Tam told the Financial Times that the company’s Walker S2 robots currently match just 30–50% of human productivity and only in specific tasks such as stacking boxes and quality control.
Yet manufacturers are still racing to order them to avoid losing out to competitors. "You can imagine . . . if Tesla has the advantage of deploying their own human robots into the manufacturing line, that means maybe BYD, they are staying behind."
According to Counterpoint Research, 16,000 humanoid robots were installed globally in 2025, mainly for data collection and research, as well as logistics, manufacturing, and automotive use. A separate report by Omdia said global shipments jumped nearly 480% in 2025 to 13,318 units. UBTech delivered 1,000 humanoid factory robots last year, ranking third globally in shipments, and aims to produce 10,000 by the end of this year.
Despite the rapid growth, Marco Wang, a Shanghai-based researcher at Interact Analysis, said many deployments remain experimental, with "a lot of challenges" before commercial operation.
Interesting Engineering noted that humanoid robots face multiple constraints that limit productivity on factory floors. Most rely on torso-mounted or backpack batteries, limiting active work to a few hours, as locomotion, balance, and manipulation consume large amounts of energy.
Most general-purpose humanoids handle only moderate loads comparable to humans, as higher payloads complicate balance and actuator design. Humans also remain faster and more efficient in complex environments, adapting quickly to varied items while bipedal robots move more cautiously.
UBTech humanoid robots work in a factory. Video courtesy of UBTech
Tam said one challenge UBTech aims to solve this year is developing a multifunctional hand, as current Walker models require humans to swap appendages for different tasks. The company also aims to raise Walker’s performance to 80% of human efficiency by 2027.
Kelvin Lau, an analyst at Daiwa Capital Markets, told Financial Times that humanoid robots should be "gradually improving," adding that 80% of human efficiency may be sufficient since robots do not need breaks or holidays.
Omdia expects global humanoid robot shipments to reach 2.6 million units by 2035, driven by advances in AI models, dexterous robotic hands, and self-reinforcement learning that are making humanoids viable for industrial, service, and eventual household use.
Tam said future generations of the Walker robot would benefit from real-world data collected in factories where the machines are already deployed. "The more human robots that could be deployed into the real world, the more real data could be collected. And then, like a circle, it . . . [will] help human robots grow."