AI on the Ballot: What Data Center Growth Means for Communities and Policymakers

On October 21, I attended the Brookings Institution webcast, “AI on the ballot: Data centers and the 2026 midterms.” Rep. Jim Himes (CT-4) opened the discussion, followed by a panel moderated by Valerie Wirtschafter with Brookings scholars Tony Pippa, Mark Muro, Kyle Chan, and Tom Wright. The conversation connected local disputes over data center development to larger questions about energy, national security, and the pace of AI policy.

Data centers are a community issue

Rep. Himes described a backlash against data centers that extends beyond places with expensive power. Residents have concrete concerns about electricity costs, water, and noise. He expects developers to make more attractive offers to host communities, while acknowledging that data centers provide relatively few ongoing jobs—the kind of durable benefit many communities want.

Tony Pippa’s account of a proposed data center and power plant near Thomas and Davis, West Virginia, showed why the process matters as much as the project. He identified three drivers of resistance: little early consultation, real environmental and economic risks, and distrust when well-resourced developers negotiate with local governments that may have limited staff or authority. In this case, state law places key siting decisions with the state and allocates 70% of the tax revenue to it.

Better deals start before approval

The panel argued that communities have room to ask for more than a late-stage benefits package. Options discussed included long-term local tax revenue, upfront community benefits agreements, investments in schools and hospitals, workforce training, and access to compute for universities and startups. Pippa urged a shift from treating benefits as compensation toward investing in a community’s capacity. These were proposals for stronger negotiations, not a claim that every project can deliver every benefit.

Mark Muro placed the backlash in a broader context: people want a say in how AI enters their lives. He pointed to a 2025 Pew survey in which only 14% of the public said citizens had substantial control over AI’s use, while nearly two-thirds wanted more management of its rollout. He also noted signs of changing industry practice, including announcements by Google, Amazon, and Anthropic that they would stop using nondisclosure agreements in this context.

Energy is the bottleneck

Kyle Chan drew a distinction between the United States and China: U.S. AI development is constrained by energy and grid connections, while China’s is constrained by access to advanced chips. China’s greater power-generation capacity does not, by itself, settle which country will lead in AI. Chip performance, efficiency, and the ability to assemble large training clusters also matter. For U.S. communities, Chan’s practical priority was faster power development and grid interconnection without shifting higher electricity costs onto residents.

Compute is also a security question

Himes expects AI data centers to be treated as critical infrastructure as more functions depend on them. He emphasized resilient, dispersed systems rather than relying solely on physical protection. He also raised concerns about advanced-chip diversion when major facilities are built in the Gulf.

Tom Wright distinguished between inference capacity, which generally needs to be closer to users because of latency, and training facilities, whose location is more flexible. That flexibility creates a strategic choice: decisions about large overseas training centers should weigh security and alignment with partner countries, not just economic incentives.

Looking ahead

Himes called for greater transparency and monitoring of frontier AI models, comparing the need for AI incident and near-miss reporting to aviation safety practices. The panel’s domestic priorities were more immediate: improve communities’ ability to negotiate, establish meaningful benefits, and expand the power supply and grid connections needed for development.

The central question is not simply whether data centers will be built. It is whether public officials and developers can give communities an early, informed role in deciding where projects belong, who bears their costs, and what lasting benefits they provide.