# Quick summary At the 2026 World Humanoid Robot Games in Beijing, Tiangong Ultra — a humanoid developed in China — ran 100 meters in 8.64 seconds, faster than the human world record of 9.58 seconds set by Usain Bolt. The result grabbed headlines, but the broader story is about rapid iteration: Tiangong ran the same distance in 21.50 seconds at last year's event.
# What happened on the track Tiangong posted an 8.64-second final after an 8.86-second semifinal. The robot's speed exceeded human sprint records, but safety and control remain imperfect: after its semifinal it collided with a padded barrier beyond the finish line, collapsed, and briefly sparked a small fire in its torso. Multiple robots at the event still crashed or fell when running at high speed.
# Why this improvement matters
# The event's design and goals
# Autonomy versus remote control More than 40% of challenges required full autonomy. Some robots were remotely controlled or given limited instructions, but the organizers and commercial developers want machines that can perceive, decide, and act without human guidance. Tasks such as finding a charging port, estimating angle and distance, and inserting a plug without damaging equipment remain difficult and require multi-step perception and manipulation.
# Commercialization and investment China is pairing experimentation with capital. XPeng's robotics division announced raising more than $900 million at a valuation above $6.3 billion. XPeng plans to scale manufacturing — targeting production of 1,000 IRON humanoids per month by year-end — and to deploy these machines first in stores and factories, with broader commercial sales planned in 2027. Early deployments will provide more real-world data to accelerate improvements.
# Where things stand now
# Bottom line The 8.64-second run is a milestone in raw speed for humanoid robots and a signal that rapid engineering cycles are producing big gains. However, crashes, safety incidents, and persistent autonomy challenges show that real-world utility — reliable perception, safe interactions, and adaptive problem-solving — remains the next hurdle.