Communities scarred by heavy industry lead US pushback against AI data centers

More than 100 people rallied outside Philadelphia's City Hall on Monday to launch a campaign aimed at keeping AI data centers out of the city. According to TechCrunch, the effort is being driven in part by residents of Grays Ferry, a Southwest Philadelphia neighborhood still living with the legacy of the oil refinery that once dominated it — and it reflects a pattern now playing out across the United States, where resistance to data center construction is sharpest in places already marked by heavy industry.

A neighborhood shaped by a refinery

Shawmar Pitts, a lifelong Grays Ferry resident and now managing co-director of the environmental justice group Philly Thrive, grew up beside what was then the East Coast's largest oil refinery. The Philadelphia Energy Solutions facility processed up to 335,000 barrels of crude per day before an explosion shut it down in 2019. Pitts told TechCrunch he only connected the chronic illnesses in his family and on his block to the refinery after encountering Philly Thrive's outreach in 2018, and he has campaigned with the group ever since.

City officials have identified two potential data center sites, one in Northeast Philadelphia and one in Grays Ferry, though no formal proposal exists yet. At Monday's rally, Philly Thrive board president Sonya Sanders — whose husband, another Grays Ferry resident, died of a rare bone cancer in 2020 — said residents want neither the pollution nor the noise, and dismissed promises of new jobs. She attributes her husband's death to refinery emissions.

The energy arithmetic behind the protests

A data center would not match the refinery's environmental footprint, but the scale of the AI buildout is generating its own numbers. BloombergNEF now projects that U.S. data centers will consume more natural gas than Germany and Japan combined by 2035 — nearly double the forecast the firm issued nine months earlier. TechCrunch cites estimates that each cubic foot of natural gas burned releases the equivalent of 60 grams of carbon dioxide, and that added data center demand is expected to produce roughly 1 million extra metric tons of greenhouse gases per day, about 12 percent of current U.S. emissions. A 2021 Harvard study found that fossil fuel air pollution contributes to one in five deaths worldwide.

Annie Fox, an attorney with the Clean Air Council, warned at the rally that a typical data center planned in Pennsylvania includes hundreds of diesel backup engines, and that officials have already indicated operators should run them during certain heat waves rather than draw from an overstressed electrical grid — a situation she expects to become more frequent as fossil-fuel-driven computing worsens the heat waves themselves.

Local governments are starting to pause

The pushback appears to be landing with officials. New York Governor Kathy Hochul signed an executive order that temporarily stops the state from approving permits for large projects, arguing that progress should not arrive with higher utility bills, a depleted water supply, or noise, and that data centers belong only in communities that want them. Moratoriums have also been approved in Denver, Indianapolis, Asheville, Charlotte, and Reno, often to give officials time to study the potential impacts before approving anything.

Resource competition and privacy feed the concern as well. Philadelphia resident Donna Greenberg told TechCrunch she worries about pollution, about operators gaining access to information they have no business knowing, and about water and electricity being diverted away from the neighborhoods that need them.

Why it matters

The AI industry's trajectory depends on a physical buildout, and that buildout is increasingly gated by local consent. Moratoriums and permit freezes can delay or relocate billions of dollars in planned capacity, and energy forecasts are climbing fast enough that siting friction has become a first-order constraint rather than a footnote. For hyperscalers, utilities, and the cities negotiating with them, community trust, grid stress, and emissions accounting now sit alongside chips and capital as the factors that determine how quickly AI can actually scale.