On 11 August, Daimon Robotics announced a several-hundred-million-yuan strategic round led by Ant Group, with existing shareholders oversubscribing. The raise came only about two months after its previous financing: a 100-million-yuan A round in June led by Inovance Industrial Investment and China Telecom.
The string of capital injections has built Daimon an investor roster spanning China Merchants Venture, Lenovo Capital, Inovance, China Mobile and China Telecom, industrial players rather than pure funds. What they appear to value is not another humanoid body, but the way embodied intelligence is extending upstream into perception, data and model infrastructure.
Touch becomes a variable
Mainstream VLA models lean on vision and language. But once a robot enters contact-intensive work, grasping, assembly, flexible-object handling, seeing alone does not guarantee reliable operation. A T-Rex study released in June designed a new architecture for high-frequency tactile signals and beat the strongest baseline by more than 30 per cent on average across 12 tasks needing fine force control and flexible-object manipulation. That suggests touch is not one more sensor, but a redesign of data, architecture and control.
Daimon is pushing further up the model stack. Its newly launched Daimon-TWM, a tactile-anchored world model, threads tactile information through physical-state understanding, action prediction and real-time control, and uses slow planning plus fast correction to improve how robots handle contact tasks.
The company spun out of a Hong Kong University of Science and Technology team and has operated since 2023, already building vision-tactile sensors, teleoperation data capture and a VTLA model, trying to close the sensor-data-model-deployment chain. Its reported Daimon-TWM performance, data scale and customer coverage are largely self-reported, and the real test remains whether the tactile gain holds across different robot bodies and industrial scenes, and whether it covers the cost of sensors, capture and deployment.
Seen this way, Ant Group’s entry looks like an early ticket. As embodied intelligence moves from can walk, can see to can work, whether touch becomes standard equipment or only pays off in high-precision niches is a question only scale can answer.
Editor’s note: This is an adapted translation of the original Gasgoo report. It has been trimmed and restructured for readability for an international business audience.