Media signals: siting pushback and AI skepticism

The fight over artificial intelligence infrastructure has moved from chip supply to local permits, power hookups and investor patience. (energy.gov) Data centers are the warehouse-sized buildings that run cloud software and artificial intelligence models, and the Department of Energy said they used about 4.4% of United States electricity in 2023. Lawrence Berkeley National Laboratory projected that share could reach about 6.7% to 12% by 2028, with total use rising from 176 terawatt-hours in 2023 to 325 to 580 terawatt-hours in 2028. (energy.gov) That power demand is now colliding with grid rules. On January 16, 2026, PJM Interconnection, the grid operator serving 67 million people, said it would pursue 12 actions on large new loads, including faster paths for projects that can bring new generation and curtail usage during system stress. (pjm.com) Federal regulators had already stepped in. On December 18, 2025, the Federal Energy Regulatory Commission ordered PJM to write clearer rules for data centers and other large loads that want to co-locate with power plants, saying the existing tariff lacked clarity and consistency. (ferc.gov) The local politics are getting harder too. WBUR reported on April 6 that Lowell, Massachusetts, passed a one-year moratorium on new data center development, while residents in Everett were pushing for a ban over noise, air pollution and power use. (wbur.org) WBUR also reported that more than 140 activist groups in 24 states are campaigning to stop or slow data center construction, and cited Harvard researchers who estimated one Virginia facility could impose up to $99 million a year in health damages on nearby residents even while meeting air rules. (wbur.org) At the same time, Wall Street is separating enthusiasm for artificial intelligence from enthusiasm for every company funding the buildout. Goldman Sachs said on December 18, 2025, that consensus estimates for 2026 capital spending by artificial intelligence hyperscalers had risen to $527 billion from $465 billion at the start of the third-quarter earnings season. (goldmansachs.com) Goldman also said investors had rotated away from infrastructure names where earnings growth was under pressure and spending was funded with debt, while rewarding companies that could show a clearer link between capital spending and revenue. The firm said the correlation among large public hyperscaler stocks had fallen from 80% to 20% since June. (goldmansachs.com) CNBC reported on February 6, 2026, that the four biggest hyperscalers were on track for combined spending close to $700 billion this year, sharpening concerns about cash burn even as executives keep raising infrastructure budgets. (cnbc.com) The result is a two-front test for the artificial intelligence boom: companies still need enough demand to justify hundreds of billions of dollars in spending, and they still need enough electricity, permits and local consent to turn those plans into operating sites. (goldmansachs.com)