Published on Dec 31, 2025

The intelligent driving industry in China is undergoing a reshuffle. It was during the second half of 2024 that Li Shufu, the founder of Geely, first realised that a profound change was under way. Embodied intelligence had started to enter a period of rapid growth.

The industry, still young, had already reached an important inflection point. AI was beginning to change everything. Not long after this realisation, Li started to reshape the Geely Holding Group.

Qianli Technology in Chongqing was established in March 2025 as the main vehicle to drive the integration of Geely's smart driving business. AI was to be elevated from an add-on to the core building block of Geely’s current and future research and development.

At the end of the group-wide restructuring, which has just been announced, more than 1,200 intelligent-driving experts had been transferred from Zeekr's intelligent driving team, around 500 transferred from another autonomous-driving company called Maichi Zhixing, and a considerable number brought over from the Geely Research Institute.

The idea is to unify this brainpower so that isolated teams no longer unknowingly reinvent the wheel but instead pool their resources. Crucially, the Geely founder appointed Yin Qi as CEO of AFARI. Yin is a former co-founder of the Chinese AI start-up Megvii.

Megvii became widely known when its facial-recognition algorithms were reported to partially outperform those of Facebook. Li Shufu expresses his highest respect whenever he talks about Yin Qi. "We should have met earlier", Li said publicly during a product launch about the AI wizard.

The technological route that the new bright talents at AFARI are now implementing for Geely is centred around the concept of model density. It means that AI models are taking on a much bigger role in the intelligent driving architecture than before, and will reduce the role of rule-based logic and high-definition maps. AI, in short, will be expected to make decisions "more like a human".

Geely’s researchers developed a VLA model (short for „vision, language, action") which combines perception via cameras, radar and lidar with semantic and conceptual understanding. It can interpret traffic scenarios and the spoken intent of the driver and execute "end-to-end" decisions about steering, braking or accelerating.

For example, when annoyed by a slow-moving vehicle, the driver can simply say "Overtake that car in front." AI in the cockpit interprets this, the VLA large model checks the surroundings, and the intelligent driving system of the car automatically performs a controlled overtaking maneuvre.

"At Qianli Technology, artificial intelligence is positioned as the primary system layer of autonomous driving, rather than a supporting feature", writes IT Home (in Chinese) in its report about the recent business unit consolidation around AFARI. "Perception, planning, and control are increasingly unified under end-to-end or semi-end-to-end AI models."

It so happens that the same underlying capabilities of VLA models in the fields of perception, planning and control are - with some specialized training - not only useful for autonomous driving, but also great for controlling the motions of industrial robots and even humanoids.

Li Shufu has always been an ecosystem builder who was never afraid to dream big. He became the first Chinese entrepreneur to acquire a major Western car manufacturer when he paid USD 1.8 billion for Volvo in 2010. In 2013 he bought 9.69 percent of the shares of Daimler, the parent company of Mercedes-Benz.Through the deep integration of AI technologies into the automotive industry, embodied intelligence is becoming a central driver of the further development of intelligent vehicles.

Now, he is sensing new opportunities brought by the convergence of automobiles and robots. The car of the future, he and his team essentially believe, will be another form of embodied artificial intelligence. AI will take over the intelligent cockpit, enable more and more natural interactions with drivers. It will be in charge of many decisions for autonomous driving.

“Geely has the potential to transform from an automaker into China’s—and even the world’s—largest robotics company,” said Li Chuanhai, Vice President of the Geely Automobile Group and President of the Geely Automobile Research Institute, during the robotics conference WAIC 2025 in Shanghai.

Technologically, embodied intelligence and autonomous driving indeed share the same foundations: front-end perception, back-end cognition and reasoning, and a brain capable of autonomous motion planning and control.

To pursue his vision, Li Shufu is partnering not only with the best in the world of automotive manufacturing like Mercedes-Benz, but also with robotics manufacturers like UBTECH.

Geely has cooperated with UBTECH on deploying a customised version of the humanoid robot Walker S1, which has started to work at Zeekr's car factory.

There is a feedback loop forming which has the potential to fundamentally reshape the automotive industry in the coming decades. AI models developed for autonomous driving can also power cobots and humanoids, increasing productivity and reducing labour costs in the car factories.

For Geely, which has set an annual sales target of three million vehicles for 2025, such productivity gains could offset early investments into AI and robotics companies in the future.

Geely's car factories, at the same time, have started to serve as early deployment scenarios for industrial robots, providing the initial scaling necessary for commercial robotics businesses, while also delivering the big data needed for their further refinement.

It is, of course, a simplification to say that cars will morph into "robots on wheels" in the future. The relationship between AI, robots and the automobile is more complex. What is undeniable, however, is that a wide-ranging convergence has begun.

"The trend happening in the automotive industry is that intelligent driving technology is becoming an AI root technology,” said Chen Qi, Geely's Chief Intelligent Driving Scientist, during this year's World Artificial Intelligence Conference (WAIC 2025) in Shanghai.