Yunqi Capital backs NovaFusion seed-plus round, using AI to reshape fusion R&D paradigm

From "build-and-test" to "digital-twin optimization," AI is rewriting the R&D paradigm for controlled nuclear fusion. Finding the optimal fusion technology path at lower cost and in less time is the critical question for the industry's march toward commercial scale.

VeloAlpha Technology, founded just four months ago, is answering that question with an AI-native fusion digital-twin platform: its VEQ module achieves roughly 20,000× faster equilibrium configuration solving than the international mainstream code CHEASE, and within two weeks of release it was absorbed and integrated into the FUSE integrated modeling platform by General Atomics of the United States.

Recently, VeloAlpha completed its second round (seed+ round) of financing, raising tens of millions of RMB, with participation from the Yunqi-SJTU AI Angel Fund among others. This follows Yunqi Capital's earlier bet on NovaFusion, marking another move by the firm in the controlled nuclear fusion track. Learn more with Yunqi Partners.

Yunqi Capital's View:

We have long tracked how AI is reconstructing the paradigm of scientific R&D. Controlled nuclear fusion is accelerating from engineering experiments toward commercialization. The traditional "build-and-test" model carries high costs and long cycles, but the deep integration of core mathematical algorithms, engineering simulation, and AI capabilities could compress the trial-and-error cost and iteration cycle of fusion R&D by several orders of magnitude — this is the key variable for the fusion industry's path to scaled commercialization.

The VeloAlpha team has spent over a decade building expertise in multi-route fusion reactor design, simulator development, and AI module development, combining full-stack technical insight with exceptional execution. Multiple modules have achieved globally leading breakthroughs and gained adoption by mainstream international modeling platforms. We look forward to VeloAlpha accelerating technology-path optimization through its fusion digital-twin platform, and bringing fusion energy closer sooner.

The following is adapted from VeloAlpha Technology

Recently, VeloAlpha completed its second round (seed+ round) of financing, raising tens of millions of RMB. In early June, the company had just closed its first round, also tens of millions of RMB. The rapid fundraising pace reflects quick progress on product and commercialization fronts. In its four months since founding, VeloAlpha has expanded to over 30 people, densely released multiple core products including VEQ, VSC, BORAY-3D, and its integrated platform, and secured multiple orders at the hundreds-of-thousands-of-RMB level.

About VeloAlpha:

Reconstructing the Fusion R&D Paradigm with AI

Beijing VeloAlpha Technology Co., Ltd. (VeloAlpha Technology) was founded in April 2026, focusing on controlled nuclear fusion digital-twin platform R&D and fusion-path optimization services. Through first-principles calculation, engineering simulation optimization, AI algorithms, and agent-based workflows, the company provides digital and intelligent solutions for fusion device design and path optimization, compressing the trial-and-error cost and iteration cycle of fusion R&D by several orders of magnitude. The core modules of the company's flagship product FusionAlpha have demonstrated generational speedups over international mainstream codes at equivalent precision in multiple benchmark tests, with the verification cost for a single fusion technology path reducible to the hundreds-of-RMB level.

Founder Dr. Huasheng Xie has nearly two decades of R&D experience in fusion, and is a well-known fusion expert, leading a team with over ten years of accumulated expertise in multi-route fusion reactor design, simulator development, and AI module development. The core team comes from Peking University, Tsinghua University, Zhejiang University, the Chinese Academy of Sciences, and other universities and research institutions, covering plasma physics, computational mathematics, and underlying software architecture.

International Recognition

VEQ Module Adopted by International Platform Within Two Weeks of Release

Among its released product lineup, the VEQ module has drawn industry attention. The module achieves roughly 20,000× faster equilibrium configuration solving than the international mainstream code CHEASE, and within two weeks of release was absorbed and integrated into the FUSE integrated modeling platform by General Atomics of the United States, replacing the original equilibrium solver module. International mainstream modeling platforms including ASTRA and OMFIT are currently also adapting the code.

The company's open-sourced algorithm version has penetration approaching 100% among domestic fusion R&D host manufacturers and research institutions, with over 50 global institutions already using its core code. As FUSE, OMFIT, ASTRA, and other integrated platforms embed VeloAlpha's product modules, VeloAlpha Technology's global technology penetration will further expand.

AI-Native Team + Deep Industry Knowledge

Rapidly Producing Internationally Leading Results at Scale

The core challenge in controlled nuclear fusion simulation lies in modeling the macroscopic and microscopic characteristics of plasma. The company revealed that it has conquered the major key modules involved, with releases expected in the near term. By year-end, VeloAlpha Technology plans to roll out multiple important advances across the full module stack of its fusion digital-twin platform, establishing systematic generational-leading technical advantages globally.

At VeloAlpha Technology, the team focuses on solving the speed and extrapolation problems of underlying simulators, and using AI to comprehensively and deeply transform the R&D paradigm. The team has already established an AI-native R&D workflow, which — combined with senior experts' deep knowledge of the industry and technology and full-stack fusion technical understanding — can now support undergraduates in producing internationally leading results within two months. The team has developed the capability to rapidly produce high-level, high-quality fusion digital-twin platform R&D, with overall progress running more than twice as fast as the development plan set in April.

Use of Funds

AI-Native Team Expansion and Overseas Market Development

Following this round of financing, VeloAlpha Technology will continue building its AI-native high-level R&D team, recruiting and cultivating top talent globally, expanding its technical and knowledge advantages, while simultaneously launching overseas market expansion.

Global controlled nuclear fusion is currently in an accelerated commercialization phase, with over 190 fusion devices under construction or in operation, and more than 8 device routes and 4 fuel routes advancing in parallel. The traditional build-and-test model faces challenges of high cost and long cycles: single experimental reactor R&D investment typically exceeds $2 billion, with construction cycles of 3 to 10 years, and over 60% of large-scale fusion devices globally have failed to meet design targets. Accelerating fusion-path optimization through a digital-twin platform is precisely the answer VeloAlpha Technology offers.

VeloAlpha — Solving the essence of physics through ultimate algorithms.