Barclays Lists 21 Nuclear Stocks to Watch for the AI Infrastructure Demand Surge

AI infrastructure spending from Western hyperscalers and AI labs could exceed $1 trillion and top out around 2028, with Barclays estimating an incremental ~$300B above current consensus for the sector. The article highlights nuclear-exposed names: NextEra’s subsidiary completed a $3.75B debt offering, Constellation secured a long-term nuclear power supply deal for Walmart, and Duke Energy received up to a $61.8M U.S. DOE grant (alongside multiple firms reporting Q1 2026 EPS beats). Risks cited include power constraints, permitting hurdles, and labor shortages—supportive for nuclear/AI power infrastructure demand but not without execution risk.

Analysis

The tradeable edge is not “more AI power demand,” it is where the bottleneck converts into cash within 12-24 months. That favors service, grid, and equipment names with existing install base and recurring work over pure SMR optionality; the market is likely to keep overpaying for long-dated reactor narratives while underappreciating the boring spend on transmission, switchgear, cooling, and interconnection. If hyperscaler capex stays elevated, the first earnings surprises should show up in backlog conversion and pricing power at engineering/infrastructure vendors rather than in new nuclear megaproject revenue.

Within power, the clearest winners are utilities and independent power producers with hard, contracted nuclear or data-center adjacency, but the upside is more defensive than explosive. Their value is in scarcity: firm carbon-free baseload with existing permits and sunk capital, which should support higher forward multiples and better contract terms, while rate-regulated names still face financing drag and regulatory lag. The second-order loser is the “concept” basket of pre-revenue reactor developers, where the capital intensity and timeline mismatch make them highly sensitive to any delay in DOE funding, permitting, or utility procurement.

The consensus is missing that most AI infrastructure dollars will not go to reactors first; they will go to grid enablement, balance-of-plant, and utility capex. That makes the best risk/reward a relative-value long in infrastructure execution names versus long-duration nuclear speculation. The main falsifier is a sharp slowdown in hyperscaler capex or evidence that data-center load growth is being deferred beyond 2026, which would compress the whole theme’s multiple almost immediately.

Key Decisions for Investors

• Long PWR / short SMR as a 3-6 month relative-value pair: own the contractor with visible backlog and execution monetization, fade pre-revenue nuclear optionality that requires years of financing and permitting.
• Buy GEV on pullbacks, but treat it as a backlog-and-services trade rather than an SMR pure play; best entry is after any post-coverage-initiation dip, with a 3-12 month hold tied to order conversion and margin expansion.
• Overweight TLN versus broader regulated utilities (NEE, DUK, D) for 6-18 months: direct data-center power monetization should re-rate faster than rate-base names that need regulatory approval to earn their equity cost.
• Use WCC as a second-order winner on any weakness: if the thesis is real, the fastest revenue accrual is in electrical distribution and MRO, not reactor construction. Favor it over higher-beta nuclear developers.
• Watch CCJ/NXE as long-duration uranium optionality, not immediate catalyst names; only add aggressively if utilities sign incremental fuel-cycle contracts or if uranium spot clears a sustained higher range.