A humanoid maker investing in a magnet developer ties AgiBot to the motor materials its robot joints depend on, with CleerScience targeting mass production in 2027.
CleerScience, a Beijing startup that applies AI to materials research, has closed an angel round of nearly RMB100 million, according to Pedaily and a funding roundup from 36Kr on November 10. Both name the investors as China Fortune-Tech Capital (中芯聚源), Hillhouse’s GL Ventures (高瓴创投), humanoid robot maker AgiBot (智元机器人), Finewill Capital (风物资本), Orient Renaissance Capital (东方嘉富) and several funds with industry backing. Neither report says who led the round or what the company is now worth.
The company was founded in 2026. Less than six months after it was founded, CleerScience says it has signed orders worth tens of millions of yuan. Pedaily said its customers and partners include leading organizations in consumer electronics, energy, semiconductors and aerospace.
AgiBot’s presence on the register is the detail to note. CleerScience’s nearest commercial product is a rare-earth magnet, and Pedaily reports that the startup is working with several leading robot makers on magnets for the motors that drive humanoid hips, knees, shoulders and hands. AgiBot shipped about 8,400 humanoids in the first half of 2026, ranking first worldwide in research firm SAG’s count, so it has a direct interest in that supply.
The startup is selling a materials lab run by AI models
CleerScience describes its goal as an "AI Science SuperFactory": software and lab equipment that take a new material from simulation through experiments and process scale-up to factory production, with each stage feeding data back to the models. In plain terms, it wants to shorten the cycle of designing, making and testing a material, which in industry often takes years.
The data comes from instruments. CleerScience builds what it calls AI-native instruments and adapts existing ones so their measurements can be read directly as training data. Pedaily reports that it has gathered tens of millions of records from real experiments. Its models interpret characterization methods, the lab tests that show what a material is made of and how its atoms are arranged, including X-ray diffraction (XRD), electron microscopy, Raman spectroscopy, XPS and NMR. The company says its XRD model ranked first on MatSciBench, a benchmark from the Institute of Physics at the Chinese Academy of Sciences, and that it will soon release a single model covering more than 20 characterization methods.
The equipment plan has three stages: AI models embedded in single instruments, then one device that measures composition, structure and shape together, and finally an all-in-one unit in which AI plans experiments across design, synthesis, characterization and testing and schedules equipment around the clock. Revenue is to come from three lines: model subscriptions and AI equipment services, contract materials research for clients, and materials whose intellectual property CleerScience owns itself.
Founders from US chipmaking set the industrial focus
Chief executive Dong Siming (董斯鸣) studied at Tsinghua University and earned a PhD in materials science at Stanford, according to Pedaily. He spent more than 15 years in semiconductors and industrial research, led an AI-driven process development project at a leading US chip company and ran an AI business unit there with nearly US$200 million in annual revenue. Co-founder and chief operating officer Zhao Zhenlin (赵震林) worked as an applied scientist at a US plasma technology company and led research and production-line construction for heavy rare-earth magnets in China. Co-founder Wang Huan (王欢), the chief scientist, holds a Tsinghua PhD and works on multimodal models.
The company has raised money quickly. In July the company announced a first round of tens of millions of yuan from a single investor, Topview Innovation (鼎峰科创), part of SummitView Capital (武岳峰科创), Cyzone reported. Li Feng, a founding partner of the investor, said then that what made the startup scarce was its "deep industrial insight and ability to turn research into production" (in EastFrontier’s translation). Hillhouse set up GL Ventures in 2020 to back early-stage companies, KrASIA reported at the time.
Local government has also taken notice. A working group of Beijing’s general AI industry partner program visited the company on September 16, according to the Beijing Municipal Bureau of Economy and Information Technology. The bureau’s account lists products for predicting material properties, characterization and process optimization, and joint work on alloys, rare-earth magnets, battery materials and semiconductor materials.
A 2027 magnet launch will test the model
The first product with a date is a new generation of rare-earth permanent magnet material. Pedaily reports that it has moved into pilot-scale conversion, that a pilot base is planned in Beijing and that CleerScience aims for mass production in 2027. The company puts global demand for such magnets at about 170,000 tonnes by 2030, a market it values at more than RMB50 billion, across electric vehicle drive motors, wind turbines and servo motors. For robot joints, it says the aims are small size, high torque and stability at high temperature.
Further out, CleerScience lists high-entropy alloys for fusion and high-end equipment, and semiconductor materials including photoresists and new memory devices. Those lines have no announced timetable.
The upcoming multi-method characterization model is the nearest milestone. The clearer test comes after it: whether a Beijing pilot line produces magnets that a robot maker such as AgiBot puts into its joint motors.