China's embodied intelligence sector is moving from technical showcases toward tougher tests of real-world commercialization, as rapid advances in humanoid robotics performance fuel both investor enthusiasm and rising expectations for practical industrial applications.

The trend was on full display at a humanoid robot half marathon in the tech hub Beijing E-Town, where autonomous robot "Lightning" completed the race in April in a net time of 50 minutes and 26 seconds.

The time was faster than the human men's half-marathon world record and cut nearly two-thirds off the 2:40:42 result posted by Tiangong Ultra, last year's robot champion.

The 21.0975-kilometer course featured flat roads, steep slopes, consecutive bends, speed bumps and other complex urban road conditions, including 22 regular curves and several near-90-degree turns, placing high demands on the robots' balance, path planning and joint response speed.

"A marathon on an urban course is an extreme stress test for core robot capabilities including motion control, autonomous navigation, endurance, heat dissipation, environmental adaptability and system reliability," said Liang Hongjun, an official of the intelligent manufacturing and equipment industry division at the Beijing Municipal Bureau of Economy and Information Technology.

Liang said the marathon provides a standardized task that allows teams to improve technologies around a common goal, complete a full technical route through competition and accelerate the application of results.

At the inaugural event a year ago, 20 teams took part and only six finished, with a completion rate of less than 30 percent. This year, the number of teams surged to more than 100, nearly five times the previous level, while the completion rate exceeded 45 percent.

"The leap in performance is essentially a comprehensive validation of hardware reliability, algorithmic stability and systems engineering capability," said Yan Weixin, chief scientist at Shanghai Artificial Intelligence Research Institute, adding that this is not a single-point breakthrough of any one technology, but rather the co-evolution of all relevant technical dimensions.

A clear trend seen at this year's half-marathon was the shift from remote-controlled running to autonomous navigation, with robots increasingly able to read the road, avoid obstacles and plan routes without human control.

Among more than 300 participating robots this year, autonomous-navigation entries accounted for nearly 40 percent. By contrast, last year, apart from the champion Tiangong Ultra, which used follow-based semi-autonomous navigation, all other teams relied on remote control.

To encourage more autonomous teams to finish, the race recorded real-time results for autonomous entries, while remote-controlled teams had their times multiplied by a factor of 1.2.

Zhao Mingguo, a researcher at the Department of Automation of Tsinghua University, said this year many teams had successfully transferred autonomous driving technologies from vehicles to robots, using lidar, pure-vision navigation and antennas to improve environmental perception.

"Which technological route will ultimately adapt best to different demands still requires further testing through more real-world tasks," Zhao said. "Over the next one to two years, indoor and outdoor robot navigation technologies are likely to advance even faster."

Behind the robots' improved performance lies rapid progress in core technologies. The winning Lightning robot developed by Chinese smartphone brand Honor delivered a maximum motor torque of 600 newton-meters, compared with about 420 newton-meters for last year's competitors.

In addition, its battery system supported more than 10 kilometers of running on a single charge, requiring only one battery replacement during the entire race.

Thermal management technologies have also advanced significantly. While many robots in last year's half marathon relied on simple air-cooling systems, Lightning adopted a liquid-cooling system capable of sustaining motor speeds above 20,000 revolutions per minute while circulating more than four liters of coolant per minute, similar to how blood circulation regulates human body temperature.

"Judging from the race performance, China's humanoid robots have moved beyond the entry-level stage of 'being able to run' and entered a more practical stage of 'running steadily and farther'," said Hao Lulu, a senior analyst at CCID Consulting.

Hao said some humanoid robots can now handle slopes and sharp turns independently, while leading autonomous-navigation models have real-time obstacle avoidance capabilities and dynamic balance algorithms that support stable posture adjustment at high speed.

Beyond the racetrack, momentum is also building in capital markets. According to the Embodied Intelligence Development Report 2025 released by the China Academy of Information and Communications Technology, China's embodied intelligence and robotics sector recorded 744 investment deals in 2025, with total financing reaching 73.54 billion yuan ($10.82 billion).

Since the start of 2026, total financing in embodied intelligence has climbed to more than 34.5 billion yuan, according to financial information platform Sina Finance, with as many as 23 companies entering the "10-billion-yuan club".

Notably, the national AI industry investment fund made its first move in the embodied intelligence track by participating in a 2.5 billion yuan financing round for major humanoid manufacturer Galbot — a move widely viewed by the industry as a sign that embodied intelligence is increasingly a key component of the country's national AI strategy.

"The robust growth of China's robotics industry is the result of multiple factors working together, including top-level national strategic planning, industrial policy support, and capital market responsiveness," said Yu Huan, a fund manager at Great Wall Fund.

Meanwhile, a growing number of robotics companies are preparing for capital market listings. Unitree Robotics had its STAR Market IPO application accepted by the Shanghai Stock Exchange in March, while companies including Leju Robotics, Deep Robotics and Galaxea are also accelerating restructuring plans ahead of potential listings.