Can small modular reactors (SMRs) reignite growth in the nuclear industry, by halving capex to $3,000-4,000/kW, thus reducing levelized costs of zero-carbon electricity to 8-10 c/kWh? This 18-page report presents five considerations around nuclear SMR costs, to help understand the risks of persistently high capex requirements. We still see gas and solar dominating future load growth.
After 24-years in the wilderness, the British rock band, Pulp, released a new album in 2025, which suddenly rocketed to #1 in the UK charts. In one song, entitled Grown Ups, a pub discussion is recalled, where a friend “said he’d moved near the motorway, because it was good for commuting. And I laughed in his face because I thought he was joking. But then I looked in his eyes, and I saw, he was not joking. He was trying. Trying so hard to act like a grown-up”. This somewhat reminds us of the nuclear industry in 2025, for reasons on pages 2-3 of this report.
To emerge from its own quasi-wilderness, and rocket to #1 in powering AI data-centers (!), the nuclear industry particularly needs to improve its capex costs. The capex costs of nuclear projects are discussed, on a top-down basis, a bottom-up basis, and with a review of past projects on pages 4-5.
Small modular nuclear reactors (SMRs) aim to avoid the very high capex costs of on-site labor, when constructing nuclear power plants, by simplifying and modularizing key components, so that they can be built with more of a factory approach. We are currently tracking 35 next-gen nuclear companies. But there are five considerations for de-risking their capex ambitions.
(1) Smaller components cost more. We tabulate the costs of different types of nuclear-relevant equipment and how cost varies with unit sizing, on pages 6-7.
(2) Smaller reactors use more materials, equipment and possibly even more labor per MW of output, according to some studies, reviewed on pages 8-9.
(3) Safety requirements of nuclear reactors will always be more stringent than for other forms of power generation. We reviewed the 55 General Design Criteria from the US Nuclear Regulatory Commission to quantify where SMRs could save costs versus large PWRs/BWRs (80% of these safety standards are basically non-issues for gas, and 95% are non-issues for solar). See pages 10-14.
(4) Learning curve effects will therefore be needed, but what rate of learning is realistic? And how would this compare with the learning rates in new energies more broadly? (pages 15-16).
(5) Operational and reliability factors may also need to adapt, in order for SMRs to power AI data centers, compete with high-availability gas generation, or be deployed in densely populated areas (pages 17-18).
Our outlook for nuclear SMRs has somewhat shifted based on this analysis, including for de-risking our nuclear growth forecasts. Specific SMR concepts, that are discussed in particular, include those from NuScale, Oklo and Holtec.