EN / 中文

Shenzhen APEC: What Are the Robotics LRobotics Takes Center Stage at 2026 APEC CEO Summit: From AI Hype to Real-World Scalingeaders Discussing?

by hangjiashuojiqiren·November 19, 2026

Hangjia Shuo Robotics

On November 17, the 2026 APEC CEO Summit opened at the Qianhai International Convention Center in Shenzhen. Representatives from 21 economies, over 2,000 political and business leaders, and 115 Fortune Global 500 companies attended.

Artificial Intelligence was the most debated topic of the summit, but the discussions on robotics were more grounded. Unitree's Wang Xingxing, UBTECH's Zhou Jian, Gree Electric's Dong Mingzhu, Xiaomi's Lei Jun, and AgiBot—which has integrated robots into amusement parks and post offices—did not talk about "disruption" this time. They discussed capacity, orders, and scenarios.

Wang Xingxing: Models are approaching a breakthrough; can capacity keep up?

Unitree went public on the SSE STAR Market in August this year, becoming the first "Embodied AI" stock listed on A-shares.

Wang Xingxing described 2026 as the "most exciting year," noting that the robotic achievements visible to the public are merely above the waterline, while iterations below the surface are progressing much faster. However, he immediately tempered expectations.

The industry's biggest pain point is model generalization capability—robots require retraining when switching scenarios, and a single device cannot adapt to diverse environments. He judged that technical progress this year is approaching a critical threshold, stating, "It might accelerate very soon."

The real test comes after the breakthrough. "We currently produce only tens of thousands of robots per year, but in the future, we may need to produce millions or even tens of millions. This remains a massive challenge." Manufacturing, reliability, and after-sales service must all scale accordingly. He added that future robot demand for computing power may far exceed most people's imagination.

Zhou Jian: There is a gap period for general-purpose humanoid robots; scenario data is the hard currency

Zhou Jian of UBTECH spoke more pragmatically: General-purpose humanoid robots will not arrive quickly; there is still a gap period. During this time, whoever possesses more scenario data and deeper understanding of edge-side applications will gain the advantage.

The path he outlined is "one strategy per domain," avoiding generic solutions. Robotic education curricula in hundreds of schools in Vietnam, Wukong robots in Seoul kindergartens in South Korea, cutting stations in Southeast Asian rubber plantations, and ultra-biomimetic companion robots for the elderly all rely on secondary development by local partners. He made a sobering point for the industry: Scenario differences across Asia-Pacific economies are too significant to "use one solution to capture the entire market."

He places greater emphasis on the scarce links in mass production chains: "Breakthroughs in single-point technologies are not difficult; what is truly scarce is bridging the gap from technology to large-scale mass production, and then converting it into productive forces that Asia-Pacific partners can actually use." The Greater Bay Area's greatest advantage lies in its mature supply chain—actuators, reducers, motors, and joint modules can all be sourced in Shenzhen; UBTECH's listing in Hong Kong allows it to leverage capital and scientific research resources.

He also addressed ethical boundaries. In September, UBTECH released the world's first "White Paper on Ethical Governance of Humanoid Robot Technology," proposing three goals: controllable actions, appropriate anthropomorphism, and moderate interaction. On the product side, Walker S2 is the world's first humanoid robot capable of autonomous battery swapping, operating continuously for 7×24 hours, and has been deployed in five industries: aviation, automotive, 3C electronics, smart logistics, and semiconductors.

Dong Mingzhu: Tell me what components you need, and I will make them

The most dramatic moment occurred between Dong Mingzhu and Zhou Jian.

Gree Electric has been developing robots for over a decade, with all robots on its production lines being self-developed. In 2019, the Gree Robot Band even performed on CCTV's Spring Festival Gala. At the summit, she proactively invited Zhou Jian to collaborate: "We didn't know each other before, but today we met through this platform. Whatever components you need, Gree will provide." She further clarified: She had just discovered high industrial complementarity with UBTECH, noting that Gree's reducers can be adapted for humanoid robots.

In the past, robot companies sought component manufacturers; now component manufacturers are volunteering themselves—upstream players are transitioning from order-takers to partners.

Not on display stands, but on assembly lines

The most noteworthy robotics scenes of this summit did not take place within the conference venue.

Lei Jun stated that Xiaomi has been working on robots for six years. Currently, embodied AI robots are "interning" at Xiaomi's automotive factory, capable of working fully for8 hours and independently charging twice during that period. He rated the potential of embodied AI as exceeding that of the automotive industry.

AgiBot partner Wang Chuang noted that nearly 500 AgiBot robots are already in daily operation at Chimelong Paradise in Zhuhai, and have also been deployed for actual work on factory production lines.

At Qianhai Stone Park and Nanshan Robot Valley, Cross-Dimension Intelligence's "Xiao Wei" is brewing craft beer, performing quality inspection for mobile phone packaging, and selling goods in agricultural support stores; Xsquarerobot's "Quantum No. 1" uses the same embodied brain to arrange flowers on display stands and sort electronic components on high-speed conveyor belts.

A more tangible application is in logistics. Robot Era's embodied AI logistics robots are operating regularly at 11 postal nodes across eight cities in Guangdong, supplying over 1,500 items per hour, with 45 units batch-deployed at the Hangzhou Postal Center.

Making bodies usable again

Han Bi Cheng of BrainCo brought the discussion back to the body: enabling those without hands to write, draw, and even play piano using neural-controlled prosthetics, and allowing those without legs to walk, run, climb rocks, and surf. He emphasized that people with disabilities and low-income groups are those who most need this technology.

Brain-computer interfaces and embodied AI are two ends of the same link: one end makes machines smarter, the other makes bodies usable again.

Several clear signals

Connecting these robotics-related remarks with industry data reveals clear trends.

First, there is a gap between optimism at the summit and reality on the balance sheet. Wang Xingxing said models might break through next year and capacity needs to catch up, while industry data shows: global shipments of humanoid robots in the first half of the year were approximately 22,000 to 25,000 units, a year-on-year increase of nearly 300% to 432%. The top five vendors are exclusively Chinese manufacturers, with AgiBot and Unitree together accounting for about three-quarters of the share.

Second, there is still distance between shipment structure and "actual work." According to the CAICT "Report on Embodied AI Development (2026)," entertainment/commercial performances and scientific education/data collection account for over 60%, while smart manufacturing accounts for only 13%, and warehousing/logistics for 5%; a considerable proportion of devices have an average monthly uptime of less than 10 days. In the first half of the year, there were 218 publicly announced winning bids in China totaling CNY 1.723 billion, with orders from industrial enterprises accounting for only 21.1%. The capacity concerns raised by Wang Xingxing are, in a sense, inflated by these devices that are "turned on but rarely moved."

Third, hot orders but cold deliveries are the most practical bottleneck. A report by 36Kr indicates that the ratio of intended orders to actual deliveries is only 10% to 20%. The first hurdle to crossing the "millions and tens of millions" target shouted by Wang Xingxing is not the production line, but this conversion rate.

This article is compiled based on public network information, for reference only, and does not constitute investment advice.