AI power demand accelerates China’s nuclear and fusion build-out

China is constructing more than half of the world's 58 new conventional nuclear reactors while supporting private companies in small modular reactors and fusion. The immediate industrial pressure is electricity: Chinese data-centre consumption is forecast to reach 774TWh by 2030, making dependable generation part of the country's ability to expand AI computing.

State-linked startups illustrate the range of the bet. Junhe Atomic is developing small modular reactors. NovaFusionX, founded by former US-based researcher Guo Houyang, is pursuing a field-reversed-configuration fusion system and targets net energy gain by 2029. That date is a company objective, not a demonstrated result; no commercial fusion plant anywhere produces sustained net electricity after accounting for the full facility.

China's advantage is coordination. Nuclear construction, grid planning, materials suppliers, university research and state finance can be directed toward a long programme. Fusion now appears in the five-year plan, and estimates cited by the FT suggest Chinese commercial investment may exceed US levels. The same system can also obscure weak projects and prioritize capacity targets over economic evidence.

Conventional reactors can contribute during the current AI build-out; fusion almost certainly cannot do so at comparable scale on the same timetable. The strategic logic is therefore layered: use established nuclear technology to reduce near-term dependence on imported fuel and volatile fossil prices, while treating fusion as a long-duration option whose scientific and engineering risks remain very large.