Americans’ AI hate wave might just be gathering steam: Data centers could hike power costs in some states over 50% by 2030

U.S. electricity costs may rise 6% to 29% by the end of the decade, with Virginia generation costs potentially up 57%, as data center demand drives a larger share of power consumption. The study says data centers’ share of U.S. electricity use climbed from 1.9% in 2018 to 4.4% in 2023 and could reach as much as 17% by decade-end. With utilities requesting a record $31 billion in rate increases in 2025, the article highlights rising consumer costs, heavier reliance on natural gas, and growing opposition to new AI data centers.

Analysis

The market is still treating AI infrastructure as a software-capex story, but the binding constraint is increasingly physical: interconnection queues, gas turbine lead times, transformer availability, and local permitting. That shifts value away from pure-play AI beneficiaries and toward the bottleneck suppliers that monetize grid scarcity, especially firms with pricing power in transmission, switchgear, metering, and utility engineering services. The second-order effect is that every incremental GW of data-center load raises the marginal cost of serving not just the hyperscalers but the entire regional rate base, which creates a political feedback loop that can slow buildouts faster than economics alone.

This is bullish for legacy generation and grid-enabling infrastructure over a 12-24 month horizon, but the trade is not simply long “energy.” If policymakers reintroduce targeted subsidies or fast-track transmission, the benefit shifts from gas-heavy incumbents toward renewables, storage, and equipment vendors. The key risk is that the market overestimates the speed at which utilities can pass through costs; if regulators force data centers to self-fund transmission upgrades or require dedicated power contracts, the consumer inflation pass-through weakens and the headline utility pain becomes more localized than national.

The broader contrarian point: consensus is likely underpricing the duration of the bottleneck but overpricing the uniformity of the outcome. Regions with unconstrained renewables and strong transmission buildout can absorb load with less margin pressure, while congested regions become the real winners for merchant power and gas peakers. That dispersion argues for pair trades rather than a simple sector beta expression, and for option structures that monetize volatility around regulatory decisions and siting approvals over the next 6-18 months.

Key Decisions for Investors

• Long VRT / short XLK on a 6-12 month horizon: own the grid and thermal-management bottlenecks that get paid regardless of which AI model wins; trim if AI capex slows materially or interconnection backlogs start clearing faster than expected.
• Long CEG or NRG vs short a data-center-exposed REIT basket over 3-9 months: power scarcity should support merchant generation margins while rising site opposition and higher utility costs pressure expansion economics for colocations.
• Buy 6-12 month call spreads in ET or KMI: if gas becomes the marginal fuel for incremental load, volumes and contract extensions should improve; structure with spreads to limit downside if subsidies tilt the mix back toward renewables.
• Long ETN / hub-and-spoke utilities equipment names vs short XLU for 6-18 months: capex follows congestion, not sentiment, and equipment suppliers can reprice faster than regulated utilities can earn returns.
• Event-driven hedge: buy downside puts on select AI-infrastructure proxies into major state siting hearings or PUC rulings; the risk/reward is favorable because permit denial can gap projects, while upside from approvals is usually slower and more incremental.