San Francisco lawmakers voted unanimously on Tuesday to extend the city's freeze on new data centers from 45 days to nearly a year, with the possibility of a further two-year extension, delivering a significant check to the city's booming artificial intelligence industry just days before Anthropic's stock market debut. The measure, approved by the Board of Supervisors, reflects growing concerns over the energy-intensive facilities that underpin the AI sector and whether the city should permit new data centers at all.
The extension follows a planning department report issued last week that identified major gaps in the city's understanding of existing facilities, including an incomplete inventory and unresolved compliance issues. The report raised questions about electricity and water consumption, pollution, and the concentration of data centers in neighborhoods already burdened by environmental hazards, but stopped short of recommending whether the city should allow additional facilities. No pending applications are currently on file, according to the report.
"We really don't know the impact of data centers to our communities," Supervisor Shamann Walton said before Tuesday's vote. Walton, who authored the measure, argued that the absence of pending applications is not a reason to let the moratorium expire. The Land Use and Transportation Committee had recommended extending the moratorium for 10 months and 15 days. Mayor Daniel Lurie's office declined to comment on the vote.
The debate carries substantial economic stakes for a city positioning itself as the capital of the AI boom. San Francisco is home to a growing cluster of AI companies, and proponents argue that some computing infrastructure should remain available locally to support firms developing and testing those technologies. Mayor Lurie has actively courted tech leaders as part of his economic strategy, with OpenAI CEO Sam Altman joining his transition team and business executives later enlisted to advise on downtown recovery efforts. AI companies have also driven demand for office space, helping downtown leasing rebound after the pandemic.
However, the planning department's report warned that large data centers could compete with housing, hospitals, and other development for limited electricity capacity. It found that a single 50-megawatt data center could consume as much electricity as 25,000 new homes, and that the city is already nearing its transmission capacity. Public speakers at Monday's committee hearing echoed concerns about pollution and the concentration of existing facilities in vulnerable communities, with Bayview-Hunters Point resident and community organizer Priyana describing the debate as "a political question about who gets to make decisions in our city."
The Building and Construction Trades Council, in a letter from leader Rudy Gonzalez, opposed a permanent ban and urged the city to use any extension to establish clear rules and a path to approval for projects that can cover infrastructure costs and protect workers, ratepayers, and surrounding communities. Gonzalez called for distinguishing large commercial facilities from smaller sites used for research, testing, and development, while cautioning that data center projects should not come at the cost of housing construction. Walton acknowledged the letter and pledged to work with unions on data center restrictions going forward, stating there is "every intent of being inclusive of community, every intent of including labor, every intent of including everyone."
With Tuesday's vote, San Francisco joins Oakland and Seattle in imposing temporary moratoriums while officials develop regulations, rather than following cities such as Alhambra, Monterey Park, and Calexico, which have adopted permanent data center bans. Developers have nevertheless explored potential projects at the shuttered downtown San Francisco Centre mall and in the Dogpatch neighborhood near the waterfront as recently as this summer. The extension keeps new approvals on hold while the city considers what rules should govern future development, a decision with significant implications for the region's AI infrastructure landscape.
