San Francisco has emerged as a major center of the artificial-intelligence industry, with companies such as OpenAI and Anthropic helping anchor the sector in the city. The gap matters because data centers connect local land use with electricity demand, water use, backup power and other infrastructure considerations. Patrick Hannan, a spokesperson for the Planning Department, said, “We do not track the number of data centers in San Francisco.” That admission puts a sharper edge on the city’s current debate because officials lack a comprehensive city-maintained count of existing data centers as developers consider additional projects. The issue also reaches beyond bookkeeping, because a consistent inventory would give planners a clearer baseline for evaluating additional data-center proposals. For San Francisco, the mismatch between technological prominence and planning visibility has become increasingly difficult to ignore.

The city’s planning framework adds another layer to the problem because San Francisco does not have a dedicated land-use definition for data centers. That distinction matters because a dedicated land-use category can give planners a clearer basis for identifying and evaluating data-center projects. Data centers can occupy industrial properties and carry substantial electrical and other infrastructure requirements that may differ from conventional industrial uses. A zoning system that treats those uses through broader categories may give policymakers less visibility into the infrastructure characteristics of individual data-center projects. That gap can leave the city evaluating individual projects without a dedicated citywide data-center category. That challenge has become more visible as developers and investors examine San Francisco properties for potential data-center projects.

A 49.5-Megawatt Proposal Raises the Stakes

Nestory Park Capital lists a San Francisco flagship data-center project on its website, describing it as a 1.6-acre facility with 49.5 megawatts of capacity. The company’s portfolio lists a flagship San Francisco facility at 160,000 square feet on 1.6 acres with 49.5 megawatts of capacity. The The San Francisco Chronicle separately reported that the firm has explored 1215 Michigan Street in Dogpatch as a potential location for a 160,000-square-foot data center, although Nestory Park has not filed official plans and the property has not changed hands. That distinction matters because Nestory Park’s listed project remains different from an approved or formally filed city development application. Still, the proposed scale illustrates why data-center classification has become a practical planning question for San Francisco.

The 49.5-megawatt figure also changes the way policymakers can think about the physical footprint of computing. A data center does not need the acreage of a traditional hyperscale campus to create significant electricity and infrastructure requirements. High-density data centers can concentrate substantial computing capacity and electrical demand within relatively compact properties. Nestory Park says its strategy focuses on converting existing properties into data centers and establishing relationships with utilities, construction partners and technology tenants. That model could make urban redevelopment relevant to infrastructure investors while prompting cities to examine whether existing zoning categories adequately address data-center projects. San Francisco now has to consider whether the physical size of a project remains an adequate proxy for its infrastructure significance.

Walton Pushes for a Regulatory Reset

Supervisor Shamann Walton has responded by introducing legislation that would impose a 45-day moratorium on new data-center development in San Francisco. The proposal would pause new data-center approvals during the 45-day period while the city considers how to regulate the use. The legislation entered the Board of Supervisors process on Sept. 15, 2026, and Walton introduced it with Supervisors Connie Chan, Jackie Fielder and Chyanne Chen as sponsors. The measure remains subject to approval by the San Francisco Board of Supervisors. Its immediate significance therefore lies in giving city officials additional time to consider how data centers should be defined and regulated.

Walton has framed the data problem as a basic requirement for effective regulation of data centers. “The Planning Department should know and needs to know how many data centers are in San Francisco and how many applications for conditional use permits,” Walton told The Standard. “My legislation will surely make sure that this is the practice moving forward. No excuses for not having this information.” His proposal also focuses attention on the terminology regulators use because officials cannot consistently apply a development standard without defining the activity that triggers it. “Until we have a universal definition, it’s going to continue to be problematic,” Walton said. The debate therefore centers not only on whether San Francisco should host additional computing infrastructure but also on whether its planning system can identify and regulate that infrastructure consistently.

The Definition Matters as Much as the Moratorium

A formal definition can establish which projects fall within a city’s data-center rules and what additional review those rules require. It can also create a framework for examining electricity demand, water use, backup generation, noise and other operating characteristics that cities may want to address separately. Cities including San José have incorporated explicit data-center provisions into their planning frameworks, including public-hearing requirements and separate treatment of large energy-use projects. The contrast becomes significant because San José maintains a dedicated planning process for data centers and currently lists numerous applications under review or already approved. For investors, a defined category can clarify which planning requirements apply before a project reaches the permitting stage. For residents and infrastructure planners, a defined category can create a clearer framework for identifying data-center projects and evaluating their infrastructure requirements.

Erin Ryan, a law professor at USC, described land-use definitions as a foundation for deciding whether a community wants to encourage or restrain a particular form of development. “Creating a specific rule forces the deliberation of sound land use policy in advance of the crises that data centers can create for communities trying to cope with the resulting problem of scarce resources after the facilities have been built,” Ryan said. Her point shifts the discussion from a binary choice between development and restriction toward the mechanics of planning before infrastructure arrives. A city can use a dedicated category to establish conditions that support development, limit specific impacts or require additional information from applicants. Such a framework can also make community-benefit expectations more explicit when officials determine that new computing capacity aligns with broader economic goals.

Gilroy Shows What Happens When Rules Lag

San Francisco’s debate comes as other California communities confront similar questions after approving large computing projects under older zoning frameworks. In Gilroy, Amazon Web Services received approval for a project covering roughly 438,500 square feet across two data-center buildings and a security building. The city’s project materials identify a 49-megawatt PG&E connection for the first building, along with backup generation, a substation and other electrical infrastructure. Gilroy’s planning process later became a focus of public attention as residents and officials examined whether the city’s existing development rules provided sufficient oversight for data-center projects. The city now offers a detailed public project page that documents approvals, appeals, community outreach and subsequent council discussions.

The comparison became even more relevant as Gilroy moved toward a temporary moratorium and work on dedicated standards for future data-center development. The city said the pause would give officials time to develop objective zoning, siting, development and operating standards for future facilities. The ordinance prevents the city from accepting, processing or approving applications and entitlements for new data centers during the moratorium period. Gilroy’s action illustrates how local governments can move from applying existing development rules to considering dedicated standards for data-center projects. It also gives San Francisco a nearby example of how a city can respond when existing planning rules do not specifically address data-center development.

San Francisco’s AI Economy Meets Its Physical Limits

Large data-center projects have often targeted locations where developers can assemble suitable industrial land and secure the electricity and infrastructure needed for the facilities. San Francisco’s economic relationship with AI creates an unusual planning tension because many of the companies driving the sector operate in the city while the physical infrastructure supporting computing often sits elsewhere. San Francisco’s AI economy has been anchored by technology companies, research, investment and talent, while the city has had fewer large data-center projects than the major South Bay data-center market. A proposed 49.5-megawatt urban facility changes the conversation because it brings a meaningful slice of AI’s physical layer closer to the city where many of the sector’s companies operate. That convergence forces policymakers to consider how the city should evaluate the infrastructure requirements of data-center projects proposed within its boundaries.

Those decisions will involve questions about land use, electricity, water, environmental review, community impacts and the economic effects of proposed projects. A permanent regulatory framework could establish requirements before developers commit significant capital, rather than leaving the city to address individual projects without dedicated data-center standards. It could also give developers clearer standards for converting existing industrial properties into data-center facilities. For communities, the same rules could clarify what information applicants must provide and what forms of review apply to proposed facilities. Those mechanisms do not determine whether a project proceeds, but they determine how transparently and consistently the city evaluates it.