Humanoid robots to see 'large-scale deployment'

Defining 2026 as the inaugural year for embodied artificial intelligence deployment, executives from Shanghai-based robot company Agibot said that the industry this year will witness embodied intelligence transitioning from technological validation to large-scale commercial application.

Yao Maoqing, partner and senior vice-president at Agibot, made the comments during an interview on Monday. This process will also indicate the industry's progress from customer pilots to closed-loop business models, he said.

In June, Agibot achieved production of 15,000 humanoid robots, making it an industry leader in both production speed and total output.

The three-year-old company has already accomplished a key milestone from laboratory demonstrations to real-world industrial production lines. From June 23 to 28, its flagship product underwent continuous 24/7 prototype validation on a tablet manufacturing line at Longcheer Technology in Nanchang, Jiangxi province.

Eight Agibot robots performed high-speed and high-precision tasks of material loading, unloading and transfer tasks, with accuracy controlled within 1 millimeter. The robots can automatically distinguish between qualified and non-qualified products with a 100 percent success rate.

This deployment demonstrates that the robots are now performing actual production functions in real industrial environments, rather than remaining at the concept demonstration stage, according to Yao.

"The years of 2026 and 2027 will witness large-scale robot deployments in factories. Agibot products will expand into work scenarios in the fields of consumer electronics, semiconductor packaging and testing, the automotive industry chain, logistics and warehousing," said Yao.

Agibot expects its total shipment to come between 40,000 and 50,000 units this year. The first workstation took about four months to go live, while subsequent workstation replication was shortened to one to two weeks, demonstrating strong technological transfer and scenario replication capabilities, he said.

Robots not only address challenges such as recruitment difficulties and high employee turnover, but can also significantly reduce training costs and quality consistency risks caused by workforce fluctuations. Immune to emotional variability and unscheduled absences, robots can also greatly enhance production management stability, explained Yao.

Agibot released the industry's first open-source large-scale dataset for embodied AI in 2024 and has established a presence in world model research. This year, the company plans to further expand its dataset scale, aiming to build the world's largest and highest-quality data platform to support the training and iteration of robots' "brains", according to Yao.

According to Zhan Kun, senior vice-president of supply chain at Agibot, it took the company more than two years to see its production capacity grow from zero to 5,000 units. But it only took the company six months to scale up from 5,000 to 15,000 units, which can be largely attributed to the strengths of new suppliers in large-scale manufacturing.

Automotive industry chain companies and listed companies accounted for approximately 60 percent of Agibot's new suppliers introduced last year, a ratio which has continued to rise in the first half of this year, said Zhan.

Agibot has also facilitated the rapid growth of a number of startups, some of which have evolved from early-stage ventures to public companies. They have become Agibot's long-term partners capable of mass production and delivery, he said.

"China enjoys an absolute global advantage regarding its manufacturing delivery capabilities. The diversity of data arising from China's comprehensive range of manufacturing sectors will provide unique advantages for training embodied AI large models," said Zhan.

Therefore, the primary bottleneck currently constraining large-scale commercial adoption is not supply chain manufacturing capacity, but rather the ability to identify more commercially viable deployment scenarios to establish closed-loop business models, he said.

Another major challenge is the supply of core materials such as chips. The company is thus seeking multi-source supply assurance and in-house research and development of critical parts, he added.

shijing@chinadaily.com.cn