Half of all new electricity demand in the U.S. last year came from data centers—just as public opinion of them plummets

U.S. energy demand rose 2% in 2025, with data centers accounting for about 50% of all electricity demand growth and expected to remain half of growth through 2030. The surge is fueling political backlash over higher power bills, water use, and local property impacts, with at least 16 data centers blocked or delayed last year worth $64 billion. The article points to growing regulatory and community resistance that could slow AI-related infrastructure buildout and pressure utilities, grid investment, and data center developers.

Analysis

The market is still pricing data-center buildout as a straight-line capex winner, but the more important second-order effect is regulatory and local-friction optionality. If permitting delays and statewide restrictions spread, the bottleneck shifts from compute demand to grid interconnection and siting, which compresses the timeline for revenue conversion across the AI infrastructure chain. That is bearish for the “picks-and-shovels at any valuation” trade, because backlog quality matters more than headline order growth once projects start slipping by 6-18 months.

The broader loser set is less obvious: regulated utilities, transmission operators, and rate-sensitive consumer sectors face a political backlash loop if AI-linked load growth keeps showing up in bills. Even where utilities can pass through costs, the lag creates working-capital stress and increases scrutiny on capex recovery, especially in election-heavy jurisdictions. That argues for a nearer-term widening of dispersion between utilities with clean load growth and those exposed to contested AI clusters.

The cleanest contrarian angle is that the public backlash may be more important for valuation multiples than for physical demand. Demand is not disappearing over the next 12 months, but the market can re-rate the entire AI infrastructure basket lower if investors start discounting higher permitting costs, longer payback periods, and political intervention. In that setup, the best longs are not the obvious data-center enablers but the beneficiaries of forced grid upgrades and power reliability spend, while the most vulnerable names are those whose bull case depends on uninterrupted hyperscaler absorption of capacity.

For SPGI specifically, the direct exposure is modest, but the adjacent implication is that project finance, infrastructure screening, and regulatory due diligence demand should improve as local opposition rises. That is a small but durable tailwind, not enough for a major fundamental rerate, but enough to keep the name resilient versus pure-buildout proxies if the anti-data-center narrative intensifies.

Key Decisions for Investors

• Short the most levered AI data-center infrastructure names on strength over the next 1-3 months; prefer a basket short versus semis/AI hardware long, because the risk is not demand collapse but schedule slippage and multiple compression.
• Go long XLU names with heavy transmission/grid-investment exposure and limited data-center concentration for 6-12 months; the trade benefits if the market starts paying up for reliability and regulated capex recovery while penalizing contested load growth.
• Pair trade: short data-center REIT / power-adjacent siting beneficiaries vs long industrial electrification plays over 3-6 months; if permitting friction rises, the value accrues to grid modernization rather than new-footprint capacity.
• Maintain a tactical long in SPGI on any infrastructure-regulation headlines; the setup is defensive with asymmetric upside from higher diligence, permitting, and credit/risk-analysis demand, but size modestly since the direct revenue impact is limited.
• For event risk, buy 3-6 month downside protection on the most AI-exposed utilities ahead of state-level legislative catalysts; the payoff is highest if affordability becomes a campaign issue and rate-case scrutiny widens.