China’s AI Power Plan: Nuclear Build-Out Now, SMRs Next, Fusion Later

China is responding to the prospective electricity demands of AI with a portfolio rather than a single technology: a large conventional-nuclear construction program for firm power, small modular reactor (SMR) development for potential future flexibility, and fusion investment aimed at a far more distant commercial opportunity. The distinction matters. Wood Mackenzie projects China’s data centers could consume 774 terawatt-hours (TWh) in 2030—roughly four times current use and about 6% of national electricity demand—but fusion will not be a material source of power on that timetable. 55

The near-term answer is conventional nuclear

China already has 64 operable nuclear reactors with 63,985 MWe of capacity, and 38 reactors totaling 39,694 MWe under construction. 17 That pipeline is the deployable nuclear component of the response to rising power demand from computing and other sectors.

Large reactors offer dispatchable, low-emissions generation, but they are not a stand-alone answer for data centers. Meeting rapid load growth also depends on the wider power system: grid build-out, renewable generation, storage, efficiency and other generation sources. The International Energy Agency expects renewables to meet nearly half of global incremental data-center electricity demand through 2030, with nuclear becoming more important later in the decade and beyond. 57

China is also advancing the Hainan Changjiang Multi-Purpose SMR Technical Demonstration Project. 18 SMRs could eventually offer more flexible siting and standardized production than large plants, but their economics at commercial scale remain unsettled. The IEA expects SMRs to enter the data-center supply mix after 2030, not as a major solution to immediate AI demand. 57

Junhe Atomic: a small-reactor developer with limited public detail

“Junhe Nuclear” appears to refer to Junhe Atomic (JHATOM), a privately owned Chinese nuclear-technology company founded in 2025. The company says it develops and seeks to commercially deploy SMR and microreactor systems. 19

Reporting cited in the available material says the Shanghai-based company has attracted 300 million yuan in venture investment, is in the design phase, and is targeting a demonstration project around 2031. 5036 These are early-stage plans, however—not evidence of a financed, licensed, under-construction reactor or a contracted data-center power project.

For prospective customers and investors, the key uncertainty is execution: the available sources do not establish a specific reactor design, construction start, grid-connection date, or commercial offtake agreement. Junhe therefore illustrates China’s interest in advanced fission, but should not yet be treated as a confirmed near-term provider of AI power.

NovaFusionX: well-funded, but still pre-commercial fusion

Shanghai-based NovaFusionX was founded in 2025 and is developing compact, modular fusion systems based on a field-reversed configuration (FRC). The company is positioning the technology for distributed applications, including data centers. 3345

Its reported funding ramp has been rapid. In April 2026, NovaFusionX said an Angel+ round of 700 million yuan brought cumulative financing to 1.2 billion yuan (about $175.6 million) within a year. 42 By August, it reported another 1.2 billion yuan pre-Series A financing, bringing total funding to 2.4 billion yuan and valuing the company at more than 10 billion yuan post-money, according to reporting on the transaction. 34

The important timeline is more modest than the financing headlines: NovaFusionX is targeting first plasma for its Nova One machine by the end of 2026. 38 First plasma demonstrates an early operational capability in an experimental device. It does not demonstrate net electricity generation, a commercial reactor, regulatory approval, or dependable data-center supply.

The company’s founder has argued that AI-driven demand requires fusion developers to compress commercialization timelines to five to 10 years, rather than 30 to 50 years. 37 That is an ambition, not an independently validated deployment schedule. Fusion has not yet produced commercial electricity anywhere.

China and the U.S. are funding fusion differently

Estimates of China’s fusion support vary because public and quasi-public funding is difficult to track. A U.S. bipartisan commission estimate puts Chinese investment since 2023 at at least $6.5 billion, almost three times the U.S. Department of Energy’s Fusion Energy Sciences funding over the comparable period; other assessments place the upper range near $13 billion. 49 These figures should be read as estimates, not audited totals.

The United States remains ahead in private fusion investment. One industry tally reported about $8.05 billion across 42 U.S. companies, compared with roughly $5.14 billion across eight Chinese firms. 8 The models differ:

• China: coordinated public, provincial, university, industrial and private-capital participation, alongside a large domestic nuclear-industrial base. 67
• United States: deeper venture-backed company formation, supplemented by federal science programs and increasingly commercialization-focused support. 8

U.S. federal funding is also increasing. DOE’s Fusion Energy Sciences program received $790 million in FY2025, while ARPA-E committed $135 million over 18 months in 2026 for fusion commercialization projects. 53

Neither funding model removes fusion’s core technical and commercial risks. High investment does not establish that any particular approach will reach reliable, economic power generation first.

Why China’s nuclear industry could gain an advantage

China’s potential edge is not proof that it has solved fusion. It is its ability to link research, manufacturing and repeat construction around a sizable home market.

Repeat construction creates industrial learning

With 38 reactors under construction alongside an operating fleet of 64, China has a sustained project pipeline for reactor vendors, component manufacturers, construction teams, operators and regulators. 17 Repetition can support standardization, workforce retention and procurement scale—attributes that can reduce execution risk over successive projects.

Coordinated capital and institutions can speed prototypes

Analysts describe China’s fusion push as combining central and provincial funding, universities and large industrial players. 67 That structure could make it easier to move from laboratory systems to prototypes and manufacturing capacity, although it does not guarantee superior reactor performance or lower costs.

A domestic platform could support exports—but conditions apply

A mature domestic construction base could help Chinese suppliers compete internationally if they can offer standardized technology, credible delivery schedules and financing. Yet export success would still be constrained by nuclear safety requirements, nonproliferation rules, geopolitics, financing terms, fuel-cycle arrangements and the economics of each project.

The bottom line

The 774 TWh forecast makes power availability a strategic issue for China’s AI expansion. 55 In the 2030 timeframe, the consequential nuclear story is the continuing deployment of conventional reactors, with SMRs as a possible later complement. Junhe Atomic is an early-stage advanced-fission company with a reported 2031 demonstration goal, while NovaFusionX has raised substantial capital and targets first plasma in 2026—but neither has demonstrated a bankable route to supplying data centers. 193638

Fusion is strategically significant in the U.S.–China technology contest, but it remains a long-duration wager. China’s strongest potential advantage lies in converting proven nuclear technologies into repeatable industrial projects while using its public and private funding ecosystem to pursue advanced fission and fusion in parallel.