New Study Finds No Link Between Rising Electricity Rates and AI Data Centers

The Institute for Energy Research finds no statistically significant correlation between the number of AI data centers and consumer electricity prices; the top 10 data-center states averaged 14.46¢/kWh in 2025 versus 14.39¢/kWh for other states. U.S. data-center consumption rose from 76 TWh in 2018 to ~176 TWh in 2023 (+131%), and Meta’s metered usage climbed from 6.97 TWh (2020) to 18.06 TWh (2024, +159%), yet these demand surges do not explain regional price spikes. The report undercuts arguments for broad moratoria on data centers, though legislation and political actions (a Ratepayer Protection Pledge signed by seven firms and a proposed federal moratorium) mean policy risk remains.

Analysis

The IER finding exposes a key structural truth: hyperscalers drive where power is built, not how retail tariffs are set. Because hyperscalers site by lowest delivered cost and execute large-scale PPAs or build owned generation, the marginal load from AI is being internalized by the corporate procurement stack rather than socialized through residential rate increases; that shifts the economics away from utilities’ commodity margins and toward regulated cost recovery and fixed charges.

Second-order winners are companies that can monetize both compute and energy — firms that lock in long-term PPAs, deploy behind-the-meter generation, or own the data-center real estate. Conversely, businesses that depend on unchanged retail tariff structures (small co-location operators, municipalities with thin tax bases) will see margin pressure as regulators reallocate costs through demand charges or higher fixed customer fees over the next 12–36 months. Transmission constraints create a multi-year bottleneck: until lines are built, localized congestion will raise delivered power for new campuses, favoring players with balance-sheet capital to pay for customer-funded transmission upgrades.

Policy and headline risk remain the wildcards. A federal moratorium or aggressive state rate-case could pause greenfield data-center projects within months, compressing capex plans and creating a 20–30% re-rating risk for exposed builders and hyperscalers with heavy near-term commitments. Offsetting catalysts that would reinforce the current trajectory include accelerated corporate RPP-style commitments and large utility-approved PPAs; monitor state PSC dockets and FERC transmission orders over the next 3–18 months for tradeable inflection points.

Key Decisions for Investors

• Long MSFT (initiate 2–4% portfolio position, accumulate on up to 5% pullback). Timeframe 6–12 months. Rationale: strongest integrated ability to monetize AI and internalize power costs via PPAs/onsite generation. Target 20–30% upside vs 10–15% downside on macro/regulatory shock; set stop at -12%.
• Buy GOOGL (GOOG) 12–18 month call spread (modest notional). Timeframe 12–18 months. Rationale: high margins on AI services with durable energy procurement reducing opex volatility; use spread to cap premium and target ~2–3x payoff if adoption accelerates. Risk limited to premium paid.
• Pair trade — Long ORCL / Short META, equal notional, 6–12 months. Rationale: ORCL benefits from hybrid/cloud enterprise demand and lower exposure to raw compute capex and regulatory heat; META is most exposed to capex-heavy AI builds and political backlash. Target 15–25% relative outperformance; cut pair if META posts sequential margin recovery or ORCL misses cloud cadence.
• Tail hedge: Buy 3–6 month ATM puts on META (small, defined-risk position). Timeframe 3–6 months. Rationale: insures against near-term moratoria or punitive state rulings that would disproportionally hit the company with largest incremental consumption; loss limited to premium, payoff if share price falls >10–15%.