Shenzhen is using the Asia-Pacific Economic Cooperation (APEC) summit this week to showcase its technological ambitions, as the city of 18 million residents hosts global leaders including Donald Trump and Vladimir Putin.
While the metropolis highlights its manufacturing prowess, a large gap remains between public enthusiasm and the current capability of AI-driven machines.
According to a survey conducted this year by Ipsos, 83% of Chinese respondents expressed enthusiasm for AI-powered products and services, compared to just 33% in the United States and France, Geo.fr reported. This optimism contrasts with the technical reality on the ground, where industry professionals acknowledge that robots still struggle with complex domestic tasks like ironing or cleaning.
In a local park, AI² Robotics has deployed humanoids that prepare coffee and ice cream for 3 to 4 euros.
While these units can manipulate containers and operate levers, they remain limited; a human must still manually refill coffee beans and empty the grounds tray, Geo.fr reported. The company claims to have 200 such server-humanoids deployed across China, with one park location selling up to 300 coffees daily.
The push for “embodied AI”—machines capable of perceiving and physically interacting with their environment—is supported by heavy industrial investment. Chinese tax authorities reported that spending on these technologies surged by 230% year-on-year during the first five months of 2026.
| Metric | Figure |
| Industrial spending growth (Jan-May 2026) | 230% |
| AI enthusiasm in China | 83% |
| AI enthusiasm in U.S. and France | 33% |
The industry is currently focused on high-stakes or repetitive labor. At a Shenzhen sorting center, robots use AI to recognize packages and coordinate movements, while at the battery manufacturer CATL, Spirit AI humanoids are tasked with connecting high-voltage components. Despite these deployments, experts suggest the technology is still in its infancy.
“The robots are still like babies,” said Kang Yu, founder of the startup Strange Attractor, which collects human movement data to train AI models. She told Geo.fr that while the machines struggle with instinctive tasks, she anticipates a “massive deployment” of AI-enabled robots across China in 2027.
Other observers remain cautious regarding the economic viability of the current generation of hardware. Dermot McGrath, founder of the innovation consultancy ZenGen Labs, said that because these machines often require human intervention when facing unexpected obstacles, they may not yet provide the intended personnel savings.
“The more a robot requires human attention, the less it allows for staff savings,” McGrath explained to Geo.fr. Whether these systems can transition from controlled, repetitive environments to the unpredictable conditions of everyday life remains the primary challenge for the industry.