- AI data centers are the primary demand catalyst driving renewed US nuclear interest: their share of US electricity could reach 17% by 2030 (EPRI), up from ~5% currently, creating extraordinary demand for 24/7 carbon-free baseload power that solar and wind cannot match without expensive battery pairing — tech companies are signing 20-25 year power purchase agreements (Microsoft with Three Mile Island, Google with Duane Arnold) that provide the revenue certainty historically absent from nuclear financing; public support has followed, with 59% of Americans favoring more nuclear in 2025 (Pew), up from 43% a decade ago.
- The Trump administration set a goal to quadruple US nuclear capacity to 400 gigawatts by 2050, with 10 large reactors under construction by decade’s end; it committed more than $80 billion to Westinghouse AP1000 reactor construction, signed an executive order directing the NRC to evaluate new licenses within 18 months (vs. the prior 36-month average), and selected 11 projects for its DOE reactor pilot program including Sam Altman-backed Oklo — which also signed a deal with Meta alongside TerraPower, backed by Bill Gates, which holds a construction permit for a reactor in Wyoming.
- Four nuclear startups achieved criticality — the ability to sustain a controlled fission reaction — by the July 4 deadline set by the DOE pilot program: Antares Nuclear, Valar Atomics, Aalo Atomics, and Deployable Energy; proven plant restarts are advancing faster than new builds: Constellation Energy targets a 2027 reopening of Three Mile Island (Microsoft PPA), NextEra targets 2029 for Duane Arnold in Iowa (Google PPA), and Holtec is planning to reopen Palisades in Michigan with state and federal support — while Peter Thiel-backed General Matter and a Centrus Energy subsidiary received $1.8 billion to build domestic HALEU enrichment capacity.
- The US faces a critical uranium supply constraint: Russia supplied roughly one-fifth of US nuclear fuel as recently as 2024, but an import ban tied to the Ukraine war is in effect with waivers expiring in 2028; HALEU fuel (5-20% enriched uranium-235) required by many SMR designs is only produced at commercial scale by Russia and China; the sole US commercial enrichment facility — Urenco’s New Mexico plant, a British-Dutch-German consortium — is expanding capacity by 50%, but the fuel supply gap represents a potential bottleneck that no amount of capital can quickly resolve for the most advanced new reactor designs.
What Happened?
The convergence of AI data center power demand, Trump administration nuclear policy, and private capital has created the most favorable environment for US nuclear power since the 1970s. The country’s 94 operating reactors — 90% built in the 1970s-80s, with only 3 new reactors completed this century (including the Vogtle units, which came in 7 years late and 2x over budget) — are being supplemented by restarted plants and a new wave of SMR startups. The DOE pilot program selected 11 projects across different reactor technologies. Four startups hit the July 4 criticality milestone. Google separately invested in Commonwealth Fusion Systems, a fusion startup that uses high-temperature superconducting magnets rather than fission.
Why It Matters?
Nuclear sits at the intersection of three of the most powerful economic trends of 2026: AI infrastructure buildout, energy security, and the return of industrial policy. The Vogtle cautionary tale has not deterred new investment because the demand environment has fundamentally changed — tech companies willing to sign multi-decade PPAs provide the revenue certainty that historically made nuclear financing impossible. The supply chain constraints (skilled labor, uranium enrichment, specialized components) are real but addressable with capital and time. What cannot be shortcut is physics: nuclear plants take 10-15 years from conception to operation under the best conditions, meaning the choices made now will determine the US electricity supply in the 2035-2040 window when AI power demand is projected to peak its growth rate.
What’s Next?
Near-term milestones to watch: the Palisades restart (Holtec targeting 2026) and Three Mile Island (targeting 2027) will be proof of concept for the restart strategy and reveal whether regulatory fast-tracking actually works. For SMRs, NuScale has NRC design approval; TerraPower’s construction progress in Wyoming is the leading indicator of whether the SMR industry can beat the consensus 2032-2035 first-commercial-operation timeline. The HALEU fuel supply is the sleeper risk: without a domestic alternative to Russian and Chinese production at commercial scale, the most advanced SMR designs face a fuel constraint that mid-2026 investment decisions cannot resolve in time for early-2030s deployment schedules.
Source: Bloomberg
