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

AI-driven electricity demand is forcing major utility capex and higher rates, widening bill-payment stress: U.S. utilities’ unpaid bills rose from ~$15B (2022) to ~$25B (2025), while shutoffs are increasing. In Virginia, power prices near data centers jumped over 260% (Bloomberg, late 2025), setting up regulatory risk for rate approvals. NextEra Energy (NEE) agreed to buy Dominion Energy (D), which could lift earnings if regulators approve rate increases; the article also highlights non-regulated AI power plays (Constellation, Brookfield Renewable, Bloom Energy), though Bloom is down from a prior ~+800% run as investors reprice growth.

Analysis

AI load growth is bullish for electricity demand, but the equity winners are not the same as the headline beneficiaries. The first-order benefit to regulated utilities is rate-base expansion; the second-order cost is political friction from higher bills, rising bad-debt expense, and slower tariff approval, which compresses allowed-ROE upside and makes smaller names like D, BKH, and NWE more fragile than the earnings narrative implies. In practice, capex can rise faster than cash conversion, and that mismatch usually hits the multiple before it hits EPS.

The cleaner expression is outside the fully regulated model. CEG should monetize incremental load fastest because merchant and contracted pricing captures scarcity without waiting for every dollar of investment to be blessed by a commission, while BE and BEPC benefit if data centers keep adopting behind-the-meter or grid-avoidance solutions. BE is already a crowded trade, so the better risk/reward is to wait for a pullback and let backlog conversion prove margin durability rather than chase momentum.

Contrarianly, the market may be underpricing affordability backlash: if delinquencies keep rising, regulators may force longer amortization periods, lower allowed returns, or special data-center tariffs, pushing the burden onto project timelines rather than utility earnings. The thesis is falsified if state commissions approve faster cost recovery, bad-debt trends stabilize, or AI load converts into materially higher realized ROE within 1-3 quarters rather than just larger capex budgets.

Key Decisions for Investors

• Long CEG / short D for 1-3 months into the next round of regulatory headlines; expect 10-15% relative outperformance if merchant power stays tight and Virginia policy remains restrictive. Exit if D secures pre-approved data-center riders or if CEG guidance stops implying pricing power.
• Own NEE as the best regulated-quality exposure versus BKH/NWE/D, but do not chase it aggressively; it should outperform the small-cap regulated complex if affordability politics intensify over the next 3-6 months.
• Initiate BEPC on pullbacks, not at momentum highs; use a 6-12 month horizon and prefer it over BE for lower execution risk and better yield support. Thesis fails if contracted growth stalls or if market prices compress faster than backlog converts.
• Stay underweight BKH and NWE until there is clarity on merger synergies and rate-case recovery. These are the names most exposed to political backlash and bad-debt drag, with limited upside if AI demand merely shifts cost burdens onto ratepayers.