BYD and Samsung-backed humanoid robotics firm PasiNiao starts IPO guidance process

According to a September 21 filing on the official website of the China Securities Regulatory Commission (CSRC), PaxiNi (帕西尼), an embodied AI robotics company valued at over 10 billion yuan, has declined to disclose its chosen listing venue. In June of this year, reports emerged that PaxiNi had been in contact with investment banks to evaluate the possibility of a Hong Kong IPO.

A Growing Queue of Robot Companies Seeking Listings

Amid the surging interest in embodied intelligence, a growing number of robotics companies are racing to seize the IPO window. To date, at least eight embodied robotics companies have publicly disclosed their IPO progress or plans. These include:

• Yunshenchu (云深处)
• Leju Robotics (乐聚)
• Zhiyuan Robotics (智元机器人)
• PaxiNi (newly added)

Their target listing venues are spread across the STAR Market (科创板), the ChiNext Board (创业板), and the Hong Kong Stock Exchange. Some other companies have chosen to keep their listing progress confidential.

Shareholding Structure and Funding History

According to the filing, PaxiNi's controlling shareholder is founder and CEO Xu Jincheng (许晋诚), who holds 37.07% of the company's shares.

Tianyancha data shows that PaxiNi, founded in 2021, has completed 10 rounds of financing to date. Most recently, in mid-September, it closed a B++ round worth several hundred million yuan, with strategic investment from Samsung. Prior to this, BYD also signed a strategic cooperation agreement with PaxiNi, announcing deep collaboration in data collection and robotics to drive substantial breakthroughs in industrial manufacturing scenarios.

Large-Scale Embodied Intelligence Data Collection

PaxiNi is also expanding its large-scale embodied intelligence data collection efforts. In an interview with Southern Metropolis Daily in July, Xu Jincheng stated that the company has deployed data collection projects in multiple cities, including Wuxi, Tianjin, and Suqian. These projects involve data collection in public spaces, factories, and private locations with consent, aiming to cover as diverse a range of real-world environments as possible.

In his view, the richness of modalities directly impacts the volume of effective data available for model training. By integrating multi-dimensional sensors into data collection devices, the quality and effectiveness of the acquired data can be exponentially improved.

Xu told reporters that PaxiNi has consistently adopted a human-centric data collection strategy, ensuring that the collected data is better suited for human-robot interaction.

Human Data Collection Methodology

According to demonstration footage on the company's official website, PaxiNi's so-called "human data collection" method involves collectors wearing data acquisition devices on both hands to capture information such as vision, touch, and contact point pose during operational actions. The company claims that this method outperforms traditional teleoperation-based collection and simulated data synthesis in terms of collection efficiency, fidelity of motion and touch, and cross-embodiment naturalness.

Source: Southern Metropolis Daily