The AI-Driven Rise in Power Bills Are Causing a $25 Billion Problem for Utility Stocks

AI-driven electricity demand is expected to materially accelerate, but it’s coming alongside tighter customer affordability: U.S. utility unpaid bills rose from ~$15B (2022) to ~$25B (2025), with more shutoffs as bills climb. Data-center-adjacent electricity prices in Virginia rose by 260%+ over five years, pressuring regulators on rate increases needed for capex. The article argues regulated utilities face a balancing act (e.g., NextEra’s proposed Dominion acquisition is contingent on rate approvals), while unregulated/alternative power plays (Constellation, Brookfield Renewable, Bloom Energy with a $6B 2026 product backlog) may be better positioned—though Bloom is also up 800%+ over the past year and is pulling back.

Analysis

The market is likely overestimating the equity benefit of AI-driven load growth for regulated utilities and underestimating the political cost of financing it. If bad debt is already rising before the full AI load ramp, then the first-order winner is not the rate base owner but the supplier that can monetize megawatts without waiting for a multi-year rate case cycle. That favors contracted/merchant generators with pricing power and long-duration PPAs, while distribution-heavy names absorb the near-term backlash from higher bills and collection pressure.

The second-order risk is that regulators respond to affordability headlines by stretching amortization periods, trimming allowed ROEs, or requiring more cost deferral. That creates a timing mismatch: capex shows up immediately, but earnings accretion lags 12-24 months and can be diluted if load forecasts prove too optimistic. In contrast, entities exposed to bilateral power contracts can reprice faster and benefit from scarcity without being the public face of rate inflation.

Contrarianly, the consensus is treating AI power demand as a uniform bullish factor across the utility complex. It is not: the structural winners are the names with flexibility in siting, contracting, and balance-sheet capacity, while the losers are the utilities with the least ability to pass through costs quickly. The AI buildout may also accelerate behind-the-meter and non-grid solutions, which could cap regulated load growth even as total power demand rises. Over 6-18 months, that split should widen.

Key Decisions for Investors

• Long CEG vs short D or NWE for the next 3-6 months: CEG has faster monetization of AI demand through contracted pricing, while regulated utilities face political and regulatory lag. Falsify if state commissions approve above-trend ROE/rate increases without delay.
• Use NEE as a quality hold, not a momentum chase: own only if buying into weakness after the market digests rate-case scrutiny. The upside is slower but more durable than pure regulated peers; risk/reward improves if the stock de-rates toward utility-average multiples.
• Avoid chasing BE after the prior run; consider only a tactical short on strength if valuation remains detached from backlog conversion. The thesis breaks if backlog converts into margin expansion over the next 2 quarters and management raises 2026-2027 guidance.
• Watch BEP/BEPC as a cleaner AI-power alternative to regulated utilities: long on pullbacks for a 6-12 month horizon if contract pricing and asset availability remain intact. Key falsifier: rising financing costs or stalled project economics compress FCF yield.
• If forcing a sector expression, prefer long CEG / short a basket of D + NWE + BKH rather than a broad utility short. That isolates the difference between merchant/contracted power and rate-base exposure to affordability backlash.