Published on Dec 17, 2025

On December 11, Xiaomi put its new home-appliance factory into operation in Wuhan, central China. The plant is fully automated. Every 6.5 seconds, a new air conditioner rolls off the line. Quality inspection is carried out ‘100 percent’ by artificial intelligence, Xiaomi Group President Lu Weibing announced on his Weibo account. (in Chinese).

This is already the third state-of-the-art factory Xiaomi has built within two years, following a smartphone plant in Changping, northwest of Beijing, which has been operating since late 2023, and an electric-vehicle factory in Yizhuang, in the southeast of the Chinese capital, where production began in the first quarter of 2024.

What all three plants have in common is that artificial intelligence controls core production processes.

In the China AI2X Briefing, we refer to this type of plant as an industrial “AI factory”. Regardless of the fact that AI plays a larger role in some factories and a somewhat smaller one in others, this terminology is intended to highlight an important milestone in the evolution of mass production.

Industrial AI factories are production facilities in which artificial intelligence centrally controls and continuously improves manufacturing, logistics, quality assurance, and process optimisation.

These plants do not merely produce smartphones, air conditioners, and electric vehicles in modern factory halls. Something more fundamental is happening.

A new era of industrial manufacturing is beginning in China. In largely human-free factory halls, AI agents supervise a small army of robots. Driverless AGVs (automated guided vehicles) rush back and forth delivering components. The articulated arms of cobots grab them, cut, mill, and screw.

In this way, more and more products are being assembled in the People’s Republic, from ballpoint pens to smartphones, from umbrellas and washing machines to complete electric vehicles. It is the birth of AI-based serial production.

The “workbench of the world” is transforming before our eyes into a laboratory of rigorous automation and digitalisation, in which AI is used to integrate and gradually improve all operational processes, and thereby reinventing the future of industrial work.

In these spotlessly clean and strikingly quiet factories, no artificial lighting is required. And yet the term “lights-out factories” does not fully capture the significance of this development. Nor does “smart factory”, where data flickers across screens at every machine, quite do it justice anymore. They can therefore be described as AI factories.

Midea, which describes itself as “the world’s largest manufacturer of home appliances”, has not only built its own AI factory but has also recognised the historical significance of this new form of production. The company has been certified in London by the World Record Certification Agency as the “world’s first factory with AI agents and multi-scenario coverage”.

As reported by Chinese technology portal 36Kr, Midea employees proudly accepted the corresponding certificate on a stage in London. This makes it official, at least symbolically. The first genuine AI factory belongs to Midea. It is located in Jingzhou, on the middle reaches of the Yangtze River, and it produces washing machines.

There may be a faint scent of public relations hovering over this announcement, like that of detergent in a laundry room, but this is by no means merely clever marketing, nor is it a version of “AI hype”.

Rather, Midea’s washing-machine factory represents a comprehensive attempt to place 38 different production processes under the control of a total of 14 AI agents. This means that production is no longer merely automated, as in a conventional smart factory, but can additionally be comprehensively monitored and autonomously improved in real time.

Among the 38 processes controlled by AI agents are logistics, including the delivery of materials using AGVs, the categorisation of each unit arriving on the flexible production line, washing machine or tumble dryer, and which model, using cameras and sensors, the activation and monitoring of designated screwing processes with superhuman precision, and quality inspection based on data transmitted to a central computer.

Humanoid robots, also controlled by the central computer and its algorithms, cooperate with AI agents responsible for DMS, TPM, EHS, and quality assurance. The AI does not merely coordinate these processes. Over time, it improves them independently. In other words, the AI factory learns autonomously.

Lei Jun, the founder and CEO of Xiaomi, described this using the example of the company’s smartphone factory in Changping. “This platform completely blew our colleagues away when they first saw it,” Lei said. “What is most impressive is that the platform can identify and solve problems while simultaneously helping to improve the production process.”

The efficiency gains achieved through the integration of artificial intelligence and automation are also striking. Midea states that the 14 AI agents in its Jingzhou washing-machine factory can complete tasks in minutes that previously took hours.

Thanks to a continuous feedback loop of machine perception, decision-making, and optimisation, “efficiency has increased by more than 80 percent on average”, Midea employees from Jingzhou said at the certification ceremony in London. Response times in production planning had been reduced “by 90 percent”.

Even if some of these figures may be slightly exaggerated, which would not be entirely surprising in China, where assertive self-promotion is clearly part of doing business, the overall trend nevertheless appears credible. Robots are fast and do not need lunch breaks, and AI calculates faster than even the most experienced production manager sitting in a control room. Even if the devil remains in the details in AI factories, and even if reports of malfunctions and setbacks emerge in the months and years ahead, it is already difficult to imagine that AI will not continue to spread across more and more factory floors.

Competition will ensure this. Those who do not invest in such modernisation of their production will, within a few years, no longer be able to compete on either price or quality.

Over the next five years, many CEOs and operators of pioneering plants of this new dimension in China believe that AI will not only transform the production of advanced high-tech goods but all traditional industries as well.

“Every industry deserves to be rebuilt with AI,” Xiaomi founder Lei Jun said recently in an interview with a Beijing daily newspaper. The new technology is simply far superior to traditional manufacturing methods, regardless of which goods are being produced in series.

Lei cited as an example the large die-cast components produced through gigacasting for the company’s electric vehicles. These parts are difficult to inspect with the naked eye. Using X-ray systems and AI-based image-processing models, however, inspection can be completed within two seconds, “ten times more efficient and more than five times more precise than manual inspection”.