Chinese street dancer Wu Yufei spent most of his 13-year career performing solo. Now, his dance partners are humanoid robots.

The 27-year-old Sichuan native and his robotic crew, Homies, have reached the final of America’s Got Talent, combining street dance with machines developed by Chinese robotics company Unitree.

Wu told Agence France-Presse (AFP) that months of rehearsals have changed how he sees the robots.

“It almost feels as if there’s a kind of team spirit there,” he said from Pasadena, California, ahead of the final.

Robot dancing has surged in popularity in China since Unitree humanoids appeared during the country’s Spring Festival gala, prompting businesses and event organizers to book robots for performances.

But teaching a robot to dance is far from automatic.

Human movements must first be captured, simulated and translated into instructions that the machines can execute. A 30-second routine can take days or even weeks to refine, according to robotics motion-capture specialists.

Robots excel at spinning and other power moves, but flexibility and musical improvization remain major weaknesses. Most routines are pre-programmed, meaning the machines cannot easily adjust if the music or choreography suddenly changes.

For choreographers, that means designing routines around what robots can actually do.

Despite the limitations, the performances have become a showcase for China’s rapidly growing humanoid robotics industry.

Industry figures see dancing as more than entertainment. Complex routines demonstrate balance, motor control, reliability and coordination, all of which could eventually translate into more practical applications.

China accounted for more than 97 percent of global humanoid robot shipments in the first half of 2026, according to figures cited by AFP, while Morgan Stanley has forecast the country’s humanoid robotics market could reach $2 billion this year.

Still, commercial uses beyond entertainment remain limited, and executives acknowledge that humanoid robots may be years away from becoming broadly useful to consumers.

For Wu, one advantage is already clear: once one robot learns the choreography, the same movements can be copied almost instantly across the entire crew.

“Their knowledge can be replicated extremely quickly,” he said.

“Not only does a single robot learn quickly, but they can all learn the same things at the same time, which is quite terrifying.”