PHYMI, an embodied artificial intelligence startup founded earlier in 2026, has raised close to $100 Million in seed financing. The round was led by IDG Capital and included Yunqi Partners, Fosun RZ Capital, Glory Ventures, Didi Global, and Hesai Group. Led by former DeepRoute.ai executive Liu Nianqiu, the company announced the funding on Wednesday as it works to develop Physical Agents for open and dynamic real-world environments.
A New Direction for Embodied Intelligence
While large multimodal models and agentic AI have performed strongly in screen-based digital environments, PHYMI is focused on closing the gap between software intelligence and the physical world. The startup is developing Physical Agents that combine cognitive reasoning with physical actions to perceive, navigate, and manipulate in unstructured settings. Its ambition is to move artificial intelligence from providing answers toward achieving complex real-world objectives.
From Executing Actions to Achieving Goals
PHYMI does not define capability by single-point actions at fixed workstations, in single scenarios, or within preset workflows. Instead, it targets long-horizon movement and manipulation across spaces where environments and human needs can change at any moment. This requires a Physical Agent to understand goals, decompose tasks, plan dynamically, and adjust its behavior as new conditions emerge.
Leadership and Technical Foundation
Founder and CEO Liu Nianqiu spent more than seven years at DeepRoute.ai as a vice president and partner. During that time, he contributed to the deployment of map-free autonomous driving solutions and supported the company's technical shift toward end-to-end and Vision-Language-Action architectures. Earlier in his career, Liu served as a senior engineer at Intel and launched ventures in the smart hardware industry.
A Hybrid Team with Production Depth
PHYMI's core members hail from institutions including Tsinghua University, Shanghai Jiao Tong University, Fudan University, Nanyang Technological University, and The Chinese University of Hong Kong. The team has led work on large models, VLA systems, real-robot reinforcement learning, robot platforms, data infrastructure, and mass-production delivery. This mix of frontier research and engineering execution is designed to carry the Physical Agent from laboratory prototypes into real-world operation.
Full-Stack Physical Agent System
PHYMI has built an integrated stack that aligns models, hardware, data, and interaction under one architecture. PHYMI BRAIN covers multimodal perception, world modeling, reasoning, planning, and action generation, while PHYMI BODY provides the physical platform for stable real-world operation. PHYMI DATA turns every interaction into reusable experience and PHYMI LINK supports collaboration among people, agents, and environments.
Investor Confidence and Market Context
Yunqi Partners characterized the Physical Agent as a long-term direction as artificial intelligence shifts from answering questions to achieving goals in the physical world. The firm believes PHYMI combines frontier model vision, complex software and hardware engineering, and the operational experience required for large-scale delivery. The round drew a varied group of financial investors, industrial capital, technology companies, and entrepreneurs.
Funding Allocation and Next Steps
PHYMI will use the capital to accelerate research and development for its core Physical Agent offerings and to build real-world dynamic data systems. The company also plans to advance product engineering, expand scenario deployment, and grow its team. These steps are intended to help the startup create embodied AI products that are reliable, practical, and affordable for everyday use.
PHYMI's seed financing underscores growing investor interest in embodied AI that can operate beyond controlled digital environments. Its focus on long-horizon tasks and real-world goal completion places the company at the center of the next phase of intelligent systems. If PHYMI delivers on its full-stack vision, it may help transform advanced artificial intelligence from a screen-based tool into a capable physical partner.