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

Governors in New York and Texas this month imposed new limits on data center development, reflecting a rare area of agreement across party lines on the need to manage the rapid growth of facilities powering artificial intelligence. New York Governor Kathy Hochul enacted a one-year statewide moratorium on hyperscale projects, citing concerns over electricity rate increases for residents and the need for better local benefit negotiations. Texas Governor Greg Abbott announced a pause until state energy regulators complete audits of existing operations, following years of unchecked expansion in the state.

These moves come as data centers have become the dominant driver of new electricity demand in the United States. Analysis from Global Energy Monitor shows the country is now developing nearly three times as much gas-fired power capacity as China, with projects in the pipeline reaching 378 gigawatts after a 50 percent increase since January. Roughly half of that capacity ties directly to data center needs, as tech companies seek to bypass grid connection delays through private behind-the-meter plants. The five largest hyperscalers plan more than 750 billion dollars in capital spending this year, with the bulk directed at AI infrastructure.

The shift toward natural gas carries clear emissions consequences. Adding the planned capacity could raise US power sector emissions by as much as 20 percent, according to estimates tied to the buildout. This marks a reversal from prior trends where China led in gas plant construction. Tech firms have increasingly pursued dedicated fossil fuel generation rather than relying solely on renewables or existing grid resources, a pattern accelerated by the scale of AI training and inference workloads.

At the federal level, the Trump administration has proposed changes to Clean Air Act implementation that would make air pollution disclosures for data centers voluntary in many cases. The rule would remove mandatory public notification requirements for minor sources of pollutants such as nitrogen oxides and carbon monoxide, leaving decisions to state discretion. Advocates note that without consistent disclosure, communities lose leverage to assess health impacts or pursue legal challenges. Several Democratic and Republican lawmakers have introduced or co-sponsored measures aimed at shielding ratepayers from data center-driven cost increases, including bills backed by Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez as well as Senator Josh Hawley.

The political response extends beyond energy policy into local and state contests. Candidates in Michigan and Pennsylvania have highlighted data center opposition as a way to connect with voters concerned about land use, water consumption, and utility bills. Pennsylvania Governor Josh Shapiro recently added stricter conditions to new projects after previously promoting them. This cross-partisan skepticism has created volatility for both parties, with some Republican strategists warning that unchecked support for the projects could affect Senate control.

The underlying tension stems from the sheer physical requirements of the AI buildout. Data centers compete directly for turbines, cooling equipment, and suitable sites while utilities revise demand forecasts upward by a factor of six compared with projections from three years ago. State-level pauses in New York and Texas signal an attempt to insert regulatory guardrails before the infrastructure commitments become irreversible. How these efforts interact with federal permitting changes and the private power strategy adopted by major tech companies will shape both the pace of AI deployment and the emissions trajectory of the US electricity sector in the years ahead.