GenRobot AI, a Beijing startup that builds the data pipelines robot makers use to train their machines, has raised $100 million in a Series A+ round backed by carmaker SAIC Motor, search giant Baidu and Malaysia's Eastworlds, the company disclosed on Nov. 2.

The round closed in the third quarter, according to TechNode and Sina Finance, which cited the company. It is the latest sign that investors in China's embodied AI boom are spreading their money beyond robot hardware to the supply chain behind it. For investors in Asia, the cap table also says something: a Chinese carmaker, an internet platform and a Malaysian backer are all writing checks into the same data layer.

The deal arrives as Asia's venture market keeps tilting toward artificial intelligence. AI took 55% of the region's third-quarter startup funding, Capital Times reported last month.

Who is backing GenRobot

Sina Finance listed two groups of investors. On the industrial side were SAIC Motor, Eastworlds, Baidu, Shoucheng Holdings and Galbot, also known as Galaxy General Robotics. Financial investors included GL Ventures, Sunhelm, CITIC Goldstone, China Securities Investment, CCB International and Huatai Zijin.

The company did not disclose a valuation.

The mix matters. Galbot is itself a robot developer, so at least one of GenRobot's backers is also the kind of company that buys training data. SAIC brings a carmaker's factory floors, the sort of setting GenRobot records.

The round extends a rapid run of fundraising. In June, Gasgoo reported that Ant Group, Didi and Delian Capital led several back-to-back rounds totaling hundreds of millions of yuan, with existing investors including Shunwei Capital and BV Baidu Ventures also taking part. In late August, PEdaily reported that the company had closed a Series A led by Momenta, which PEdaily described as a leading physical AI company, without giving the amount. At that point, GenRobot said it had raised more than $200 million across seven rounds in roughly a year since its founding.

Taken together with the new $100 million, the company's disclosed fundraising now tops $300 million. GenRobot is registered in Beijing under the Chinese name Jianzhi, according to PEdaily.

Why robot data is drawing money

Large language models were trained on text scraped from the internet. Robots have no equivalent pool. To learn how to fold a towel, pick a part off a conveyor or stack boxes, a model needs recordings of those actions, ideally many thousands of hours of them.

GenRobot's pitch is to collect that data from people rather than from robots. According to TechNode, its recording systems capture visual footage along with hand and body movements in homes, factories and warehouses. Its Matrix platform then synchronizes, cleans and labels the recordings so robot developers can use them for training and testing.

The hardware includes a wearable capture device called Gen DAS and a skill-capture unit called Gen Controller, PEdaily reported. The company said its sensors record sound, touch, pressure and force feedback alongside the video. It has also trained its own "data foundation model," which it says tracks hands, body pose and depth from the footage.

Because the data comes from human demonstrations rather than a specific machine, the company describes it as body-agnostic: in principle, it can be used to train robots of different designs. GenRobot told PEdaily that clients had already pretrained new models entirely on its data and seen them transfer across settings in manipulation tests. That claim comes from the company and has not been independently verified.

Earlier this year, the company released an open dataset of more than 13,000 hours of such recordings, according to Gasgoo, which said it was drawing more than 1 million downloads a month.

What investors should watch

The bet is that data, rather than motors or chassis, becomes the bottleneck as humanoid and service robots move from demonstrations to real work. GenRobot frames its mission as moving AI from predicting the next word to "completing the next action," as Gasgoo reported.

The risks are familiar from other parts of the AI supply chain. Robot makers with deep pockets may decide to build their own data operations, and the value of any dataset depends on whether models trained on it perform in real factories and homes. GenRobot has not published revenue figures.

The company said in August that it plans to expand its data collection capacity, upgrade its foundation model, scale up testing on clusters of real robots and grow overseas. The Eastworlds stake gives it a Malaysian partner as it looks abroad.

For later investors, rounds like this one set a reference point for how the data and tooling layer of robotics is priced, and the exit routes are the usual ones (see how startup investors cash out in Asia). GenRobot has not announced any listing plans. Its next disclosure on customers or revenue will show whether its human-recorded data is turning into paying contracts at scale.