China Embodied AI Nears Critical Leap from Usable to Good

China’s embodied AI industry is approaching a pivotal inflection point, transitioning from functional prototypes and pilot programs toward large-scale commercial deployment. With over 3,000 registered enterprises, a landmark IPO, and a major government-backed deployment initiative, the sector is making what industry insiders describe as a “critical leap” — moving from being merely “usable” to genuinely “good.”

Industry at a Glance

As of May 2026, China has 3,025 registered embodied AI-related enterprises, according to Qichacha data cited by Xinhua News. In 2025 alone, 408 new companies were registered — a 119.35% year-on-year surge. Notably, companies established for over a decade account for 36.26% of the total, signaling a mature industrial base rather than a speculative bubble.

The city of Tianjin has emerged as a key hub, forming a complete industrial chain covering core components, body manufacturing, and system integration with 104 enterprises. The industry chain revenue exceeded 27 billion RMB in 2025, and Q1 2026 output value grew 10.1% year-on-year.

Policy Momentum: The MIIT-SASAC Action Plan

In June 2026, China’s Ministry of Industry and Information Technology (MIIT) and the State-owned Assets Supervision and Administration Commission (SASAC) jointly launched the “2026 Annual Humanoid Robot and Embodied AI Real-Scene Training Special Action,” as reported by the 21st Century Business Herald. The initiative targets a bold goal: by the end of 2026, key products should complete application verification and achieve normal deployment in representative scenarios, with a 10,000-unit scale deployment capability.

The action plan covers 10 provinces and municipalities plus central state-owned enterprises, focusing on industrial manufacturing, public services, and special operations. It operates through “Innovation Application Consortiums” that link scenario owners, robot makers, component suppliers, and research institutes.

Landmark IPO: Unitree Technology

In a bellwether moment for the industry, Unitree Technology, a leading humanoid robotics company, passed its IPO review on the STAR Market in June 2026, planning to raise approximately 4.2 billion RMB. Unitree founder Wang Xingxing told Xinhua that for embodied AI to experience a “ChatGPT-like revolutionary moment,” multiple key technical challenges still need to be overcome, with insufficient generalization capability being the industry’s most recognized core bottleneck.

The World Intelligence Industry Expo 2026

The 2026 World Intelligence Industry Expo, held in Tianjin starting May 28, featured for the first time a dedicated Embodied AI pavilion, according to Interface News. Over 80 enterprises showcased more than 150 embodied AI products, from fighting robots with fluid motion to dexterous hands capable of threading a needle.

Galileo (Tianjin) Technology demonstrated robotic dogs capable of autonomous stair navigation and slope traversal, already deployed in energy and emergency response sectors for 24-hour inspections. Manager Tang Ziqi told Xinhua the company has secured tens of millions of RMB in intent orders at multiple international exhibitions in 2026.

Surging Investment

The investment landscape has been extraordinary. According to OFweek, at least 500 million RMB per day has been flowing into embodied AI in 2026. The first two months alone saw cumulative financing exceed 37.3 billion RMB. In 2025, the sector recorded 325 financing rounds totaling nearly 40 billion RMB, a 326% year-on-year increase.

Tencent News reports that investment logic is shifting from “concept chasing” to “industrial depth.” Capital is increasingly flowing toward companies with demonstrated delivery capabilities and real-world deployment, with a growing focus on upstream components, sensors, and dexterous hands.

The Challenges Ahead

Despite the momentum, significant hurdles remain. Zhang Yuehai, Chief Advisor at Beijing Xinjiang Suyan Technology Services, told Xinhua that “70% of our project time is spent on data cleaning; only 30% is actually used for AI training.” Industrial data exists in vast quantities but remains fragmented — inconsistent formats, poor labeling, and difficult multimodal alignment plague the sector.

Zhu Shiqiang, Dean of Zhejiang University Robotics Institute, described the industry as being in a “lively but awkward” stage — many prototypes, but few truly large-scale application products. Key bottlenecks include multimodal large models (the “brain”) and dexterous hand core technologies that have not yet achieved a breakthrough.

Xing Jie, Chairman of Youshi Capital, noted that the domestic humanoid robot industry hasn’t been stuck on core technology, but rather on the “last mile” of industrialization — the gap between a working prototype and a reliable, cost-effective, mass-producible product.

Talent Development

To address the growing talent gap, nine top Chinese universities — including Harbin Institute of Technology, Beijing University of Aeronautics and Astronautics, and Shanghai Jiao Tong University — have launched dedicated embodied AI undergraduate programs in 2026, as reported by 36Kr. Most programs follow a “small but elite” model, with six of the nine schools admitting only 30 students nationwide. Industry demand is intense: recruitment platform data shows embodied AI positions offer an average annual salary of 333,800 RMB, with top roles reaching 15 million RMB.

Market Outlook

IDC predicts that by 2026, China’s humanoid robot application scenarios will increase to more than three times current levels, with market size approaching $1.3 billion. Domestic institutions predict overall embodied AI robot shipments could reach 100,000+ units in 2026. The State Council Development Research Center projects the market could reach 400 billion RMB by 2030 and exceed 1 trillion RMB by 2035.

What to Watch

The coming months will be critical. Key questions include whether the 10,000-unit deployment target by end of 2026 will be achieved, how the Innovation Application Consortium model performs in practice, and whether Chinese companies can move beyond closed industrial scenarios into open consumer environments. The industry’s ability to overcome data quality, model generalization, and standardization challenges will determine whether 2026 truly becomes the “first year of commercial verification” for China’s embodied AI sector.