Yunqi Capital Co-Leads Round in Feijiekesi: NVIDIA PhysX Co-Founder Rebuilds Foundational Physics AI Engine | Yunqi Portfolio

Getting models to truly see and understand the world is becoming the critical breakthrough point for the evolution of physical AI.

Recently, Yunqi Capital co-led the Pre-A3 funding round for physical AI infrastructure company Feijie Kesi (飞捷科思). Feijie Kesi builds a full-stack technical foundation spanning physical simulation, data generation, model training, quantitative evaluation, and industrial deployment — all centered on a differentiable physics simulation engine — exploring the infrastructure opportunity as AI advances into the real physical world. Learn more in this edition of "Yunqi Partner."

Yunqi Capital's Investment Thesis:

Physical AI will serve as a critical underlying layer supporting the continued development of embodied intelligence, robotics, and industrial intelligence. As AI applications move toward deeper task execution, high-precision, low-cost, trainable, and transferable simulation environments will directly impact model training efficiency and real-world deployment capabilities.

Feijie Kesi's entry point through a differentiable physics engine offers strong underlying technical scarcity, platform potential, and global competitiveness. We look forward to the company's continued breakthroughs in physical AI infrastructure, enabling more intelligent systems to enter the real world.

The following is excerpted from Feijie Kesi:

This round was co-led by Yunqi Capital and Zhongxin Zhidao Capital, with continued follow-on investment from Oriental Fortune Capital, Silicon Harbor Capital, and others. Additional participants included Taida Fund, Changlei Investment, Changshi Investment, Pangu Ventures, and Zhongying Venture Capital. Combined with the previous two rounds, the Pre-A round is now officially complete, with total funding reaching several hundred million RMB.

The concentrated backing from both new and existing investors fully validates the capital market's strong recognition of Feijie Kesi's breakthroughs in underlying physical AI technology, commercialization potential, and long-term growth value.

Built on its core founding team, Feijie Kesi is rooted in tackling fundamental physical AI technology challenges. The company was founded by Professor Lihua Zhang, a principal architect of NVIDIA's PhysX physics engine, giving it world-class physics engine R&D experience and technical insight.

Earlier, on March 27, Feijie Kesi officially released Fysics, China's first fully self-developed differentiable physics simulation engine. Together with the one-stop simulation training platform MoziSim, the omnimodal physical AI foundation model OmniFysics, and a bidirectional standardized evaluation benchmark system, these form a complete physical AI full-stack technical foundation, fully closing the loop across "physical simulation, data generation, model training, quantitative evaluation, and industrial deployment."

Proceeds from this round will focus on core technology iteration and industrial ecosystem building. The company will prioritize upgrading Fysics engine's high-fidelity solving capabilities, further improving simulation accuracy and computational efficiency in complex scenarios and dynamic physical environments; and continuing to iterate the OmniFysics omnimodal physical AI foundation model, strengthening the model's generalization across scenarios and devices.

The next-generation physics engine Fysics is both the core lever for advancing physical AI and the critical bridge connecting virtual and physical worlds, as well as the underlying core infrastructure supporting breakthrough deployments in embodied intelligence and robotics. Feijie Kesi focuses on solving current industry pain points and challenges, driving industry transformation through technical breakthroughs and empowering industrial upgrading.

We look forward to Feijie Kesi becoming the "new operating system" for the global physical AI field, achieving large-scale substitution, and propelling Chinese technology in physical AI core infrastructure from following to leading!