Sarah Kreps argues that elected officials should be cautious of immediately rejecting proposals for local data centers. Instead, officials should capture potential benefits and protect residents by negotiating for better terms.
In the current political climate, consensus on any issue is a rarity. And yet increasingly, Republicans and Democrats are speaking with one voice on the issues that come with the construction of data centers that power artificial intelligence. Both parties have capitalized on across-the-aisle voter antipathy, which is arguably a proxy for skepticism toward AI. Opposition to physical data centers directly affecting their communities offers voters more agency than opposing the rise of AI itself.
That the discourse has turned negative is not surprising, even if election year politics sharpens the tone. Big Tech is always an easy target, and data centers have contributed to increases in local energy and utility costs. With inflation continuing to validate concerns about price increases of any kind, the political incentives further align in opposition. However, communities sacrifice tax revenue and municipal renewal by their opposition, and if they feel current tradeoffs aren’t fair, they have more bargaining power than they realize to demand better terms without having to throw out the baby with the bath water.
The turn to zero sum thought
Technological change has often brought analogous cycles of backlash. For example, when global ride-sharing service Uber first appeared on the scene, taxi drivers protested amid anxieties about their jobs and future competition. Similarly, while the public broadly supports renewable energy, wind projects have elicited local opposition, with the prominent Nantucket case percolating through the legal system for years because of impact on the local community. Innovations can arrive somewhat suddenly, and certainly before we fully understand their impacts. Initial enthusiasm can produce an equal and opposite reaction of anxiety.
Data centers are following that pattern. I watched that cycle unfold firsthand in Lansing, New York. In 2019, the owners of the coal-fired Cayuga Power Plant on Cayuga Lake in upstate New York proposed retiring the plant and repurposing the plant as a data center. Upstate New York has an unusually low-carbon power mix, anchored by Niagara hydropower and the large nuclear complex near Oswego, which made the coal-fired plant superfluous and offered reliable alternative power for a data center.
Environmental groups lauded the proposal as a “big clean-power win,” and Fossil Free Tompkins, a local advocacy group, said they were “very excited about the project.”
The project first stalled in 2020 because the developers only received two megawatts of low-cost hydropower rather than the 25 megawatts it needed to begin construction. It then ran into a local land use and zoning moratorium that lasted throughout 2025 before New York Governor Kathy Hochul threw the project into further uncertainty with a one-year moratorium on new hyperscaler data centers this past July. The coal plant had long since shuttered but left the smokestack aesthetic on the lake. No data center is close to operational.
By 2025, when the land use and zoning moratorium went into effect, the political landscape had changed considerably as the widespread use of AI and concerns about its consumption of energy and water, its impact on employment and pollution, and other threats to society were on the rise. In 2018, data centers consumed 76 terawatt hours of electricity (or 1.9% of national energy consumption). By 2023, consumption had jumped to 176 terawatt hours, or 4.4%, with ongoing construction projects across the country making that trajectory even more formidable.
At the same time, electricity was becoming more expensive. At the national level, the average retail price consumers pay for electricity rose one-third from 2019 to 2025. Locally, the electricity provider, NYSEG, had implemented sharply higher delivery charges and proposed increases that would amount to an additional monthly increase of a staggering 24% per month.
The national debate invariably shaded the local concerns. Lansing residents already faced starkly higher electricity bills and the prospect of an energy-consuming data center could hardly have assuaged concerns about future costs.
Ironically, for years communities had competed to attract data centers and offered generous tax incentives. Now they have lurched to the opposite conclusion that they should keep the data centers out. Indeed, the United States has more than 4,000 data centers, which receive tax incentives in 38 states. Spending on data centers in 2025 was $61 billion. Yet, now more than 16 states, including Virginia and Pennsylvania, which at one time wooed data centers, have introduced legislation restricting the construction of new data centers, against the backdrop of public opinion polls showing strong opposition to local data centers.
Rediscovering win-win logic
There is a different way to see the problem. Drawing on economist Ronald Coase, who had the insight that when an economically valuable activity imposes costs on someone else, there may be gains from bargaining for both local communities and data center owners. The local community, in this case, need not accept an externality or prohibit the construction of new data centers, but can negotiate who bears the costs.
It is this Coasean logic that should inform the construction of data centers more than it currently has. Companies need something that the local communities hold, which is the land, infrastructure, and the permission or siting. Those assets have become increasingly valuable, which gives the communities leverage.
Admittedly, the logic requires a clear accounting of the economics of data centers that is complicated by the fact that these are local builds each with their own circumstances. Some things are fairly constant, such as the fact that each data center produces hundreds of temporary construction jobs but comparatively fewer longer-term jobs. One study of the contribution to tax revenue in Virginia showed that data centers accounted for between less than 1% to 31% of total local revenue; in Loudoun County, with the highest concentration of data centers, data centers represent 40% of the county’s overall budget.
