Nuclear Power Surge Driven by AI Demand and Policy Support

US nuclear power is poised for significant growth, with capacity expected to increase by 63% to 159 gigawatts by 2050, amounting to a $350 billion-plus investment. This surge is largely fueled by the increasing demand for reliable, carbon-free energy from AI data centers. Policy support and decreasing cost differentials with renewables are key factors propelling this growth, with early advancements expected from plant upgrades, restarts, and the eventual scaling of small modular reactors (SMRs) after 2035.

The rise in net nuclear capacity, projected to climb by 52 gigawatts from 2036-2050 at a 2.7% compound annual growth rate, will be primarily driven by policy incentives and the need for clean, dependable power by AI data centers. The forecast suggests that most capacity additions will occur post-2035, with potential constraints including permitting timelines, financing risks, and underdeveloped supply chains for SMRs. The scenario outlined by Bloomberg Intelligence anticipates new reactors beginning operation around 2038, aligning with historical installation rates seen in the 1970s-1990s.

By 2050, it is projected that there will be a net addition of 61 gigawatts of nuclear capacity, elevating nuclear power’s share of US electricity generation to 20% from the current 18% in 2024. In the shorter term, the US could see an increase of nearly 9 gigawatts in nuclear capacity by 2035, reaching just over 106 gigawatts. This growth trajectory will likely be driven by upgrades at existing facilities, reactivating retired units, and the completion of projects such as the V.C. Summer reactors in South Carolina. Major technology companies like Meta, Microsoft, and Amazon have entered into long-term nuclear power agreements to support their AI and cloud computing operations.

The estimated cost of adding 63 gigawatts of nuclear capacity by 2050 stands at $354 billion, with assumptions of new reactor start-ups commencing in 2038. The Vogtle plant in Georgia, the first nuclear project completed in the US in fifty years, had a construction period of 15 years and an estimated cost of $13,876 per kilowatt. Although AI data centers are pushing for the revival of shuttered reactors, only a few closed facilities are currently viable for restart by 2030, notably Holtec’s Palisades, NextEra’s Duane Arnold, and Constellation’s Three Mile Island Unit 1.

Private investment in new nuclear capacity may necessitate additional support beyond existing tax incentives, such as federal loan guarantees similar to those provided for the Vogtle reactors. The Trump administration’s executive orders advocating for streamlined reactor licensing and a 400-gigawatt nuclear target by 2050 signal strong government backing. As the cost discrepancy between nuclear and renewables is expected to diminish after 2027, nuclear power could become more competitive, especially with the expiration of key solar and wind tax credits.

The Energy Information Administration’s forecast, which suggests a decline in US nuclear capacity by 2050, contrasts with Bloomberg Intelligence’s more optimistic outlook driven by AI data center demand and government backing. The potential for nuclear power expansion is further supported by recent policy enhancements, such as expanded nuclear tax credits and accelerated permitting processes. Multiple states are considering new nuclear reactors, with companies like Constellation eyeing plant development in regions like upstate New York where they already have a presence.

Tags: regulatory

Read more on bloomberg.com