A kung fu show demonstrated dramatic progress in China’s humanoid robotics over the past year. But a bigger challenge lies ahead: how to make these humanlike robots useful in real-life environments.
Key takeaways
- China wants to build on its tech prowess to make robots that can thrive in the real world.
- The challenge is getting from teleoperation to full automation.
- We’re around two years away from some meaningful breakthroughs.
Download the Analyst Survey 2026 PDF
Aside from fireworks and dumplings, watching the televised annual Spring Festival Gala, or Chunwan, has been a holiday tradition for every Chinese family on the eve of the new year.
For decades, the four-hour show was a jumble of songs, dance routines, and skits – all performed by human beings. But starting from last year, humanoid robots took centre stage, giving demonstrations of what Chinese robotics can do. In 2025, dancing robots twirling handkerchiefs wowed viewers. This year, we saw something more stunning.
Dozens of humanoid robots practiced martial arts, mimicking action star Jackie Chan’s ‘Drunken Boxing’ style – made famous by his iconic 1978 movie ‘Drunken Master’ – which requires incredible balance, flexibility, and strength. They also performed nunchaku, aerial flips, and executed three consecutive single leg backflips. The performance, well synchronised to music and the human performers, was far more complex than the handkerchief dancing the year before. It was not just a kung fu show, but a message to the world about how fast China’s humanoid robotics is evolving.
Manufacturing expertise and government support have given China the edge in developing robotics. China’s robust hardware supply chain, much of it built up through the electric vehicle sector’s R&D on everything from sensors to batteries, has helped local companies to iterate faster than western competitors. According to the draft of the 15th five-year plan, which sets strategic objectives and policies for 2026-30, Beijing is prioritising several emerging technologies, including advanced chips, robotics, batteries, and the brain-computer interface.
Despite the marvellous choreographed demonstrations, a bigger challenge lies ahead: how to make these humanlike robots useful in real-life environments. That is no mean feat. The machines need to figure out on their own how to get a job done and how to work with other robots, without laborious programming for each situation. Brain function hurdles, co-ordination between brain and hardware, and a lack of data for 3D real-world movement mean widespread commercial deployment still faces technical and cost challenges. It’s one thing for a desktop-bound large language model to make a mistake, but a very different thing for a robot that’s physically moving among humans to do so.
Mixing with the robots
I recently visited more than a dozen Chinese robotic original equipment makers in mainland China. Some are developing versatile robots aimed at manufacturing, logistics, hospitality, and household applications. With big leaps in hardware technology, the robot makers are already good at balance, motion control, and dynamic locomotion. Their focus is now on refining the foundational model, which is a large-scale artificial intelligence system trained on massive, diverse datasets that will act as a ‘brain’ for robots, allowing them to plan movements, co-ordinate with each other, and adapt to new situations.
Most players still rely heavily on teleoperation and videos to gather real-world data and refine foundation models. This is both time consuming and costly, but recent progress related to the generation of synthetic data might become increasingly useful in training the models, accelerating the whole refining process.
With continued improvement in robotic foundation models, I believe humanoid robots can start becoming productive in factories or warehouses, performing repetitive tasks such as box carrying or sorting, within a couple of years.
The number of robot developers has increased rapidly in China, which could intensify competition within the sector. But competition will drive startups to innovate, iterate faster, and improve their products to stand out.
I look forward to the robot show at the next Spring Festival gala.