In Lansing, New York, the data campus, if fully taxed at its estimated value (based on the initial build-out), would produce $8.7 million in local property taxes, about 15% of the school’s operating budget. But those contributions are not predetermined. Localities can negotiate payments in lieu of taxes, which are often tax incentives in which case the payment would not exceed what the company is otherwise required to pay in property taxes. But the company could pay for additional infrastructure improvements that of course benefit the project but also the broader community, or contribute to municipal services, in which case the payment would be higher.
A more difficult problem arises with costs whose magnitude is more uncertain or contested. Environmental and infrastructure costs, for example, are not one-size-fits-all. Facilities vary in their size, cooling, energy requirements, and the communities that host those data centers also vary in their grid capacity and electricity generation. A data center that repurposes a power plant using industrial land, closed-loop cooling, and a regional electricity supply of hydro and nuclear energy, such as that in Lansing, imposes lower costs on the community. In comparison, Amazon’s massive site in Indiana required razing 1,200 acres of farmland and Google’s data center in The Dalles, Oregon, which was an early generation system built beginning in 2006, used 355 million gallons of water in 2021 for evaporative cooling. Google’ data center in particular has contributed to the reputation of data centers as water guzzlers, but water consumption is not an intrinsic feature of data centers. Newer technologies that rely on closed-loop systems recirculate their cooling water. The Cayuga Data Campus was that type of system and required an initial charge of 300,000 gallons but no additional water draws for cooling.
Internalizing externalities that vary so much by location and technology becomes a difficult but not impossible task. Companies can finance infrastructure that offsets the costs, as Google announced in 2026. They can adopt closed-loop systems, fund wetland restoration where construction disrupts farmland, and invest in sound barriers to mitigate noise. These examples suggest context-specific responses that do not require settling on what the externality’s dollar amount might be. They also point to the reason why the early practice of blindly welcoming data centers with tax subsidies and now rejecting data centers largely based on national narratives is unwise. Both of these approaches assume that the costs and benefits are fixed.
The Coasean question amounts to finding a bargain that makes both sides better off. A data center should not raise the electricity bills of local families. The same should apply to water infrastructure, noise mitigation, roads, and other costs.
But companies have resources that the local communities often need. I watched my own community cut back on snow removal, close classrooms, and raise taxes as business has left the region and the tax base has contracted. In places like this, the relevant concern should not just be whether a new industrial project imposes costs, which themselves can be monitored and mitigated. It should also consider the possible benefits. In Lansing, those potential gains were substantial: millions of dollars for local schools, additional funding for grid upgrades, and proposals to improve the community greenspaces and trails. Rather than shut down the proposal altogether, the community would have been better served to assess the externalities and figure out how to elicit benefits.
The logic underlies the idea of a “data center dividend,” in which the revenues that data centers generate reinvest in the local communities in the form of tax relief, electricity bill credits, or even direct payments. The approach gives communities direct stake in the gains associated with underwriting the AI economy, helping to reverse the dynamic of privatizing the gains and socializing losses.
One caveat to this framework is that it assumes communities can make a dispassionate assessment of the costs and benefits of their local project. However, the issue has taken on a Manichean political valence: “We hear what the firm is saying but do not trust corporations” is an expression often heard in the Lansing context. Such logic makes any type of compromise seem like capitulation.
The point of contracts and regulation is that trust and incentives do not always converge. For example, rather than talk past each other on what the company might or might not do in the future, the two sides could negotiate around the disagreement on water usage, electricity costs, or noise. Trust in the company’s word does not need to be part of the equation as long as the two sides can negotiate protections.
Such bargaining is increasingly at odds with the current populist moment, where opposition to data centers has become low-hanging political fruit on both sides of the aisle. It is an easy way to signal that elected officials are standing up to Big Tech and defending household budgets. Right now, however, their opposition is tantamount to political theater rather than stewardship. Acting in the interest of their constituents would mean helping to negotiate better terms for technological developments that protect residents from costs and capture more of the benefits.
Author’s Disclosure: In 2025, the author chaired an independent assessment of the proposed Cayuga Data Campus in Lansing, New York. The assessment was commissioned by the project developer, and the author was compensated for her time. The author retained independent authority over the assessment’s methodology, assumptions, and findings. You can read our disclosure policy here.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
Subscribe here for ProMarket’s weekly newsletter, Special Interest, to stay up to date on ProMarket’s coverage of the political economy and other content from the Stigler Center.
