2022 will be the first year of the "arms race" of intelligent driving computing power, with L3 penetration turning upward. We sorted out the top ten data of new car intelligent driving computing power in 2022 as follows. With the continuous boarding of lidar, the increase of the number of point clouds, and the increase of the number and accuracy of cameras and ultrasonic radar, the redundancy and priority of large computing power have become the consensus of the main engine factory, so as to improve the depth and accuracy of vehicle end neural network algorithm and bring the accuracy of real-time driving scene data recognition L3 / L3 + improvement of landing certainty of high-order autopilot. We believe that 2022 will enter the first year of the "arms race" of intelligent vehicle computing power, provide fertile hardware soil for vehicle intelligence, and accelerate the upward turning point of L3 and L3 + intelligent driving penetration in the automotive industry.
Qualcomm pushed the 5nmsoc chip into the automotive industry, and the industrial iteration continued to accelerate. At CES in early 2020, Qualcomm announced that it would launch snapdragonride's complete autopilot computing platform to confront chip giants such as NVIDIA, so as to further extend its business territory from mobile phone to cockpit and intelligent driving terminal. On January 26, 2021, Qualcomm released SOC and accelerator chips with 5nm process technology and snapdragonride platform with over 700tops computing power in an online activity with the theme of "redefining cars", with power consumption ranging from 5W to 130W. It is expected to ship in Q2 of 2022. China Greatwall Technology Group Co.Ltd(000066) automobile has reached cooperation with Qualcomm, and wey in 2022q2 will become the first mass production vehicle equipped with ride platform. We believe that the first entry of 5nm process into automobile means that the development of intelligent automobile has kept up with the iterative pace and development context of consumer electronics, and the acceleration signal of industrial development is very obvious.
NVIDIA Orin continues to promote the multiplication of onboard SOC computing power, and the dilemma of "prisoner game" between OEMs will become more and more obvious. NVIDIA is the computing power overlord of intelligent driving chip track. From 2018 to the end of 2021, NVIDIA launched a total of four generation autopilot chips, which has been doubling in performance. In 2023, the fourth generation of 1000TOPS Atlan sample was launched. It can be compatible with the software stack based on Xavier and Orin. No need to redesign the software, it can make use of the new Soc to achieve a substantial improvement in vehicle performance. In the future, it is expected to further promote the acceleration of intelligent vehicle industry. According to the public information of each main engine factory, at present, the main engine factories that clearly adopt Orin chip include ideal automobile, SAIC Zhiji, Weima automobile, SAIC R, Volvo, lutes, Gaohe automobile, Jidu automobile, Xiaoma Zhixing, etc. We believe that with the off-line mass production of the corresponding models of the above main engine manufacturers in 2022, the dilemma of "prisoner game" between main engine manufacturers will become more obvious, hardware redundancy and computing power will be further strengthened, and manufacturers related to the industrial chain will see a significant acceleration of performance.
Investment suggestion: we maintain the "buy" rating of Huizhou Desay Sv Automotive Co.Ltd(002920) , Thunder Software Technology Co.Ltd(300496) , Navinfo Co.Ltd(002405) . We suggest paying attention to Guangting information and Jingwei Hengrun (IPO).
Risk tip: the risk that the penetration rate of automobile intelligence is lower than expected; Risks of economic downturn; Supply chain management risk.
