➤ Deep Dive
_________
The Quiet and Fast Rise of China’s Industrial AI Agents
Too Busy for Doomsday Scenarios, China’s AI Agents Report for Work
Published on Sep. 28, 2026

U.S. President Donald Trump and his Chinese counterpart Xi Jinping met in Washington this week, and artificial intelligence was on the agenda. Trump, fixated on rivalry with China, was as blunt as always. “Whoever wins AI, WINS!” the U.S. president declared.
Chinese President Xi Jinping, not comfortable with speaking freely, opted to release a statement shortly after his arrival at Joint Base Andrews on September 23. His words were more diplomatic. The two countries “should be partners rather than rivals,” Xi argued. Yet he is answering the hostility Washington throws at his country with firm countermeasures.
While both superpowers are entangled in an AI race that is increasingly reminiscent of the Cold War, European industry has to consider its best interest. This is a good moment to remember that the same partnership offer that Xi has just extended to Trump is availabe to European companies doing business in China.
Leaving ideology aside and looking closely at what is happening on Chinese factory floors, you will find that Chinese robot makers and data center operators are welcoming foreign partners, many of whom come from Europe.
Siemens with factory software such as its digital twins, Schaeffler with state-of-the-art bearings and motors and ABB with drives and controls for cooling towers and pumps are all part of the AI-driven industrial transformation in China.
The cooperation starts with building the factories that China needs. In Liuzhou, in the southern region of Guangxi, Siemens Digital Industries Software and the Chinese robot maker UBTech have jointly deployed a digital “smart brain” that powers the production of humanoid robots.
While humanoids deployed in the warehouse and carrying parts to the production line made headlines for UBTech in Liuzhou, the digital twin developed together with Siemens is the more consequential use of technology in this case. AI meets mature factory simulation to power the commercialization of physical AI.
The manufacturing operations management system they developed closes a digital loop across work-order scheduling, material distribution, and quality monitoring. Each robot carries a unique serial number for one-item-one-code traceability, according to UBTech.
While these manufacturing details may be less exciting for the media and the general public than “robots making robots”or the ambitious production capacity of 10,000 Walker S and Cruzr series humanoid robots a year that the Chinese robot maker is building, they are what moves the needle for productivity gains.
Schaeffler, the huge family-owned supplier of bearings, electric drives and chassis systems based in the Bavarian town of Herzogenaurach, chose the components route. In March, the German automotive supplier, which has cleverly rebranded itself as a “Motion Technology Company”, entered a strategic partnership with Leju Robotics, another up-and-coming robotics maker in China.
It is Schaeffler’s first strategic partnership with a Chinese humanoid maker, and it couldn't be more timely. As the automobile and robotics industries are converging and both are getting a boost from AI, this is an opportunity spotted and put into action. This is the best antidote to German “AI angst” and doomsday scenarios.
Schaeffler and Leju Robotics have teamed up to open an independent humanoid company in China. Schaeffler brings eight product families to the table, from rolling bearings, screw drives and precision gears to motors, sensors, thermal management and battery management, which covers what a humanoid needs for actuators, sensing, motion control and energy management.
The Chinese and German partners want to put humanoids to work in factory inspection, equipment operation support, logistics and human-robot collaboration, and to collect the data for model iteration along the way.
“With Leju Robotics, we are gaining a strong, innovative partner in one of the most dynamic growth markets for humanoids,” said CEO Klaus Rosenfeld. By 2035, Schaeffler plans to integrate a mid-four-digit number of humanoids into its own production.
The empowerment of industrial vision technologies needed for the deployment of AI in factories with the help of top hardware “made in Germany” is as another good example. Foxconn, the electronics manufacturer from Taiwan with huge production in mainland China, uses Zeiss imaging systems to spot defects at the micrometer level.
This process is now fully automated, according to Shi Zhe, Foxconn Group's chief digital officer. He highlighted over twenty years of collaboration between the two companies. Shi noted that Zeiss has propelled the joint progress of “AI + optical inspection”.
Zeiss from Oberkochen in Germany, which calls itself “an internationally leading technology enterprise operating in the fields of optics and optoelectronics,” has expanded from an optical equipment supplier into a technology enabler and co-developer of standards, Shi said. Together, he continued, the partners are shifting manufacturing from experience-driven to intelligence-driven.
This makes a lot of sense. A vision model's effectiveness depends on its training images and those on the production line. Zeiss provides optics and resolution, while Chinese engineers develop models, data pipelines, and production scale.
For Zeiss, this partnership transforms a product business into a share of the customer's process. Kechuangban Ribao summed up Zeiss’s China strategy as a shift from “selling products” to “building an ecosystem.”
