The archive · Climate & Energy · Strategic decision · 2025–2026
Nova Fusion bets compact FRC fusion beats tokamaks for AI power; ¥2.4B in 16 months
Shanghai startup Nova Fusion raised ¥2.4B across three rounds in 16 months betting FRC-SMR small modular fusion can power AI data centers before tokamaks can.
Nova Fusion (诺瓦聚变)
What the business is
A Chinese private fusion startup building FRC-SMR (field-reversed configuration, small modular) reactors as distributed power plants, starting with its first device, Nova-1, aimed at AI data center demand.
Starting capital：¥2.4B (~$340M) total: ¥500M angel (Aug 2025, record for Chinese private fusion), ¥700M angel+ (Apr 2026), ¥1.2B Pre-A series (Aug 2026)
How it started
Guo Houyang, a U.S. Physical Society fellow and former chief scientist at the world's largest FRC company TAE, founded Nova Fusion in Shanghai in April 2025. He decided the time was right because AI's power hunger had compressed fusion's commercial window from about 30 years to 5–10, and Helion had already proven FRC's engineering path.
What happened
Nova-1 finished engineering design and core-component development and moved into final assembly. The company targets first plasma discharge by end-2026, China's first private 100 million °C fusion temperature in 2027, and Q>1 by 2029; investors benchmarked it against Helion Energy, which has raised over $2.1B at a $15.5B valuation.
How it ended up
Still running: Nova-1 is in assembly, with first plasma targeted for end-2026 and a 50–100MW demo plant planned for the early 2030s.
Background
Nova Fusion is a Shanghai-based private fusion startup founded in April 2025 by Guo Houyang, a U.S. Physical Society fellow who led the U.S. Department of Energy's largest FRC research program and served as chief scientist at TAE, the world's biggest FRC fusion company. It is China's first company focused on FRC-SMR: field-reversed configuration, small modular reactors.
The bet is that compact magnetic-compression reactors can reach net energy gain (Q>1) years before gigawatt-scale tokamaks, at a fraction of the cost — roughly 1/100, with a target power cost of ¥0.07/kWh, about a tenth of China's current industrial rate. The first customer is meant to be AI data centers, whose 50MW-scale demand and 5–10 year power window shaped the company's roadmap.
In 16 months the startup completed three rounds totaling ¥2.4B — ¥500M angel (August 2025), ¥700M angel+ (April 2026, with Alibaba continuing), and a ¥1.2B Pre-A series (August 2026, with Meituan Longzhu, Hillhouse, Lenovo Capital and Guozhong Capital) — reaching a post-money valuation above ¥10B and setting three records for Chinese private fusion. Nova-1's four core subsystems are developed and entering mass production; first plasma discharge is targeted for end-2026.
What has to be true
- AI data centers created a 5–10 year power demand window that favors small, distributed reactors over multi-decade tokamak megaprojects.
- FRC's magnetic-compression design is structurally simpler and far cheaper, giving a capital-constrained startup a plausible path to commercial power.
- Guo Houyang's rare full-stack FRC experience from TAE and U.S. DOE programs made the team credible to top-tier investors.
- China's industrial supply chains let the company develop every core subsystem domestically, including record-size quartz vacuum chambers.
- The company deliberately front-loaded fundraising in parallel with engineering milestones, so it could start Nova-2 before Nova-1 even produced plasma.
What can be applied
When a market window compresses, a cheaper, faster technical route can outrun the incumbent one — but record funding only buys time until the device must actually produce plasma.
Aftermath
As of September 2026, Nova Fusion has 60+ engineers and all four core subsystems — solid-state pulsed power, modular high-field magnets, large quartz vacuum chambers, and a nanosecond control/acquisition system — developed in-house and entering mass production. Nova-1 is in final assembly, with first plasma targeted for end-2026 and China's first private 100 million °C fusion temperature in 2027. The plan is to start Nova-2 ahead of Q>1 and upgrade it into a 50MW commercial demo plant around 2030, selling power first to AI data centers.
Sources
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card