Ant Group leads strategic round in Daimon Robotics, first tactile bet as startup unveils world-first ‘physical interaction brain’

Daimon Robotics said on August 11 that it has closed a strategic funding round worth hundreds of millions of yuan, led by Ant Group, with existing shareholders following on. It is Ant's latest heavyweight bet in the embodied AI arena — and the first time the company has extended its investment reach into the tactile sensing layer.

Ant has previously placed dense bets across embodied intelligence, backing Unitree, Xinghait, Lingxin Qiaoshou, Sudu Tech and Wuji Tech, spanning robot bodies, embodied brains and core components. Daimon, a representative company in embodied tactile sensing, completed a 100-million-yuan Series A only two months ago, with backers including China Merchants Group's CVC, Lenovo Capital, Inovance, China Mobile and China Telecom.

The company's technical lineage runs back to Carnegie Mellon University. Matthew T. Mason, known as the “father of dexterous robotic manipulation,” argued that a hand's dexterity is mostly about the brain, not the hand. Wang Yu, Mason's first PhD graduate, brought that school of thought back to China, co-founding HKUST's Robotics Institute with Li Zexiang in 2015 and helping Daimon start scaled operations in Shenzhen in late 2023.

Wang serves as co-founder and chief scientist. CEO Duan Jianghua leads industrialization, chief AI scientist Yuan Weihao brings both large-model and physical-robotics experience, and Du Yipai, the architect of Daimon's monochromatic optical tactile sensor, heads R&D and tactile engineering.

On the product side, Daimon recently launched Daimon-TWM (Tactile-grounded World Model), which it calls the world's first tactile-anchored world model and “physical interaction brain.” Native tactile signals run through understanding, reasoning, prediction and verification, forming a “tactile nervous system” that spans fingertip perception, brain-level reasoning and action control. The 10B-parameter model runs real-time inference on an NVIDIA RTX 5090 and is designed to deploy across robot bodies and arms.

Daimon-TWM stacks three layers — a physics-cognition module that builds physical common sense, a reasoning-and-decision module that predicts how contact states evolve, and an instant-control module that corrects actions at 100Hz. In physics-cognition benchmarks, average success rates on contact-intensive tasks roughly double those of π0.5 without disturbances and improve tenfold under disturbances; in a broken-glass cleanup demo, the model adjusts grip angle and force in real time based on tactile feedback.

On data, Daimon has built what it calls the world's largest tactile-inclusive physical-world dataset, Daimon-Infinity; the first open-sourced 10,000 hours have drawn nearly 5 million downloads on Alibaba's ModelScope. With China Mobile, it is building an outbound data-collection network, including the world's first “embodied data collection 5S store” in Chenzhou, Hunan, with 1,000 devices planned in the first phase.

Commercially, Daimon claims first place globally on eight key metrics, has served more than 200 customers worldwide (over 50 overseas), and has delivered to OpenAI, Figure, Physical Intelligence, Skild AI, Meta, BMW and Google DeepMind.

The deal underscores a broader shift: capital is flowing from robot bodies toward tactile perception. As players race to make dexterity scale, the open question is whether Daimon's data expansion to millions of hours by year-end and its cross-body deployment plans can turn the “physical interaction brain” from demo into mass production.