AI is driving energy prices. How NJ must protect ratepayers | Opinion

An opinion piece by Fred Fastiggi, principal and managing director of Shoreline Energy Advisors, warns that New Jersey ratepayers are facing rising electricity costs due to the massive energy demands of the artificial intelligence (AI) industry and its data centers. Fastiggi uses the "boiled frog" metaphor to illustrate how these costs are slowly increasing, likening proposed special electrical tariffs for AI owners to "whistling past the graveyard" if they don't sufficiently insulate non-AI users. The article explains that AI models consume vast amounts of electricity during training, with individual data centers requiring 500MW or more—half the size of a New Jersey nuclear generating station. National studies predict a significant increase in US data center energy demand, from 25GW in 2024 to over 80GW by 2030, potentially consuming 12% of all US electricity by 2030. This surge, combined with retirements of old power plants and delays in connecting new ones to the grid, is straining the Pennsylvania-Jersey-Maryland (PJM) Regional Transmission Operator, which serves New Jersey and 12 other states. Fastiggi criticizes the traditional utility rate-making process, which socializes infrastructure expansion costs, potentially leading to non-AI customers subsidizing infrastructure built specifically for AI. He also raises concerns about stranded costs if projected AI energy demands do not materialize, citing a past instance in Newark with PSE&G. To prevent electric prices from reaching a "boiling point," Fastiggi urges New Jersey legislators and regulators to adopt more aggressive, targeted measures. He suggests that prospective AI users should fund grid impact studies, be held responsible for exit fees if capacity commitments aren't met, and face penalties for underestimating demand. Additionally, he proposes strengthening specialized tariffs to make AI facility service exclusively interruptible during peak demand and relaxing policies to allow for isolated microgrids or distributed generation for data centers, thereby decoupling them from the utility grid and protecting non-AI users.