Electricity bills are rising sharply for millions of Americans, and experts say the culprit is an unexpected one. The massive buildings filled with computers that power artificial intelligence.The eastern United States grid managed by PJM Interconnection LLC saw power supply expenses jump by $7.3 billion in its most recent auction, with facilities housing AI servers responsible for nearly half of those costs. About 67 million people living in 13 states depend on this grid for their electricity.According to a Wednesday report, these computer facilities made up 45% of total power supply expenses in the summer auction. Looking at the previous auction as well, the buildings contributed roughly two-thirds of record-high prices. Combined, both auctions show these facilities added $16.1 billion in expenses.The situation has become so contentious that Pennsylvania’s governor, Josh Shapiro, suggested his state might withdraw from the regional grid system.Everyday Americans feel the squeezeKevin Stanley lives in Baltimore, Maryland, more than an hour away from northern Virginia’s concentration of server farms. The 57-year-old relies on disability income and has watched his energy costs surge about 80% compared to three years ago. “They’re going up and up,” Stanley explained to Bloomberg. “You wonder, ‘What is your breaking point?'”His experience reflects a growing problem as these energy-intensive facilities push electricity prices to new heights throughout much of the nation, forcing regular households to shoulder costs tied to digital infrastructure.The enormous buildings require vast amounts of electricity, adding to Americans’ financial strain from already climbing food and housing prices. This is creating both economic and political tension as utilities and community leaders debate how to split up the expenses. However, these same facilities are considered crucial for maintaining American competitiveness in global AI development.Technology corporations, now among the world’s most influential entities, have bet heavily on AI. The facilities, some large enough to cover much of Manhattan, contain endless rows of servers providing the computing capability and storage required to develop and operate AI systems.Energy demands will only grow larger as tech giants double down on AI infrastructureNvidia Corp. recently announced plans to invest up to $100 billion in OpenAI for new facilities. Microsoft Corp. signed a multiyear agreement valued near $20 billion with Amsterdam-based Nebius Group NV for cloud computing using a New Jersey location.OpenAI and Oracle Corp. created a partnership to construct 4.5 gigawatts of capacity, enough to supply power to millions of American residences.“The reliability crisis is here now; it’s not off in the distance somewhere,” stated Mark Christie, a past chairman of the Federal Energy Regulatory Commission who previously worked as a Virginia regulator. He noted that projected power demands, largely driven by requests to connect these facilities to the grid, are a major factor increasing expenses.The cost problem reaches beyond America. Japanese power auction prices reached record levels as officials anticipate an AI expansion, while Malaysia is raising rates for these facilities as new construction strains supply. A report from Aurora Energy Research in the UK projected that increased consumption from these buildings could drive power prices up 9% by 2040.Worldwide, these facilities are projected to use over 4% of electricity by 2035, per BloombergNEF. For perspective: if they were a nation, they would rank fourth globally in electricity consumption, trailing only China, the US and India.In America, power usage from these facilities is expected to double by 2035, reaching nearly 9% of total demand. Some experts predict this will be the largest jump in US energy consumption since air conditioning became widespread in the 1960s. This comes as the grid already faces challenges upgrading old infrastructure and adjusting to climate change. Want your project in front of crypto’s top minds? Feature it in our next industry report, where data meets impact.