Since 2021, China has continuously been the largest market for Zeiss globally. Despite challenges like centralized procurement policies and an ever-worsening geopolitical environment, Zeiss continues investing in China. It has two hubs in the Yangtze River Delta and the Greater Bay Area, with a Greater China headquarters under construction to make China a leading “manufacturing and innovation center” for Zeiss.
Martin Fischer, president and CEO of Zeiss Greater China, said at an event in China marking the company’s 180th anniversary that AI will speed up Zeiss’s transformation. The same can be said about manufacturing in China, and that is why this partnership is smart and its chances of success outweigh the risks.
Under its new five-year strategy, Zeiss plans to build AI deeply into its products and operations. Fischer emphasized that localization involves deep integration into China’s industrial ecosystem. Rather than simply adapting to China, Zeiss is drawing on the country’s technology wave to better cope with future uncertainty, Kechuangban Ribao commented.
The industrial revolution thanks to AI is more of an “electricity rush” than a gold rush, however. This is where the figures, and the opportunities, get really large. By the end of 2025, China had built 42 computing clusters with 10,000 or more GPUs each, with computing centers consuming 170 billion kilowatt-hours that year, 1.6 percent of the national total, according to the National Energy Administration.
By 2030, data centers are expected to draw 800 billion kilowatt-hours, around 6 percent of all electricity consumed in China. ABB presented these numbers at Alibaba Cloud’s Apsara Conference in Hangzhou this week, and that was no coincidence.
China's central planners are following a strategy named “suandian xietong” (算电协同), computing-power coordination. For data centers, you need energy. Lots of it. Because conventional low-voltage distribution loses too much energy, Chinese data centers are shifting to 800-volt or even 1,500-volt direct current.
This is where ABB, the Swiss electrification and automation group, is welcome with its advanced hardware and solutions. In Hangzhou, ABB showcased an all-DC 800-volt distribution solution and SACE Infinitus, a solid-state circuit breaker certified under IEC 60947-2 that clears short circuits 1,000 times faster than a mechanical breaker and lasts 100 times longer, according to the company.
Cooling is the second-largest electricity consumer in a data center. ABB’s ultra-low-harmonic drives for chillers, cooling towers, and pumps keep current distortion below 3 percent without external filters. Its AC500 controller integrates cooling, battery storage, building management, and backup power into one system that complies with the IEC 62443 cybersecurity standard.
A single rack in a large computing center can draw more than 100 kilowatts. Someone must deliver that power and remove that heat.
More and more, AI is also used to optimize the data centers providing AI computing power. AI agents coordinate photovoltaic and storage systems so that “electricity follows computing demand, while computing demand adjusts to electricity supply,” ABB explained on its Chinese-language website. (in Chinese)
The hardware and software enabling the Chinese concept of computing-power coordination is exactly the kind of products and services where ABB excels. Someone must deliver that power and remove that heat.
And China is only just getting started. To meet both its climate mitigation targets and the energy needs of its data centers, Beijing is promoting new AI computing facilities that are directly supplied with green energy generated with solar and wind.
The Chinese energy group Envision recently began operations of its “Galaxy Campus” in Ulanqab, in the northern Chinese region of Inner Mongolia. It is currently the world's largest single AI computing facility. It has a planned capacity of 2 gigawatts and an exceptionally high share of directly supplied green electricity, reported the energy portal Beijixing Dianli Wang on September 23, calling it an “aircraft carrier” of computing-power and electricity coordinatioxn. (in Chinese)
Hu Xingzan, the group’s chief technology officer for AI data centers, envisions such campuses running entirely on directly sourced renewable power and their own power systems. “The pace of grid construction is actually out of sync with the speed at which AIDC is developing today,” he said, using the industry’s shorthand for AI data centers.
Wind and solar fluctuate on one side, computing loads on the other. There is not only demand for advanced AI models and agents, but also for a lot of high technology from Germany, Switzerland or Sweden.
The Chinese-foreign partnerships formed now are not only for places like Guangxi or Inner Mongolia. They have the potential to travel the world, accompanying Chinese technology companies during their expansion abroad.
Alibaba Cloud will open a data center in the Netherlands in October, with Finland and Turkey to follow, said Li Feifei, president of its international business, at the same conference in Hangzhou. Those sites will need power distribution, cooling, and inspection too.
Washington and Beijing will continue racing and fighting. Europe’s industrial companies should not get distracted, though. They are in Liuzhou, in Hangzhou, and in Ulanqab, and they will be in the Netherlands when Alibaba opens its doors. Let the two presidents talk. Someone has to sell the shovels for the new gold rush.
