Xiaomi trials humanoid robots at its EV factory

Lei Jun, Xiaomi CEO, said in a social media post that humanoid robots have begun trial operations at the company’s automobile factory, and the company plans to deploy large numbers of them across its production facilities within five years, TechNode reported.

He said the robots are based on a general-purpose vision-language-action (VLA) foundation model called Xiaomi-Robotics-0.

By integrating multimodal perception and reinforcement learning, the robots have initially achieved autonomous operations in tasks such as loading self-tapping nuts and transporting material boxes.

Xiaomi has been expanding beyond smartphones into electric vehicles and advanced manufacturing as it seeks new growth drivers in technologies such as robotics and AI.

🔗 Source: TechNode

🧠 Food for thought

Implications, context, and why it matters.

Xiaomi’s robot trial is one small step in a giant industrial race

Xiaomi ran a factory test. Its humanoid robot installed self-tapping nuts with a 90.2% success rate, which keeps the technology in an early phase 1.
About one in ten tries failed. CEO Lei Jun said a production line needs steady repeatability, unlike research labs where misses are acceptable 2, 3.
Competition is intense. In 2025, Chinese firms shipped 87% of the 13,317 humanoid robots sold worldwide 4. Tesla shipped 150 units, and Figure AI (a venture-backed humanoid robotics startup) shipped 150 4. Agility Robotics (a maker of warehouse-focused humanoid robots) shipped 150 as well 4.

Real-world data is the new oil in the humanoid robot race

The factory rollout aims to learn more than to swap out workers. Real interactions help train Xiaomi’s vision-language-action (VLA) foundation model 3.
Chinese companies often place robots in live settings to capture operating information, then feed it into AI control updates for faster iteration 5.
More units on the floor create broader logs, which speeds tuning for perception and control software 5.
The “deployment-first” approach uses China’s manufacturing base for quick prototypes. It could let its AI move faster than rivals who keep more work in lab hardware programs 5, 4.