The Hidden Winners of the AI Power Crunch: 3 Utilities to Watch

AI-driven electricity demand is creating multi-year grid interconnection and equipment backlogs, pushing investors toward utility and power producers tied to hyperscaler contracts. Constellation Energy highlights 22 GW of nuclear capacity and multiple 20-year PPAs (including a 20-year Microsoft deal restarting Three Mile Island Unit 1 at 835 MW), while the stock is down 35% from its 52-week high and trades around 20x forward earnings. Vistra adds upside via merchant power exposure plus new 20-year Meta nuclear PPAs (2,600 MW) and AI-linked data center partnerships (Helix backed with $10B+ commitments), while NextEra pairs AI tailwinds with regulated stability, including a record 35.1 GW renewable/storage pipeline and a $100B DOE-linked AI data center campus partnership in Kentucky.

Analysis

The market is likely still underestimating where the economic rent accrues in AI power: not to the hyperscaler building the load, but to the owner of scarce, dispatchable electrons and interconnection rights. That favors CEG and VST more than headline AI beneficiaries because contracted power plus merchant optionality can reprice faster than regulated utility earnings; the key is that these assets sit in constrained nodes where replacement capacity is slow, so the scarcity premium can persist for several quarters.

The softer read is that NEE’s AI angle is more of a long-duration pipeline story than a near-term earnings step-up. In a world where investors are paying up for immediate scarcity, regulated growth plus renewables/storage can look like the right business but the wrong stock for the next 1-3 months unless the market starts to reward balance-sheet de-risking and dividend stability. Second-order winners are grid equipment, transmission, and gas pipeline names; second-order losers are merchant industrial power users and any datacenter buildout dependent on delayed interconnection queues.

The main risk is regulatory backlash: behind-the-meter arrangements and hyperscaler-specific deals invite scrutiny if retail rates or reliability deteriorate, and that can compress merchant power multiples quickly. A second tail risk is that the narrative is already crowded; absent fresh contract wins or upward revisions to forward power pricing, the trade can stall even if the structural thesis is correct. The thesis is falsified if PJM/ERCOT power forwards roll over, if contract announcements slow materially over the next 1-2 quarters, or if state/federal intervention limits the economics of co-located generation.

Key Decisions for Investors

• Long CEG / short XLU for 1-3 months: express scarcity pricing and contracted baseload torque while hedging broad utility beta; attractive if power forward curves stay firm and CEG holds its relative strength versus the sector.
• Long VST / short NEE for 2-4 months: VST has more earnings convexity to wholesale pricing and data-center optionality, while NEE’s AI upside is slower-burn and more likely to be treated as bond-proxy utility duration.
• Buy CEG on pullbacks only after confirmation of another hyperscaler-related contract or upward revision to merchant power guidance; otherwise keep as a watch item because the stock can drift if the market has already priced the narrative.
• Reduce exposure to the most obvious AI software beneficiaries on power-cost risk if datacenter capex plans begin slipping; watch MSFT and META capex commentary as the first falsifier for the power-demand thesis.
• Set an alert on PJM regulatory headlines and wholesale power forwards: if policy rhetoric turns into rate or co-location restrictions, take profits on merchant power longs quickly because multiple compression can happen before earnings revisions do.