On October 8, the strategic investment division of Ant Group disclosed that Liqing Intelligence, a Physical AI Infra platform company, had recently completed several hundred million yuan in angel round and angel-plus round financing.

Ant Group increased its stake across two consecutive rounds and led the angel-plus round, while CMC Capital, Guoqi Investment, funds under CICC Capital, and the Shanghai Artificial Intelligence Industry Series Fund participated as follow-on investors, with existing shareholders including Shunwei Capital, Century Golden Resources, Frees Fund, and the Tsinghua Alumni Seed Fund continuing to add investment.

The financing will be directed toward physical-world data pipelines, model research and development, physical simulation, capability evaluation, and cross-embodiment adaptation and deployment systems.

What Ant Group is investing in is not another robot hardware company, but the data and training segments needed before embodied intelligence can enter real-world scenarios.

Liqing Intelligence was founded in April 2026, with founder Li Yiming serving as an assistant professor at Tsinghua University's Institute for AI.

The company positions itself as a Physical AI Infra provider, seeking to offer tools such as data collection, model training, physical simulation, capability evaluation, and deployment feedback for models entering the physical world.

For robots to learn grasping, carrying, or assembly, they cannot rely solely on text and images already available on the internet.

They also need to perceive the depth of objects, the posture of hands, the forces experienced during contact, and environmental changes; collecting such real-task data is costly and difficult to cover low-frequency and complex scenarios.

Liqing Intelligence's current entry point is data collection and simulation. The company has launched its T1 series data-collection gloves and Ego data-collection devices to record multimodal information from real tasks; it then uses its self-developed physics engine to let models repeatedly trial and error in simulated environments, aligning them with a small amount of real-machine data.

Data collection is only the starting point. Liqing Intelligence also provides task design, data production, processing and quality inspection, and capability validation services to model manufacturers, and attempts to combine tasks, models, devices, and deployment systems for delivery to end customers.

This places it between model research and robot deployment. What it needs to prove is not whether it can produce a batch of data, but whether these data, simulation results, and adaptation experience can be reused across more models, more embodiments, and more projects.

Currently, Liqing Intelligence has not disclosed metrics such as customer scale, revenue, cross-project reuse ratio, or changes in deployment costs.

Whether the "infrastructure" positioning can hold still depends on whether these segments can precipitate from single-project services into replicable platform capabilities.

Ant Group's two consecutive rounds of increased investment have also brought this early-stage company into its embodied intelligence investment map.

Previously, Ant Group had invested in companies including Galaxea, Stardust Intelligence, Unitree Robotics, Titan Tiger Robot, and Lingxin Qiaoshou, covering directions such as robot hardware, joint modules, and dexterous hands.

For example, Ant Group exclusively led Galaxea's nearly 300 million yuan Series A financing; in June 2025, Ant Group participated in Unitree Robotics' financing and, through a wholly owned entity, took stakes in Titan Tiger Robot and Lingxin Qiaoshou.

These investments point more toward robots' "bodies" and "senses."

At the same time, Ant Group is also developing its own "robot brain." Its subsidiary Ant Lingbo has launched models and products for service robots and is exploring collaborative operations of multi-brand, different-configuration robots in retail pharmacy scenarios.

Ant Group disclosed that Lingbo's embodied models have been adapted to more than 20 robot configurations from 17 manufacturers.

Liqing Intelligence, meanwhile, fills in the data and simulation layer of this chain.

Data-collection devices are responsible for preserving real tasks, physical simulation is used to expand trial-and-error iterations, and cross-embodiment adaptation determines whether a set of capabilities can migrate to robots of different forms.

Therefore, Ant Group's investment logic is beginning to show clearer layers: self-developed models and robot "brains," investments in hardware, joints, and dexterous hands, and then an extension toward data and simulation infrastructure.

The significance of this investment lies not in the fact that Ant Group has already built a complete embodied intelligence system, but in that its external investments are beginning to touch the intermediate links from robot training to deployment.

An investment relationship does not equal business synergy. Public information has not yet disclosed whether Ant Lingbo and Liqing Intelligence have joint development, data procurement, or scenario cooperation.

Embodied intelligence still faces issues such as expensive real-machine data, limited model generalization, and complex adaptation across different embodiments.

Whether Liqing Intelligence can transform data, simulation, and deployment systems into lower training and delivery costs, and whether Ant Group's investment map can further form synergy, still awaits verification by subsequent projects.