take years of approvals while data centers are built in months , according to industry analysis. Now the collision between AI's breakneck expansion and the grid's glacial pace is forcing regulators, utilities, and tech giants into uncomfortable compromises—and someone will have to pay for it.
Who Foots the Bill When the Grid Runs Short?
PJM Interconnection's capacity auction cleared at $329.17 per megawatt-day for the 2026-2027 delivery year —an eleven-fold increase from the $28.92 price two years earlier. Data center load accounted for $6.5 billion, or 40%, of the $16.4 billion in costs from PJM's December capacity auction , the grid operator's independent market monitor reported in January. Most of that—$6.2 billion—is tied to data centers that haven't even been built yet.
PJM serves 67 million people across thirteen states and Washington, D.C. The capacity costs get passed to ratepayers. The average family in PJM territory faces an estimated $70-per-month increase by 2028 , according to industry estimates. Political backlash has been swift. Oregon became the first state to create a dedicated data center rate class, Virginia's SB 253 would shift distribution and capacity costs from households to data centers, and at least six states have introduced construction moratoriums , per regulatory tracking.
The question is not whether AI will consume more power—it will. AI is driving an unprecedented surge in global electricity demand set to rise more than 1 trillion kilowatt-hours per year through 2030, with data centers alone expected to account for nearly 20% of that growth , Morgan Stanley analysts noted. The question is who builds the infrastructure to deliver it, and who pays when it arrives faster than the system can adapt.
Can Regulators Move as Fast as the Industry?
The Federal Energy Regulatory Commission has set a June 2026 deadline to act in a high-stakes rulemaking that could redefine how massive new power users—including AI-driven data centers—connect to the U.S. interstate transmission system , according to an April order. The docket, RM26-4-000, was initiated after Energy Secretary Chris Wright invoked rarely-used authority to direct FERC to address the "timely and orderly interconnection" of large loads—generally defined as demand exceeding 20 megawatts.
Data center interconnection has become one of the most contested issues in U.S. energy policy, as AI-driven demand pushes power requirements to unprecedented levels and the federal regulatory framework governing how large loads connect to the grid is being reshaped , industry observers note. The rulemaking could extend federal oversight to areas historically managed by states. The National Association of Regulatory Utility Commissioners filed comments arguing that FERC asserting jurisdiction over load interconnection would interfere with state authority over retail rate cases , highlighting the tension.
Meanwhile, GE Vernova introduced GridOS for Transmission and released two new AI whitepapers addressing grid planning and autonomous grid-edge operations at its Orchestrate 2026 conference on June 9 , signaling that the software layer of grid modernization is accelerating. GE Vernova's CEO Philippe Piron said meeting rising electricity demand "will require more than adding generation, it demands a grid that can coordinate, adapt, and act faster than ever before," and that "software is now central to how utilities plan investments, operate networks, and respond to near real-time conditions."
The market for AI-powered grid optimization is responding. The escalating cost of grid congestion, estimated at $11.5 billion annually in the U.S., combined with massive capital avoidance potential, has ignited rapid growth in the market for Grid-Enhancing Technologies, and the global smart grid market, valued at $52 billion in 2025, is now projected to grow at a CAGR of nearly 17% , according to market research.
