Maniformer closes angel+ funding to scale physical AI data infrastructure

  • Round led by Guofang Capital, with participation from CDH VGC, Fortera and others
  • Startup targets “data scarcity” bottleneck in embodied intelligence with million-hour output goal

Maniformer (觅蜂科技), a physical AI data services platform incubated by robotics firm Agibot (智元机器人), has raised several hundred million yuan in a new angel+ funding round.

The round was led by Guofang Capital, with participation from Fortera Capital, Shanghai Seari Private Equity Investment and Yuanqi Robot. Existing investors Ningbo PIA Automation Holding Corporation and CDH VGC Fund also increased their stakes.

Based in Shanghai, Maniformer positions itself as a physical AI data infrastructure provider, built around data collection, processing and deployment for embodied intelligence systems. Its core team draws from autonomous driving labs, AI research groups and large internet companies.

The company is led by Chairman and CEO Yao Maoqing, who is also a partner and senior vice president at Agibot. Corporate registry records indicate Agibot holds a 75% controlling stake in Maniformer.

The data bottleneck

The funding comes as embodied AI developers face a shortage of high-quality training data, inconsistent standards and fragmented datasets.

Industry estimates suggest a gap of more than 5 million hours of high-quality usable data by 2026, a bottleneck seen as constraining sector development.

Yao said data will become a foundational production input alongside compute in AI systems, adding that Maniformer aims to reach millions of hours of data production capacity by 2026, spanning factories, logistics, retail stores and home environments.

The startup has deployed multiple data collection modalities, including teleoperation, hardware-independent capture and simulation-based generation, forming an end-to-end stack across hardware, software, platforms, scenarios and operations.

All images are downloaded from Maniformer’s official WeChat account

Human-in-the-loop capture

Its recently launched MEgo hardware line includes data-capture gloves and wearable devices that shift collection away from robot-dependent systems toward human-in-the-loop capture, lowering cost and entry barriers.

“Maniformer is an independent data services platform. The only way Agibot obtains data from Maniformer is through market-based orders,” Yao said, seeking to dispel speculation that the company would rely on its parent for funding.

The company has partnered with clous server providers such as JD Cloud, Baidu Cloud and Alibaba Cloud, and is working with the Shanghai Electric Apparatus Research Institute and China’s national data standards body to launch what it calls the “Hive Data Co-Creation Initiative,” targeting 10 billion hours of data capacity by 2030.