When the first-quarter numbers on America's data center backlash arrived this spring, they showed that at least 20 proposed data centers were canceled after local opposition in the first three months of 2026, a record so far. The loss in total value amounted to more than $41.7 billion in planned investment, according to Heatmap Pro. A separate tracker, Data Center Watch, counted 75 projects worth $130 billion blocked or delayed over the same period, equal to all of 2025, while the number of organized opposition groups more than doubled to 833 across 49 states. Perhaps more telling than any single figure is the trajectory: two cancellations in 2023, six in 2024, 25 last year, with at least $85 billion in lost project value over the last three years. The trend toward increasing local resistance to data center buildout is clear.
For a while, mainstream framing tried to treat all of this as a familiar story of NIMBYism versus economic progress. However, looking more closely at the figures, that framing has not survived contact with reality. Grist has documented that the local backlash against data centers is a more bipartisan issue than nearly any other in American politics. Brookings now describes the conflict as a fight over whether the infrastructure of AI will be held accountable to democratic institutions, while the Institute for Local Self-Reliance has published a policy framework arguing that local opposition is less a land-use problem than a popular response to the increasing concentration of monopoly power. Even Microsoft and Oracle executives have publicly conceded that Big Tech has a community problem. So, what is the solution?
Communities have learned how to derail data center development before a project plan is even formally filed for planning consent. The question that will define the next decade of American AI is what should communities consent to, and on what terms. I would argue that the solution to Big Tech's community problem is to make way for local participation and, ultimately, local ownership. There is a history lesson here worth recounting.
Local opposition to new industries in America is not without historical precedent. Even so, it is worth looking closely at who is objecting. In Heatmap's June polling of more than 4,000 registered voters, 78 percent of those who voted for Kamala Harris and 63 percent of those who voted for Donald Trump said they would oppose a data center near their homes. No region of the country came in under 69 percent opposition, and rural voters proved slightly more opposed than anyone else. When conservatives and liberals converge on the same position, it seems unlikely that what they are rejecting is technological progress. What they are rejecting, I would suggest, is a business model in which local energy, water, and land are consumed while the economic value mainly flows to an increasingly unpopular cadre of distant shareholders.
The mining industry coined the phrase “social license to operate” to describe what extractive companies lose when communities withdraw their consent. That the vocabulary fits the AI buildout so naturally says something about the current moment. Data Center Watch has found that opposition sometimes mobilizes before a project is even filed, with the mere rumor of a data center enough to trigger organized resistance. It is difficult to imagine what a more fully revoked license would look like – there is a serious and growing threat to the AI sector that no amount of PR or social media campaigning can override.
Underlying these stark polling numbers there is a clear lack of trust between local populations and Big Tech companies. Only 7 percent of voters say they trust tech CEOs to make decisions that affect their lives. Follow the money and community concerns about extraction are supported by data. Between 2020 and 2024, residential electricity prices rose 25 percent while commercial users saw far smaller increases and industrial users now pay less than they did two years ago, a divergence that a Harvard Law School report captures in a title that does much of the analysis for us: “Extracting Profits from the Public.” Utilities requested a record $31 billion in rate increases in 2025, more than double the prior year, followed by another $9.4 billion in the first quarter of 2026 alone, while the average American household paid roughly $110 more for electricity last year than the year before.
The stakes extend beyond affordability, because in a climate-insecure world, energy security and affordability have effectively collapsed into a single concept. Low-income households already spend up to 20 percent of their income on energy, against roughly 3 percent for wealthier ones. A family that cannot afford to run the air conditioning through a heat dome is energy-insecure in the most literal sense, whatever the national security briefings may say about resilient supply.
The industry's traditional sweetener of job creation, meanwhile, is structurally weak. Meta's $800 million campus in Wood County, Ohio, will support approximately 100 jobs once the construction crews leave, by the company's own announcement. Measured in jobs per square foot, data centers trail almost every alternative use of the land, from retail stores to distribution centers to factories. A warehouse facility can buy local acceptance with payroll; a data center cannot, because there is very little payroll to offer. An industry that cannot pay communities in the traditional currency of jobs will have to pay in something else. The obvious answer is equity.
