Americans would rather live near a nuclear power plant than a data center—by a lot

Gallup found 71% of U.S. adults oppose an AI data center in their local area, versus 53% opposing a nuclear plant, highlighting significant public resistance to data center expansion. The article emphasizes concerns around noise, traffic, utility bills, water use, and power demand as U.S. data centers consume about 4.4% of national electricity, with estimates rising to 12% by 2028. While the piece is not company-specific, it underscores a growing regulatory and social-license risk for AI infrastructure buildout.

Analysis

The market is still underpricing the social-license bottleneck in AI infrastructure, but the second-order implication is not “data centers get stopped” — it’s that project economics will shift toward incumbents with the best land, power interconnects, and municipal relationships. That favors integrated energy/infra ecosystems and behind-the-meter power solutions over pure-play compute builders, because the scarce input is increasingly permitting velocity, not capital. In that setting, the marginal loser is not AI demand itself but any developer reliant on greenfield zoning and public water access.

The most actionable read-through is for distributed generation and energy-management vendors: backlash against grid-heavy central buildouts raises the probability of on-site generation, microgrids, storage, and fuel-flexible backup becoming default design features rather than premium add-ons. That is constructive for companies that can shorten time-to-power and reduce visible community impacts. Conversely, large centralized utility load forecasts may look better on paper than reality if local moratoriums force geographic displacement and extend interconnection queues by quarters, not days.

For markets, the consensus is missing how quickly political resistance can reprice bottlenecks even when end-demand remains intact. This is a months-to-years regulatory risk, but the catalyst can hit in weeks if another county or state adopts a moratorium, forcing developers to re-route capex and delaying revenue recognition for equipment suppliers. The contrarian opportunity is that headlines look negative for AI, yet they may actually accelerate adoption of modular energy products and services that were previously optional.

The key overhang is valuation: if investors already own the AI-capex winners, they may be less exposed to siting friction than the market assumes, while the real beta shifts to power infrastructure names with execution leverage. The right trade is not short AI broadly; it is to fade the most location-constrained, water-intensive, or permit-dependent parts of the supply chain and own the picks-and-shovels that solve the constraint.

Key Decisions for Investors

• Long BE into the next 4-8 weeks as a relative beneficiary of the 'decentralized power for AI' narrative; use tight risk controls because the stock will remain headline-sensitive, but the asymmetry improves if state/local moratorium chatter expands.
• Pair trade: long BE / short a basket of exposed data-center developers and REITs with heavy greenfield dependence over 1-3 months; thesis is that permitting and utility access—not demand—become the binding constraint.
• Add a tactical long in grid resilience / distributed power names (e.g., microgrid and storage enablers) over 3-6 months; the upside is multiple expansion if AI siting standards shift toward behind-the-meter generation, with downside limited to policy delays.
• Avoid chasing pure AI infrastructure winners after local opposition headlines unless they have signed power, water, and zoning in hand; the risk/reward is poor on names that need multiple approvals, because one moratorium can push revenue out by 2-4 quarters.
• Monitor for a pairs opportunity in utilities exposed to large new loads versus industrials supplying on-site power equipment; if moratoriums spread, the latter should outperform on a 90-day horizon even if overall AI capex stays strong.