US opposition and regulatory scrutiny intensify for large AI data-center projects
Редакция ЦОД:Главное
Public and regulatory resistance to large AI data centers is increasing in the United States, with projects reportedly halted, delayed or cancelled and several states introducing or outlining tighter requirements in 2026. The reported constraints centre on power-system capacity, utility costs, water use, noise and community consent.
- S1 reports at least 75 US data-center projects were halted, delayed or cancelled in the first three months of 2026, with about US$130 billion in aggregate value.
- New York, Texas and Pennsylvania reportedly introduced or announced tighter conditions affecting new data-center development in 2026.
- ERCOT queue demand of about 474 GW represents proposed interconnection demand, not commissioned data-center capacity.
- Grid access, water availability and community consent could become more consequential variables in US AI data-center site selection.
Public opposition and regulatory scrutiny are emerging as material constraints on the build-out of large AI data centers in the United States, according to S1. The source reports that, during the first three months of 2026, at least 75 US data-center projects with an aggregate value of about US$130 billion were halted, delayed or cancelled. Those categories should not be read as indicating that all of the projects were permanently cancelled.
S1 attributes resistance to concerns over electricity and water tariffs, grid loading, noise, land use and the relatively limited number of permanent jobs once construction is complete. It reports that 71% of Americans oppose AI data-center construction in their local area, including 48% who are strongly opposed.
The reported policy response has varied by state. S1 says New York suspended environmental permitting for one year for large data centers above 50 MW on 14 July 2026. It says Texas froze the progress of new data centers in the grid-interconnection process on 3 August, while requiring reviews of control, funding sources, and projected water and electricity use. On 18 August, Pennsylvania reportedly imposed requirements for data-center developers covering self-supplied power, payment for their own electricity and local-community consent.
Power availability is central to the issue. S1 estimates that proposed data-center demand in the ERCOT interconnection queue had reached about 474 GW in 2026—more than five times Texas's historical peak demand. This is proposed queue demand, not operating capacity, and S1 notes that some applications may be intended to reserve queue positions rather than reflect final investment decisions.
Local examples illustrate the siting challenge. S1 cites a US$6 billion data-center project in Festus, Missouri, that did not proceed following voter opposition. It also cites a planned 50 MW project in Monterey Park, California, located about 150 metres from a residential area.
For developers, the reported trend could make early power procurement, grid-infrastructure funding and mitigation of local impacts increasingly important to project execution. If the restrictions described by S1 persist, the US AI data-center market could become more differentiated by available grid capacity, water resources and community acceptance, potentially increasing development cost and duration for new campuses.
For AI data-center developers and hyperscalers, site selection is increasingly shaped by more than land and capital availability. If these constraints persist, grid access, water availability and local acceptance could extend campus development timelines, raise interconnection and infrastructure costs, and favour locations with more workable power and regulatory conditions.
- — / 6 октября 2026 г. · КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.