Geely’s automotive empire has taken a decisive step to stitch artificial intelligence directly into its cars. Yin Qi, already chairman of Geely’s smart‑driving arm Qianli Technology, has been named chairman of big‑model start‑up JiaYue Xingchen while the latter closed a B+ financing round of more than ¥5 billion — a single‑round record in China’s large‑model space over the past year.
The appointments and the fundraising are not isolated moves but part of a deliberate strategy: to pair a proprietary “big‑model” brain with Geely’s in‑house driving and cockpit hardware. Qianli provides the vehicle terminal — the sensors, actuators and production‑grade software — while JiaYue supplies the foundational models, multimodal perception and agent layers. Company statements say the two will co‑develop world models, VLA (very large action) systems and multimodal stacks to deliver high‑“model content” intelligent driving and cockpit solutions.
Yin Qi’s dual roles formalise work that began privately: he helped found JiaYue Xingchen and since late 2024 has reorganised Geely’s smart‑driving assets, folding Zeekr’s autonomous team, Geely Research Institute’s driving group and Megvii‑linked MaiChi Zhixing into Qianli. JiaYue’s leadership team — CEO Jiang Daxin, chief scientist Zhang Xiangyu and CTO Zhu Yibo — will now work with Yin to accelerate base‑model research and roll the outputs into production vehicles.
The partnership is already past proof‑of‑concept. In 2025 the three parties previewed an Agent‑native cockpit OS at the World AI Conference, combining JiaYue’s multi‑modal models and end‑to‑end speech stacks with Qianli’s cabin software. That Agent OS is slated to reach production in Geely’s Galaxy M9. At CES 2026 the partners announced a new assisted‑driving brand, G‑ASD, underpinned by JiaYue’s models; Geely says the underlying smart‑driving system now equips 16 models across Zeekr and Lynk & Co and covers more than 300,000 vehicles.
For Geely, the attraction of owning the model stack is clear. Reliance on external cloud models limits optimisation for vehicle constraints, data flows and safety certification. Embedding proprietary models raises what Chinese industry participants call the product’s “model content” — the share of user experience and decisioning driven by in‑house AI — which becomes a strategic moat as cars compete on software and services rather than horsepower alone.
But the strategy brings risks. Developing and continuously updating large models at automotive safety levels is capital‑intensive and operationally demanding. Models must be validated across rare driving scenarios and satisfy regulators increasingly wary of handing safety‑critical tasks to opaque neural systems. There is also a war for talent and chip capacity: pure‑play cloud AI and semiconductor firms are competing with automakers for the same engineers, data and silicon.
The financing round, which market participants estimate at roughly $700m, signals investor confidence that vertical integration — cloud to car — can pay off. It also reflects a broader pattern in China: auto groups, internet giants and chip vendors are converging on a single prize, the vehicle as an AI endpoint. If Geely manages to deliver reliable, differentiated agents and cockpit experiences at scale, it could leapfrog rivals that depend on third‑party models; if it fails, the cost and reputational damage would be substantial.
Either way, the move marks an important moment in the fusion of China’s AI and auto industries. With Yin Qi straddling both organisations, Geely has put a senior executive at the intersection of model research and product deployment — a structure designed to accelerate iteration from lab prototypes to factory floors. The outcome will shape not only Geely’s competitiveness but also how Chinese automakers define the relationship between proprietary AI and automotive safety in the years ahead.