Geely climbs to the AI high ground

Author | Chai Xuchen

Editor | Wang Xiaojuan

Established car manufacturers are also vying for the jobs of new forces and tech giants.

On January 7th, Eastern Time, Geely "showed its sword" at the 2025 International Consumer Electronics Show (CES), launching the industry's first "full-domain AI" technology system for smart cars.

It is understood that this technology system is built on an AI-native operating system, covering both C-end and B-end as well as an integrated smart ecosystem of air, land, and space, aiming to create a native "smart mobility life form" that is warm, emotional, and capable of sustainable growth.

"Full-domain AI" and "smart life form" mean that Geely has extended its ambitions to the ultimate stage of intelligence.

Coincidentally, the most aggressive new car-making force, Li Auto, recently proposed transforming into an AI technology company. Chairman Li Xiang pointed out that "the intelligent era speaks of true AI, and cars will evolve from transportation tools of the industrial era to spatial robots of the AI era."

For Geely, this is not a momentary trend; the layout of AI is by no means a castle in the air. The end-to-end voice large model, AI digital chassis, FlymeAuto intelligent cockpit, and intelligent driving developed based on the aforementioned technology system will soon be applied to new models under its brand.

In the mind of Li Chuanhai, the head of Geely's Automotive Research Institute, Geely's AI strategy has a clear three-step plan.

The first phase is the intelligent assistant, which can assist the car in completing tasks and commands, achieving precise responses; the second phase is the intelligent agent, which, based on multi-modal perception, allows the car to understand the environment and autonomously complete tasks, capable of handling more complex situations like autonomous parking and maneuvering.

The final phase is the life form. "Smart cars can perceive needs, autonomously plan tasks, reason, and execute, while possessing awareness and even emotions," Li Chuanhai stated. At this point, the car becomes a warm, emotional, and sustainably growing life form that can integrate into people's daily lives. The era of "silicon-carbon symbiosis" will eventually arrive.

Currently, Geely's AI technology has already been implemented in the power sector. In October last year, Geely launched the new Thor EM-i super electric hybrid system, which uses AI algorithms for dynamic and precise real-time control to achieve ultra-low fuel consumption; in March last year, Geely released the world's first unmanned drifting technology, stabilizing the vehicle through AI during motion.

This is just the tip of the iceberg. Chen Yong, chief AI scientist at Geely Automotive Research Institute, stated that Geely will subsequently integrate AI into every research and development link and every user experience scenario.

However, the layout of AI is not achieved overnight; it is a long-term marathon.

Chen Yong revealed that Geely is currently developing a "full-cause intention understanding large model," which can integrate all factor information such as vehicle status, vision, and voice to achieve a comprehensive understanding of user needs. "For example, on a rainy day, it can comprehensively judge whether to help me park and how to park, which preliminarily realizes the second phase, truly 'thinking' like a person about the user's potential needs and reasoning and executing." As for reaching the third stage, three conditions must be met: computing power, data, and algorithms. Li Chuanhai stated that these three conditions are the technological foundation for AI development and the basic capabilities for Geely to build "automotive full-domain AI" and create a "mobility life form."

In the field of computing power, Geely has established the Geely Star Smart Computing Center, with cloud computing power iterating to 10.2 billion times/second, capable of supporting the smart computing needs of 5 million online vehicles. "We predicted early on that large models would have a huge demand for training resources, so we started testing and using domestic GPUs at the beginning of last year," said Li Chuanhai.

On the other hand, the emergence of large models is inseparable from data; AI requires high-quality data "feeding" for continuous iteration.

Chen Yong revealed that in 2021, Geely introduced technologies such as geometric prior knowledge and 3D Gaussian splashing to generate various scenarios, such as rainy and snowy weather, congested road conditions, and multi-story parking garages, achieving infinite generalization of traffic scenarios. In addition, Geely also has a style transfer model based on AIGC technology to ensure that synthetic data closely resembles reality, meeting the requirements for algorithm training.

Finally, in the fields of developing multi-modal large models, full-cause intent understanding large models, end-to-end large models, edge large models, and hybrid cloud accelerated inference frameworks, Geely is also deeply cultivating research in basic technologies.

Clearly, a few years ago, Geely lagged behind in the wave of electrification, but after catching up and re-entering the TOP sequence, it is now determined to seize the wave of AI transformation. Three years ago, Geely held the crown of "self-owned No. 1" for five years, which was taken away by BYD led by Wang Chuanfu. Unwilling to accept this, Li Shufu vowed to make a comeback.

On January 5, Geely Group announced its strategic goals for the next three years—by 2027, total vehicle sales will exceed 5 million units.

If successful, Geely is expected to enter the TOP 5 of the global automotive group sales rankings by 2027, holding a significant position in the global automotive industry. The efficiency and market transformation brought by AI are playing a key role in Geely's rise