What happened
Published June 2, 2026 by Carnegie Endowment's Nuclear Policy Program, this 20-page paper examines how Alphabet, Amazon, Meta, and Microsoft are approaching nuclear power to meet the escalating electricity demands of AI data centers. The authors find that hyperscalers have so far acted overwhelmingly as energy offtakers—preferring power purchase agreements and minority venture stakes in small modular reactor startups—rather than taking on direct ownership or project development risk. The paper identifies three 'entanglement risks' that big tech has not yet fully priced in: reputational exposure if nuclear projects encounter safety incidents, nonproliferation concerns arising from new corporate actors entering a historically regulated industry, and long-term nuclear waste obligations. The authors conclude that 'the coming decade will require technology companies to decide how nuclear fits into their energy strategies—and grapple with the obligations that follow,' warning that current cautious postures may be insufficient as SMR timelines slip and grid-capacity constraints tighten.
Why it matters
AI infrastructure planning is now inseparable from energy strategy; boards and executives who treat the nuclear-AI nexus as an infrastructure-team problem are missing reputational, regulatory, and balance-sheet exposures that belong on the enterprise risk register.
Action needed
Circulate to the CFO, Chief Sustainability Officer, and Chief Risk Officer with a request to map the organisation's AI compute footprint against energy-supply dependencies and flag any indirect nuclear-vendor exposure in the supply chain.