Forget Nvidia. This ETF Could Be the Next Big Winner From the AI Boom

The article argues AI’s power constraint could make electric utilities the next “AI winners,” citing Nvidia’s $100B+ support for OpenAI’s Ohio data center (ramping to 8 GW) as an example of surging demand. It highlights the Vanguard Utilities ETF (VPU) with ~70 holdings (e.g., NextEra, Duke Energy, Southern Company, Constellation) and notes a ~2.7% dividend yield while emphasizing upside from grid upgrades and data-center capacity expansion. Risks include regulated returns on capex and community pushback against data centers, implying the thesis is constructive but not guaranteed.

Analysis

The market is likely to overpay for the broad “utilities as AI winners” story and underpay for dispersion within the sector. In regulated names, load growth does not automatically translate into earnings because ROE is capped, capex is lagged through rate cases, and financing costs can eat most of the upside; the real economic rent tends to accrue to the few assets with scarce baseload or transmission leverage. That makes CEG the cleaner AI-power expression than a diversified utility basket, while VPU is mostly a low-beta duration trade dressed up as AI exposure.

The first-order catalyst is not just higher electricity demand, but who gets contracted vs. who gets stuck waiting on permits, interconnection, and community approvals. Over 1-3 months, headlines on data-center PPAs, rate-case outcomes, and grid expansion approvals should drive relative performance between merchant power, regulated rate-base names like NEE/AEP, and slower compounders like SO/DUK. If hyperscalers keep shifting toward behind-the-meter or direct procurement, the upside to regulated utilities is materially capped.

Contrarian view: consensus is treating “power demand” as a clean bull case, but the bottleneck can just as easily delay AI monetization rather than boost utility profits. If power constraints persist, the near-term risk is not that AI leaders collapse; it is that NVDA/MSFT/GOOGL capex cycles stretch out and multiple expansion slows while capital gets reallocated to generators and grid-linked assets. The thesis fails if financing conditions improve sharply, regulators greenlight faster recovery, or utility earnings guidance starts showing real EPS accretion rather than just asset growth.

Key Decisions for Investors

• Long CEG / short VPU for the next 3-6 months: best way to express AI power scarcity versus the market’s tendency to bid the whole utility basket; exit if regulated utility guidance starts inflecting faster than merchant power pricing.
• Buy NEE on pullbacks for a 12-18 month hold: better risk/reward than SO/DUK because it can monetize growth capex without needing a full valuation reset; invalidate if higher-for-longer rates materially raise its funding cost and delay project returns.
• Underweight SO and DUK versus AEP/NEE until rate-case evidence proves actual EPS conversion from data-center load; these names can see headline enthusiasm without matching earnings delta.
• Treat VPU as a hedge, not a primary AI beta bet; if you need utility exposure, prefer stock selection over the ETF because the expected return is likely capped by regulated economics.
• Watch for a reversal signal in upcoming utility guidance and interconnection approvals; if 2026 load forecasts do not translate into upward EPS revisions within 1-2 quarters, fade the AI-utilities rerating.