AI Demand Spurs US Gas Plants at Twice China's Pace

States across the political spectrum are moving to curb the rapid expansion of data centers, even as the facilities underpin a historic wave of private investment in artificial intelligence infrastructure. New York Governor Kathy Hochul enacted a year-long moratorium on new hyperscale projects in July, citing the need to prevent higher electricity rates for consumers. Texas Governor Greg Abbott followed in early August with a temporary halt pending audits of existing sites and their impact on the grid. Similar measures have surfaced at the federal level, including proposals from Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez, while bipartisan concerns over costs have drawn support from figures such as Senator Josh Hawley.

Analysis from Global Energy Monitor shows the United States now leads the world in gas-fired power projects under development, with capacity in the pipeline reaching 378 gigawatts as of mid-2026, up 50 percent since January. Roughly half ties directly to data center needs, reflecting a near doubling of behind-the-meter gas plants dedicated to these facilities over the past year. This shift allows companies to bypass congested transmission lines and avoid passing expenses to ratepayers, a response to longstanding regulatory delays in grid connections.

The five largest hyperscalers plan more than 750 billion dollars in capital spending this year, with the bulk directed toward AI-related builds. Utilities have revised demand forecasts sharply upward, projecting six times the peak growth anticipated just three years earlier. Private power arrangements have emerged as a practical adaptation, with voluntary pledges involving major technology firms and several state governors aimed at matching supply to localized demand without broader subsidies.

Critics highlight potential air quality effects from nitrogen oxides and other emissions, prompting the Trump administration's Environmental Protection Agency to propose changes that would make public disclosure of minor-source pollution voluntary in many cases. Proponents of the rule change argue it reduces administrative burdens that have historically slowed permitting without clear environmental gains. Communities opposing new sites often cite noise, water use, and visual impacts, leading local officials in places such as Pennsylvania and Michigan to tighten requirements or reverse earlier support.

These restrictions carry economic trade-offs. Data centers represent concentrated capital formation that drives productivity gains across sectors, yet barriers to construction risk shifting activity abroad or constraining the computational resources essential for continued technological advance. Historical patterns suggest that attempts to centrally manage infrastructure expansion frequently produce shortages and elevated costs borne by end users. Market-driven solutions, including on-site generation, have already adjusted to regulatory frictions more swiftly than centralized planning alternatives. The underlying pressure stems from genuine electricity requirements rather than speculation, and further layers of oversight may simply redistribute the burden without addressing root supply constraints.