3 Nuclear Stocks Powering the Grid Long Before AI Made Them Trendy

The article highlights Cameco, Constellation Energy, and Vistra as lower-risk ways to gain exposure to AI-driven nuclear-power demand, with the IAEA projecting global nuclear capacity could triple by 2060. Uranium spot prices climbed from $35.00/lb at end-2020 to $89.68/lb by August, supporting projected 2025-2028 EBITDA CAGRs of 15% for Cameco, 38% for Constellation, and 15% for Vistra. Constellation is expanding through its $26.6B Calpine acquisition and long-term PPAs, while Vistra is building nuclear capacity despite battery-storage and regulatory setbacks; both are positioned to serve hyperscaler electricity demand.

Analysis

The investable bottleneck is dispatchable, contracted power rather than reactor-development optionality. CEG and VST monetize hyperscaler demand now, but their sensitivity differs: CEG is more exposed to long-duration nuclear repricing and execution on large asset integration, while VST retains greater ERCOT gas-power and retail-margin torque. This favors VST in a near-term heat-load or gas-price spike, while CEG has the cleaner 6-18 month thesis if corporate PPAs continue to clear above merchant power curves.

CCJ/CCO is not a pure uranium-beta trade after Westinghouse: higher fuel prices support mining economics, but the higher-margin strategic value sits in reactor services, outage work, and new-build supply-chain procurement. That makes CCJ relatively resilient if uranium spot retraces, but it also means a simple uranium-price multiple likely overstates upside. The more important catalyst is utility contracting volumes and term-price escalation, which can tighten the nuclear fuel market without requiring a further spot-price surge.

Consensus appears to be extrapolating AI load forecasts into all nuclear-adjacent equities. The weak link is timing: data-center interconnection and transmission constraints can defer incremental load by 12-36 months, while small-reactor names OKLO and SMR face licensing, financing, fuel-supply, and construction-risk simultaneously. Conversely, grid congestion can increase the value of existing generation in constrained regions, creating a structural premium for CEG and VST that is not captured by national power-demand estimates.

Key reversal risks are lower ERCOT/PJM forward power curves, a decline in natural-gas volatility, adverse capacity-market or nuclear-credit rulings, and hyperscaler capex cuts. For CEG, monitor contracted-price disclosures, Calpine integration leverage, and any restart-capex revision; for VST, monitor ERCOT reserve margins, retail churn, and battery-replacement costs. A sustained deterioration in 2027-28 power forwards would falsify the merchant-power component of both theses.

Key Decisions for Investors

• Prefer a 3-6 month long VST / short CEG pair on equal dollar exposure if the valuation discount remains at least 25% on forward EV/EBITDA after normalizing for announced acquisitions. VST offers greater near-term weather and ERCOT scarcity upside; risk is a sharp fall in gas prices or ERCOT forward curves, which would favor CEG's more contracted profile.
• Maintain CCJ/CCO as the quality uranium exposure, but size it as a nuclear-services and long-cycle contracting position rather than a spot-uranium proxy. Add only after confirming term-contract volumes and Westinghouse margin progression at the next earnings release; reduce if uranium term pricing stalls while mine operating costs rise.
• Avoid outright longs in OKLO and SMR for the next 6-12 months; use them only as a funded hedge against upside nuclear-policy surprises, financed by a small short basket of the two versus long CCJ. A definitive commercial financing package, NRC milestone, or binding utility offtake would invalidate the short leg.
• Set alerts on ERCOT summer forward prices, PJM capacity-auction outcomes, and U.S. natural-gas prices. A 10%+ decline in regional power forwards or a regulatory cap on capacity compensation should trigger profit-taking on VST and a reassessment of CEG's multiple support.