China’s Nuclear Power Push for AI: Reactors Now, SMRs Next, Fusion Later

China’s nuclear strategy for rising data-center demand has three distinct timelines. Reactors already operating or under construction offer the most immediate potential contribution to the grid. A startup’s small modular fission reactor is targeted for the next decade, while fusion investment is funding experiments rather than dependable electricity supply.4719

How large could data-center demand become?

Wood Mackenzie forecasts that electricity use by all Chinese data centers could roughly quadruple to 774 terawatt-hours (TWh) by 2030, or about 6% of national consumption. AI is a major reason for the attention on power supply, but the forecast covers data centers generally—not AI workloads alone—and remains a projection, not a guaranteed outcome.14 Expanding low-carbon generation also fits China’s stated goal of carbon neutrality by 2060.1

Conventional reactors are the near-term option

China is reported to have 64 operable nuclear reactors and 38 under construction, with the latter representing nearly 40 gigawatts of additional capacity.4 This existing fission build-out is more relevant to electricity demand approaching 2030 than proposed fusion plants. It does not mean the new capacity will be dedicated to data centers: those facilities draw on a wider electricity system.414

Junhe Nuclear: a modular fission demonstration for 2031

Shanghai-based Junhe Nuclear has reportedly attracted 300 million yuan in venture funding and is designing its first small modular reactor (SMR) demonstration project. It aims to have that project operating around 2031 and has discussed reusing former coal-plant sites and workers.717 An SMR is a fission reactor; it should not be confused with the small, distributed fusion machines proposed by NovaFusionX. Junhe’s date is a development target, not evidence that a plant is complete or supplying power.717

NovaFusionX: substantial funding, earlier scientific milestones

Fusion startup NovaFusionX has reportedly raised 2.4 billion yuan in total: a 500 million yuan angel round, a 700 million yuan Angel+ round and 1.2 billion yuan across pre-Series A financing rounds.18 It is developing a field-reversed-configuration approach and targets first plasma by the end of 2026 and experimental net energy gain by 2029.1819 Its founder has described an ambition to shorten fusion’s commercialization window to five to ten years.19

Those milestones are not interchangeable with a working power plant. First plasma would show that the device has produced plasma; an experimental net-gain result would still leave the challenge of producing sustained, usable electricity. The company’s demonstration-power plans lie in the 2030s, so fusion should not be treated as a firm contribution to the 2030 demand forecast.1219

How China and the US compare

China’s advantage may be its ability to combine reactor construction, domestic manufacturing and coordinated public investment. In one reported Chinese nuclear project, domestic suppliers provided more than 90% of the equipment.50 For fusion specifically, a CSIS analysis estimates China mobilized at least $6.5 billion in public funds from 2023 through 2025. That is an estimate of public funding, not proof that commercial fusion is close.33

The US has a substantial private fusion sector, including investment linked to technology companies, alongside federal research and milestone-based support.2732 Its government roadmap targets early fusion-plant deployment in the mid-2030s, while identifying materials, infrastructure, supply-chain and workforce gaps that still need to be addressed.28 Public-spending comparisons alone therefore do not settle which country will commercialize fusion first.

The practical distinction is timing: China’s conventional fission fleet can contribute to the grid now and as projects are completed; Junhe is pursuing a 2031 demonstration; and NovaFusionX is still working toward experimental fusion milestones. Rising AI demand helps explain the investment in all three, but it does not make their electricity available on the same schedule.4719