Skeptics can note, fairly, that most of these canceled projects were speculative, rezonings that existed only on paper. Even so, the pattern proves the point that matters: local opposition groups hold effective veto power at the siting stage, before capital is committed, which means the ownership question is one of business model and must be answered before a project breaks ground. At the same time, given the computational resource gap between towns and hyperscalers, no local organization will ever have the capacity to train a commercial-scale frontier model; gigawatt-scale training campuses are the exclusive domain of hyperscalers and always will be. Yet the industry's own economics are shifting toward inference, the everyday running of AI for hospitals, farms, schools, and small manufacturers. Local-scale compute of that kind constitutes a workload that favors proximity to the towns and cities it serves and could be run cooperatively. This is where the opportunity for shared ownership, shared revenue, and co-investment lies.
In the 1930s, investor-owned utilities declined to electrify rural America because the returns on investment were unviable. The solution, as it turned out, was not to forgo providing electricity to rural populations but to adopt a more locally beneficial ownership model. Farmers formed member-owned cooperatives, backed by federal lending, creating an opportunity to build the grid out themselves. That legacy survives today, with nearly 900 electric co-ops providing power to roughly 42 million Americans across some 56 percent of the American landscape. Municipal broadband followed the same pattern when incumbents refused to expand fiber networks to sparsely populated rural towns.
Two decades of research on renewable energy siting finds that community-owned projects enjoy higher levels of trust than outside commercial developers, that acceptance rises when residents participate in decisions and hold a stake, and that compensation perceived as a buy-off is sometimes dismissed and can even deepen hostility. Denmark went so far as to write wind farm co-ownership into national law in 2009 for precisely this reason. A stake, it seems, works where a sweetener fails.
At present, nearly two-thirds of new data centers are planned for rural areas, where member-owned cooperatives serve much of the grid. This creates an opportunity for hyperscalers to trial more locally beneficial ownership models. Buckeye Power in Ohio offers a glimpse of what enhanced local ownership can buy. Because members own the generation through Buckeye Power, the state's co-ops were able to build a process assigning data center costs directly to data centers; as their chief executive put it, “our members will not be asked to shoulder the costs associated with serving them.”
The value at stake can be quantified. Byron, New York, a town of fewer than three thousand people, negotiated nearly $25 million over 20 years in connection with a 280 megawatt solar facility, roughly $89,000 per megawatt over the life of the deal, and the template is already migrating to compute: Lancaster, Pennsylvania has executed what is believed to be the first data center-specific community benefits agreement, worth $20.25 million in contributions over a twenty-year term. That agreement is an imperfect first draft, negotiated, residents complained, with too little community input, which is itself an argument for standing ownership over one-off agreements. Proposals now circulating would go further still, including compute royalties credited to every parcel of land near a facility. The business models are not difficult to imagine: a co-op edge data center anchoring a community microgrid, its waste heat warming greenhouses; member-priced AI services for precision agriculture and rural telehealth, sold the way co-ops already sell power and, increasingly, broadband; member dividends and AI workforce training run the way co-ops once ran electrification demonstrations for farm families.
No cooperative data center exists in America today. However, if we look abroad, we may find some inspiring examples. Germany ran cooperative data processing centers for its cooperative banks beginning in 1963, while Switzerland's Infomaniak, structured as a workers' cooperative, heats 6,000 households with reused server heat. These are proofs of concept rather than proof at scale, and the missing rung is the same one rural electrification faced, which is financing. Electrifying the countryside did not pencil for private capital either, which is why it took a federal lender. The Rural Utilities Service is authorized to lend $7 billion for rural electric infrastructure in 2026, with expected demand already exceeding that figure. Extending the mandate to community compute, an RUS for the intelligence age, is the single policy that would move this vision from aspiration to pipeline.
E.F. Schumacher spent two decades as chief economic adviser to the British Coal Board, then one of the largest organizations in the world, before concluding that “production from local resources for local needs is the most rational way of economic life.” Half a century later, as the infrastructure of intelligence arrives in the American countryside, his conclusion has returned with higher stakes.
Dr. Phillip Bruner is a professor at the University of Washington's Foster School of Business and founder of the nation's first Climate Risk Lab housed at a business school. He advises institutions on climate-related financial risk and energy market resilience.
