Crowd Wisdom Consulting: Accelerating the Landing of Intelligent Driving, It is expected that the global market share of high computing power chips will increase to 57% by 2027.

Group Wisdom Consulting predicts that the landing of Level 3 automated driving between 2025-2026 will drive large-scale computing chips to become the market mainstream. At the same time, the demand for computing power for Level 4/L5 will further raise the ceiling of chip performance. It is estimated that by 2024, the global intelligent driving large computing chip (>100TOPS) market share will be about 15%, with the acceleration of Level 3 intelligent driving in the next three years, this data is expected to increase to 57% by 2027, and the market share of ultra-large computing chips (>500TOPS) will also reach 4%. Chip computing leap, urban NOA function driving smart driving chip computing power index growth As the level of automated driving transitions from L2 to L3 and above, the demand for computing power for intelligent driving chips is showing exponential growth. L2 level usually requires less than 10TOPS of computing power, while L3 and above levels require 100TOPS or even 1000TOPS of computing power. Representing current large-scale computing chips, NVIDIA Corporation's OrinX chip (254TOPS) and Thor chip (up to 2000TOPS) are widely used in mainstream models. And Tesla, Inc.'s FSD chip and the MDC smart driving platform composed of Huawei's Hisilicon Ascend chip through software and hardware optimization, still perform well in practical applications. Chip process upgrade, integration and modularization as the iteration direction of smart driving chips With the significant increase in the demand for computing power for intelligent driving chips, the upgrade of chip processes is the key to improving computing power and energy efficiency. The more advanced the process, the higher the transistor density of the chip, the lower the power consumption and the stronger the performance. In the past two years, the 7nm process was the mainstream of intelligent driving chips, such as NVIDIA Corporation's OrinX and Horizon Journey 5. With the landing of L3, the demand for computing power for intelligent driving chips is getting higher and higher, and the 5nm and below processes will become the standard for the next generation of intelligent driving chips, especially in L4 and above level automatic driving applications, such as NVIDIA Corporation's Thor series chips. However, the risks of advanced process outsourcing (such as political risks of GEO Group Inc) may become a factor restricting the development of domestically developed smart driving chips. In addition, since the intelligent driving system involves multiple modules such as perception, decision-making, and control, the traditional solution uses multiple chips to achieve, leading to increased system complexity and cost. The high integration of smart driving chips allows a single chip to support perception, decision-making, and control functions simultaneously. In the future, as computing power continues to iterate from large computing power to ultra-large computing power, it is expected that intelligent driving chips will further develop towards higher integration and modular design directions, while modular design will also become a trend to meet the needs of different levels of automated driving. According to Group Wisdom Consulting data and forecasts, by 2024, the global market share of intelligent driving advanced process chips (7nm) is about 58%, and it is expected that as chip iterations towards integration and modularization accelerate, the 5nm process for automotive-grade smart driving chips will become popular by 2026, at which time the market share of smart driving chips using advanced processes will exceed 75%. Cabin integration and energy efficiency improvement are core competitive factors for chip manufacturers Looking forward, the integration of intelligent driving and intelligent cabin functions is an important trend for future intelligent vehicles, and for chip manufacturers, they need to provide high-performance, low-power integrated solutions. As the mainstream architecture of the next generation of intelligent vehicles, the cabin integration of chips can greatly reduce system complexity and costs to achieve the popularization of high-end intelligent driving, such as the current Qualcomm Snapdragon Ride platform and NVIDIA Corporation's Thor chip can be used to layout integrated solutions for cabins to support the fusion of intelligent cabins and automated driving functions. In addition, higher computing power intelligent driving chips often come with high power consumption, and optimizing energy efficiency is key to improving chip competitiveness, such as Tesla, Inc.'s FSD chip achieving high energy efficiency through architectural optimization, and Horizon's Journey series of chips also performing well in energy efficiency. Group Wisdom Consulting believes that in the future, chip manufacturers will need to improve energy efficiency through architectural innovation and process upgrades while increasing computing power, to meet the demand for low power consumption in intelligent vehicles.