A look at cost and viability concerns over small nuclear reactors in the US, as the government, companies, and investors bet $9B+ on SMRs to power the AI boom
Small reactors could prove too costly to be viable despite $9bn investment in power for AI boom
Context & Ripple Effects
The AI power search has moved from talks between technology companies and owners of existing nuclear plants to a broader revival for nuclear, geothermal, and storage providers. The more than $9 billion commitment to SMRs tests whether that demand can support new nuclear capacity rather than only power contracts for existing US reactors.
The stakes extend beyond a single reactor design: AI-related electricity demand has already been framed as a lifeline for energy suppliers facing difficult financing conditions, including nuclear, geothermal, and storage developers. Cost concerns now challenge whether SMRs can convert that renewed interest into commercially viable projects.
First-order effects
- SMR developers and their government, corporate, and investor backers face a sharper viability test: committed capital does not resolve the reported risk that reactor costs remain too high.
- AI infrastructure operators seeking steady power gain another prospective supply option, but must account for the possibility that SMR projects will not deliver economically.
Second-order effects
- If SMR economics remain unconvincing, power buyers are likely to place greater weight on existing nuclear capacity and other clean-power options already being pursued for data-center demand.
- Financing for advanced energy projects may become more selective, with backers putting greater emphasis on cost execution rather than AI-linked demand alone.
Third-order effects
- The episode points to AI infrastructure becoming a utility-planning issue: electricity availability and project economics may increasingly shape where and how compute capacity is built.
- If the pattern holds, the clean-power market for AI will reward technologies that can pair reliable output with credible delivery costs, rather than benefiting nuclear proposals simply from association with AI demand.
The trend: AI’s power requirements are pulling energy technologies into the compute supply chain, while exposing execution and financing risk as the limiting factors.