Qingzhou Raises Another $100 Million in Funding with Joint Investment from a Northern Automobile Manufacturer, Focusing on World Models | Exclusive Report by 36Kr

The picture shows QCraft.

Text | Xiao Man

Editor | Li Qin

36Kr learned that QCraft (hereinafter referred to as "QCraft") announced the completion of a new round of Series D financing of $100 million. This round of financing was jointly invested by a leading domestic automaker, Ningbo Ninghai Xingtaihe Fund, Liangxi Science and Technology Innovation Industry Mother Fund (managed by Bohua Capital), Huade Science and Technology Innovation, a leading automotive electronics component company and other major industrial investors and investment institutions.

The funds from this round will be mainly used in two directions: First, to increase R & D investment in cutting - edge physical AI technologies such as "world model + reinforcement learning"; Second, to strengthen organizational talent building and reserve strategic strength for the company's long - term development.

QCraft was founded in early 2019. It is an autonomous driving company in China that has always adhered to L4 - level autonomous driving and intelligent assisted driving for passenger cars. Its core team members come from international technology companies such as Google Waymo, Tesla, NVIDIA, and Facebook, and R & D personnel account for up to 80%.

QCraft's strategic focus shifts towards L4 and general physical AI

"Autonomous driving is the best entry point to physical AI," said Dr. Yu Qian, co - founder, chairman and CEO of QCraft.

In QCraft's view, when autonomous driving truly becomes a reality, it means that its value is not limited to an intelligent system that can only drive a car, but will evolve into a general intelligence for the entire physical world.

QCraft once divided the intelligent evolution of autonomous driving technology into three stages:

The first stage is the "machine intelligence" stage mainly based on the modular architecture, using the "perception - positioning - planning - decision - making - control" modules. Although the perception ability has been greatly improved, the system generalization ability is weak.

The second stage is the "human - like intelligence" stage with the end - to - end architecture as the core, emphasizing data - driven as the core. It gradually models perception, planning and decision - making to form a one - stage end - to - end structure, and the driving performance is approaching that of humans.

Finally, as the end - to - end architecture becomes the industry consensus, the introduction of the VLA model and the world model enables the system to deeply understand physical laws, human intentions and empirical knowledge. Driven by both real and generated data, the safety can reach ten times that of human drivers.

QCraft believes that 2026 will be the starting point of a new ten - year golden cycle for autonomous driving in the future. QCraft will significantly shift its strategic focus towards L4 autonomous driving and general physical AI and accelerate its overseas layout. At the same time, it will increase R & D on world model + reinforcement learning technology, drive the future with technology, and firmly move towards general physical AI.

QCraft's product evolution route is also becoming clearer. QCraft plans to promote the penetration of full - scenario urban NOA into national models priced at 100,000 yuan in 2026; promote the initial large - scale deployment of Robovan and small - scale pilot of Robotaxi. In 2027, it will promote the mass production of NOA on more than 3 million vehicles, promote the mass production of Robovan at the level of 100,000 units, and achieve large - scale deployment of Robotaxi.

QCraft has achieved one million vehicle deliveries

In the past seven years, QCraft has always adhered to the development strategy of "parallel development of L2++ and L4".

Since the layout of L2++ at the end of 2021, the installation volume of QCraft's "Chengfeng" intelligent assisted driving system has exceeded one million units. Data shows that it took nearly three years for QCraft to complete the first 500,000 unit deliveries, while it only took eight months to cross from 500,000 to one million units, with an obvious growth rate.

Currently, QCraft has reached cooperation with nearly 10 leading automakers such as Li Auto, Chery, GAC, Geely, and SAIC, covering nearly 30 models. It is expected that by 2026, the number of cooperative models will exceed 50, and the urban NOA function will be basically standard - equipped.

In the L4 field, QCraft officially entered the unmanned logistics track through the new paradigm of "mass production means operation", and relevant vehicles have been put into commercial operation in Jinhua, Wuhu, Ningbo and other places.

QCraft believes that L2 and L4 are product logics, and the underlying technical logics are highly consistent. A large - scale data closed - loop system driven by both L2 and L4 is the optimal solution to autonomous driving.

Therefore, the entire architecture system from technology to products of QCraft shares the same technical foundation - an "autonomous driving super factory" driven by AI.

The advantage of doing this is that the mass - production data of L2++ can be directly used for the training of L4 products, and vice versa, maximizing the value of data. In addition, based on a unified technical foundation, different products can have the same driving ability from the same source, and there is no need to repeatedly develop for different product forms, thus accelerating technological iteration.

Moreover, QCraft has carried out the design and R & D of a new - generation model and integrated the advantages of the VLA and the world model into the same end - to - end architecture.

With the injection of $100 million in new financing and the accumulation of one million units of mass - production data, QCraft is planning to try to push autonomous driving into a broader general physical AI field through technological reconstruction.