Which stocks will benefit from AI power demand going nuclear?

AI data-center demand is supporting expectations for nuclear power, with data centers accounting for nearly 40% of PJM's latest capacity-auction charges, or about $6.3B, and operators benefiting from near-term reliance on existing plants, restarts and uprates. Uranium-related equities have lagged: Cameco was down 8.5% over one year, while Uranium Energy, Centrus Energy and Energy Fuels fell 48.9%, 69.4% and 60.5%, respectively; the Global X Uranium ETF declined 32.1%. The article notes reactor construction timelines, valuation risk and dependence on non-binding agreements for early-stage microreactor companies.

Analysis

AI power demand is a nearer-term scarcity and contract-pricing story than a reactor-build story. The key equity distinction is who can monetize firm power soon: CEG and VST have more immediate exposure to existing generation, while BWXT and the fuel-cycle names depend on longer-dated deployment, procurement, or policy pathways. Even for generators, headline demand is not equivalent to incremental earnings: contract pricing, delivery location, congestion, and the cost of serving new load determine whether value reaches shareholders. Regulated utilities such as AEP and SO may see load and rate-base growth, but returns are less directly linked to spot scarcity than merchant generation.

The uranium-equity drawdown is not automatically a bargain signal. Miners and enrichment exposure do not capture data-center power pricing one-for-one, and new reactor demand has a long conversion path from announcements to fuel orders. Conversely, if utilities lock in fuel procurement while new nuclear demand becomes more credible, today’s weak fuel sentiment could reverse before new reactors are built. NNE’s non-binding commercial signal is not equivalent to a bankable order.

Over 1–3 months, focus on disclosed contract economics, capacity-market outcomes, and load/interconnection updates; over 6–18 months, uprates, restarts, and procurement decisions matter more. The thesis weakens if data-center projects are delayed, power contracts fail to clear attractive economics, or new supply and transmission relieve local scarcity. The contrarian opportunity is the timing gap—not a blanket claim that all nuclear-linked equities are mispriced.

Key Decisions for Investors

• Prefer a measured relative-value position: long CEG versus short URA, sized as a timing trade rather than a view that uranium demand is structurally impaired. Enter on strength in CEG only if company disclosures support incremental, economically attractive load contracts; cap risk if contract terms remain opaque.
• Keep VST as an alternative existing-generation exposure, but compare its disclosed generation mix and contract economics with CEG before choosing between them. Do not assume every data-center PPA earns a premium after delivery and congestion costs.
• Treat CCO, UEC, UUUU, and LEU as separate fuel-cycle exposures, not proxies for near-term data-center power prices. Revisit only when procurement, contracting, or policy evidence improves; verify each company’s exposure and the relevant fuel-cycle bottleneck first.
• Avoid paying for NNE’s MOU as if it were a committed deployment. Upgrade the view only on a binding order, financing evidence, and a credible delivery timeline; a canceled or delayed project would falsify the near-term commercialization thesis.
• Monitor contract disclosures, PJM capacity outcomes, project interconnection progress, and power-market pricing. If these show delayed load additions or no improvement in realized economics for generators, reduce the CEG/VST preference; if fuel procurement accelerates materially, reassess the short-URA leg.