US voters rejected AI data centres in 40 of 41 ballot measures

Voters opposed AI data-centre development in 40 of 41 local ballot measures held Tuesday across 10 states, with Ohio accounting for 21 measures. Democrats who campaigned against the build-out also won governor’s races in Texas, Ohio and Georgia, signaling local political resistance to data-centre expansion.

Analysis

Treat this as a permitting-risk signal, not evidence that U.S. AI capacity plans are being canceled. The market mechanism is a higher probability of delays, added mitigation costs, and less predictable approvals for greenfield projects—especially where local votes can change zoning or infrastructure decisions. That can favor operators with powered, permitted sites over developers whose value depends on undeveloped land, while shifting projects toward less restrictive jurisdictions or behind-the-meter power. Utilities face a two-sided exposure: data-center demand supports load growth, but project delays could leave them seeking recovery for infrastructure built ahead of demand. The electoral results do not establish that anti-buildout sentiment caused the broader wins or that governors can directly stop local projects.

Near term, expect headline volatility in data-center development and related infrastructure names; there is not enough information here to justify a broad short on AI demand. Over 1–3 months, the key test is whether local measures are legally binding and whether elected officials propose concrete zoning, power-allocation, water-use, or permitting changes. Over 6–18 months, persistent constraints could redirect capex and raise the value of existing powered capacity, but migration may dilute the effect on aggregate construction. The contrarian point: local resistance may redistribute projects rather than reduce total AI infrastructure investment. The signal weakens if measures prove advisory, approvals continue on schedule, or hyperscalers maintain build plans without material cost or timing revisions.

Key Decisions for Investors

• Avoid a thematic short of AI infrastructure on this evidence alone. Put greenfield-heavy data-center developers on a watchlist and verify project locations, permitting status, and exposure to the affected ballot jurisdictions before sizing any position.
• Consider a relative-value bias toward operators with already powered and permitted capacity over speculative land-and-development exposure; keep it conditional until local measures’ legal effect and company-level project exposure are confirmed.
• Monitor utility and power-infrastructure names for both sides of the trade: delayed data-center load can pressure expected growth, while grid upgrades and on-site generation needs may redirect—not eliminate—spending. Favor no sector-wide utility position without evidence on cost recovery and committed customer load.
• Escalate the risk if officials introduce binding restrictions or companies disclose project delays, capex deferrals, or higher permitting costs. De-escalate if measures are nonbinding and project schedules and investment guidance remain intact.