Physical AI Startup Quantum Dynamics Raises Over $100M Seed

Quantum Dynamics, a new venture focused on Physical AI, has successfully closed a seed funding round exceeding 100 million yuan (approximately $13.8 million USD). The round was co-led by prominent investors Yunqi Capital and SenseTime, signaling strong confidence in the company's mission to bridge artificial intelligence with the tangible world. Funds will be directed towards advancing core Physical AI technologies, building a robust talent pipeline, and expanding its global market reach, aiming to accelerate the development of intelligent systems for real-world applications.

Founded in early 2026 by Li Qiang, a veteran of Alibaba Group with extensive experience in logistics and international digital commerce, Quantum Dynamics is charting a distinct course in the rapidly evolving embodied intelligence sector. Unlike founders emerging from autonomous driving or pure AI research backgrounds, Li Qiang brings a unique blend of deep operational understanding from his 17-year tenure at Alibaba, where he held leadership roles at Cainiao, Taobao Tmall, and international business units. His experience includes scaling systems from inception and managing large, global R&D teams, most notably as the former CTO of Cainiao Group and CTO of Alibaba International Digital Commerce.

The company's strategic differentiator lies in its pragmatic approach: prioritizing the conversion of existing AI capabilities into tangible customer value rather than waiting for a perfect general-purpose model. This "scenario-first, foundation-later" strategy is particularly relevant as the embodied intelligence field shifts from theoretical demonstrations to practical deployment. Industry observers note that the success of Physical AI hinges increasingly on the depth of real-world data, a challenge amplified by the inability to "scrape" physical interactions like digital text or images. Quantum Dynamics aims to capitalize on this by embedding its systems directly into operational environments.

Quantum Dynamics' initial focus is on the logistics and warehousing sector, a domain Li Qiang knows intimately. The company is building a data feedback loop by deploying its AI systems in real warehouses. Each robot's daily operations, encompassing tens of thousands of pick, place, scan, and label actions, generate invaluable, implicitly labeled data on successful and failed interactions across diverse physical objects and conditions. This approach directly addresses the critical need for large-scale, real-world data, which is essential for training robust world-action models capable of understanding and manipulating the physical environment. This strategy is crucial in a market where IDC projects China's embodied intelligence robot spending to surge from over $1.4 billion in 2025 to $77 billion by 2030, with a staggering 94% CAGR.

The core technology platform developed by Quantum Dynamics is designed for continuous evolution from real-world deployments. It comprises six integrated modules, covering data aggregation, model updates, and cluster-based collaborative intelligence. Key innovations include a "World-Action Model" that learns how actions alter the physical world by integrating visual, motion, proprioception, and tactile signals, and a "World Model Corrector" for real-time error detection and replanning. The system also incorporates Test-Time Training and few-shot adaptation techniques to rapidly adjust to new environments and tasks without full retraining, enabling a single operator to manage an increasing number of robots.

Looking ahead, Quantum Dynamics plans a phased rollout, starting with B2C and SME e-commerce warehouses in 2026, targeting high-volume tasks like replenishment, picking, and packing in environments with standardized SKUs. By 2027, the company intends to expand into more complex logistics hubs, industrial production lines, retail shelves, and eventually commercial services and home assistance. This progressive expansion strategy aims to build a versatile Physical AI platform that can adapt to a wide array of real-world challenges, demonstrating the viability of a vertical-to-general approach in the physical AI domain.