Artificial intelligence is driving a rapid expansion in data-center electricity demand, with the IEA estimating that global data-center consumption was about 415 TWh in 2024 and could rise to around 945 TWh by 2030 in its base case. At the same time, modern AI workloads require higher rack densities and more demanding thermal management, making cooling a first-order design constraint rather than a secondary facility service. This paper evaluates a conceptual architecture that pairs offshore small modular reactors (SMRs) with ocean-water cooling for next-generation AI campuses. The analysis is based on a literature review of AI energy demand, SMR development status, seawater and underwater cooling systems, and the emerging discussion of nuclear-powered data centers. The paper argues that the concept is technically plausible, especially when treated as a modular offshore “energy island,” but is not yet commercially mature at hyperscale. The strongest case for the architecture is its combination of firm, low-carbon electricity and access to a vast thermal sink, which together address the two biggest infrastructure bottlenecks for AI: power availability and heat rejection. The main constraints are first-of-a-kind capital cost, regulatory complexity, marine corrosion and logistics, environmental licensing for thermal discharge, spent-fuel governance, and cybersecurity/physical security at offshore sites. Illustrative engineering estimates show that a 300 MW(e) SMR-class module can support a substantial AI campus, while deep-ocean or seawater cooling can materially reduce cooling energy and freshwater demand. Overall, the study concludes that offshore SMR + ocean cooling is best viewed as a credible long-term systems concept for coastal AI hubs, especially where land, freshwater, and grid capacity are constrained, but it requires phased demonstrations, robust environmental monitoring, and advanced safety case development before commercial deployment. (IEA)
Fahad Kareemuddin Kareemuddin (Fri,) studied this question.