Funding for China's Embodied AI Sector Cools in Fourth Quarter
Hu Shujuan
DATE:  Nov 18 2026
/ SOURCE:  Yicai
Funding for China's Embodied AI Sector Cools in Fourth Quarter Funding for China's Embodied AI Sector Cools in Fourth Quarter

(Yicai) Nov. 18 -- Investment activity across the embodied intelligence sector in China has softened this quarter, including fewer fundraisers for large language model projects, while sensor makers, dexterous hand developers, and robotics firms with clear use scenarios continue to secure fresh capital.

In the 70 funding rounds in the embodied intelligence sector since the start of last month, StarDust Intelligence and FAIRINO were the only two firms to raise over CNY1 billion (USD150 million), compared with 25 such cases in the first half of this year, according to data from new-economy venture capital statistics provider IT Juzi. Most other fourth-quarter fundraisers have been for hundreds of millions of Chinese yuan (tens of millions of US dollars).

In comparison, Bluepoint Touch Beijing Technology, a developer of force and torque sensors for industrial and humanoid robots, wrapped up a Series D3 funding round yesterday, raising more than CNY1 billion over the past six months. Founder Liu Wuyue said the company may launch another round by the end of this year.

Financing appetite remains robust among upstream component suppliers and downstream application players, Liu Wuyue said in a recent interview with Yicai, noting he has not seen a meaningful pullback in funding for these segments.

However, leading Chinese venture capital companies have also sharply cut back on the embodied intelligence sector since October. HSG has not backed any relevant projects, while IDG Capital and Hillhouse Investment Management have each made only one investment.

The post-first-half cooling is a healthy market adjustment after the sector’s red-hot fundraising run earlier this year, Liu pointed out.

Pure-play embodied artificial intelligence model developers are losing their edge, several industry insiders said to Yicai. While general LLMs can execute certain tasks, they remain slow for physical manipulation, with low success rates for workflows requiring force, tactile sensing, and closed-loop feedback control.

Editor: Martin Kadiev

Follow Yicai Global on
Keywords:   Embodied intelligence,Venture Capital