December 1, 2025 The rapid expansion of artificial intelligence is triggering the fastest and most concentrated surge in U.S. electricity demand in modern history, creating a critical structural mismatch with a power grid that cannot be built or upgraded at the same pace. This looming crisis was the central focus of the inaugural Data Center POWER eXchange (DPX) summit, convened by POWER in Denver on October 28, which brought together utilities, hyperscalers, regulators, and developers to confront the practical challenges of delivering power at the scale AI requires. The summit framed core questions around building firm capacity rapidly, allocating infrastructure costs, maintaining grid reliability amid multi-gigawatt requests, and managing community acceptance as data center footprints expand. Discussions were guided by the full lifecycle of infrastructure projects, confronting universal constraints like long-lead equipment shortages, transmission limits, and the widening gap between the speed of digital companies and traditional utility planning cycles. Former Federal Energy Regulatory Commission (FERC) Chairman Mark Christie opened the event by warning of a "dual crisis." On the supply side, he noted the premature retirement of dispatchable generation, like coal, without equivalent replacement, eroding reliable supply in markets like PJM. Demand is being reshaped by a narrow band of large customers, with load forecasts—which directly influence capacity market prices—rising without corresponding new generation. "Something's got to give," Christie stated, highlighting affordability as a key boundary. He cited a 6% year-over-year increase in consumer power bills, far exceeding inflation, and cautioned that rising electricity costs represent a "political volcano that’s on the verge of exploding." Robert Blue, Chairman and CEO of Dominion Energy, provided a ground-level view from a utility he described as "ground zero of the data center revolution." He revealed that data centers now account for 27% of Dominion Energy Virginia's sales, concentrated in Northern Virginia's "Data Center Alley," which handles an estimated 70% of the world's internet traffic. The utility's connection rate has skyrocketed from 100-200 MW annually a decade ago to about 1 GW per year for the last three years. PJM's projected annual peak growth rate has surged from 0.5% in 2021 to 6.3% today, with Dominion expecting demand to double by 2039. To meet this, Blue advocated for an "all of the above" policy, with the company's latest plan including over 33 GW of new generation over 20 years and annual transmission capital spending rising to over $2.8 billion by 2027. He noted the company already has approximately 47 GW of load in various contracting stages as of September 2025. Caroline Golin, founder of Envision Energy Advisors and former Google energy lead, emphasized the urgency for grid modernization. She argued that treating AI demand as a bubble would lock the industry into outdated models instead of using the capital influx as a catalyst for a "facelift." Golin pointed out that many U.S. utilities lack visibility into their distribution systems and are unprepared to manage or redirect load efficiently. She stressed that the immediate "crunch" is in the next three years, a period during which traditional generation and transmission cannot be built fast enough. The solution, she argued, lies in leveraging grid-enhancing technologies, behind-the-meter solutions, and unlocking an estimated 50-70 GW of latent system capacity through new business models and correct market signals to ensure the U.S. remains competitive in the global AI race. Source: powermag
U.S. Grid Faces Unprecedented Strain as AI Demand Collides with Infrastructure Limits, Summit Warns
