Nuclear energy to power data centers in Ohio, Pa.

The market is starting to price nuclear less as a “power commodity” and more as a contracted infrastructure layer for AI. That shifts the earnings quality question from spot power prices to long-duration offtake, plant life-extension capex, and regulatory execution—i.e., the real winners are owners of already-permitted baseload assets with low reopening friction, not greenfield developers. The second-order effect is that every incremental MW locked into data-center PPAs tightens available dispatchable supply in PJM, which should support local capacity pricing and make remaining merchant nuclear plants more valuable than the headline deals imply.

The hidden bottleneck is not demand, it’s schedule risk. Restarting or re-licensing plants is a multi-quarter-to-multi-year process with heavy dependence on NRC timing, equipment refurbishment, and grid interconnection; that means the equity reaction likely front-runs cash flow by 12-24 months. In the interim, turbine/generator, switchgear, cooling, and nuclear-services suppliers may capture a better near-term economic uplift than the plant owners themselves, because capex is immediate while revenue recognition is delayed.

Contrarianly, the consensus may be underestimating the political/operational fragility of “clean, reliable” nuclear headlines. Any safety incident, NRC delay, or cost overrun would quickly re-rate the theme, and the market is likely overconfident that AI demand alone can justify every restart at current valuations. The better trade is to own the assets with contracted downside protection and avoid paying full multiple for early-stage advanced nuclear concepts that still face financing, licensing, and commercialization risk.