Chinese tech giant Xiaomi has moved significantly closer to the practical implementation of the “lights-out manufacturing” (fully automated factory) concept. Yesterday, the company released the results of tests involving its own humanoid robots at an electric vehicle plant, demonstrating that industrial automation is ready to transition to a qualitatively new level.
For the global tech and investment community, this test marks a pivotal moment, signaling that autonomous systems are maturing faster than anticipated and are poised to disrupt traditional manufacturing supply chains and labor models.
Xiaomi CEO Lei Jun confirmed in an official statement that during the testing phase, the robots operated autonomously for three hours, performing complex assembly operations without any human intervention. This is the first documented instance of such prolonged, continuous operation of industrial humanoid robots in a real-world factory environment, as opposed to a laboratory setting.
By the Numbers: 90.2% Efficiency with a 76-Second Cycle per Car
According to data published by the South China Morning Post, the tests were conducted on an active assembly line. The robots were tasked with installing fasteners in vehicle floors. The key performance indicator — the success rate of simultaneous dual-sided installation — reached 90.2% within a production cycle of 76 seconds per vehicle.
The company noted that the primary challenges included the need for precision alignment with locating pins, handling varying nut slot designs, and compensating for magnetic fields that could affect grip stability. Beyond fastening, the robots also successfully performed auxiliary tasks: removing protective films, attaching branded emblems, and moving containers of materials.
Technological Foundation: Tactile Vision and Multimodal Models
From an engineering standpoint, the control architecture is of greatest interest. The robots operate based on a tactile fine-tuning model called TacRefineNet. A key feature is the execution of operations based on tactile feedback, without relying on machine vision or 3D object models. This represents a fundamentally different approach to ensuring assembly precision.
Concurrently, a vision-language-action (VLA) system is employed to interpret production tasks and generate action sequences. Essentially, the robot doesn’t just execute a pre-programmed routine; it “understands” the context of the task and adapts its actions accordingly, which is critical for handling non-standard situations.
Strategic Outlook: Five Years to Industrial Deployment
Commenting on the test results, Lei Jun emphasized that “this is the first step towards the stable application of Xiaomi’s humanoid robots in the field of intelligent manufacturing.” In the medium term, over the next five years, the company plans to deploy a significant number of these robots in its own factories. While specific quantities and investment timelines remain undisclosed, the strategic direction is clear: a transition from laboratory demonstrations to industrial-scale implementation.
It’s worth noting that this announcement comes amid a broader surge of activity in China’s humanoid robotics sector. Industry sources report that just this week, at least two specialized companies announced the completion of major funding rounds: Galbot secured approximately $363 million*, and Noetix Robotics raised about $145 million* with participation from an entity affiliated with battery giant Contemporary Amperex Technology Ltd. (CATL).
Xiaomi Robots Assembling Cars – Video
The video below showcases how Xiaomi’s humanoid robots are assembling vehicles of the same brand. We present it for your viewing.
As you can see, they work without coffee breaks, discussions about yesterday’s game, or any pauses at all.
Assessment and Conclusions
For the professional community, the test results mark a significant milestone in the evolution of manufacturing technology. The 90.2% success rate during prolonged autonomous operation indicates that humanoid robotics are reaching the threshold of industrial viability. While fully unmanned factories are still a distant goal, it is precisely these integration projects on active lines that allow for data accumulation and algorithmic refinement.

For businesses focused on high-tech industries and international cooperation, this is a clear signal to closely monitor the development of autonomous assembly technologies. Xiaomi’s experience demonstrates that hardware and artificial intelligence are now capable of taking on functions requiring high precision and tactile sensitivity, which will inevitably reshape the structure of added value in the industry in the near future.
*Note: For calculation convenience, approximate exchange rates as of March 6, 2026, have been used: 1 USD = 43.81 UAH, 0.79 GBP, 1.44 CAD. All amounts in other currencies are approximate and provided for reference only. For current rates, please refer to official sources or use this currency converter.
