The rise of independence is unstoppable, and intelligence will become an important winner and loser

High-end models have begun an intelligent arms race, and intelligent acceleration is also expected to shorten the vehicle replacement cycle. With the continuous improvement of the level of autonomous driving, the requirements for AI chip computing power of cars are also getting higher and higher. According to the analysis of automatic driving computing power demand proposed by NVIDIA, the chip computing power demand of L0-L2 level automatic driving is below 10 TOPS, and the chip computing power demand of L3 level automatic driving is between 30 and 100 TOPS. L4/L5 level automatic driving requires chip computing power requirements above 100/1000 TOPS, respectively. On the one hand, with the improvement of automatic driving level, the number of sensors (such as cameras, millimeter-wave radar and lidar) is more and more, and AI chips need to process more and more data; On the other hand, the number of scenarios and decisions processed by AI chips have increased geometrically. According to the cost of the automated driving system (including sensors and domain controllers), the cost of the L0/L1 level assisted driving system is between $275 and $325, and the cost of the L2/L2+ level automated driving system is between $450 and $550 and $750 and $1,200. The cost of an autonomous driving system at level 3 is more than $4,000.

Will the replacement logic of smart phones play out in the era of smart cars? We may have underestimated the demand for car swaps. Compared to the development of the smartphone era in the past 10 years, the iPhone will have a new hardware upgrade after the new flagship phone is launched every year, and the latest iOS system is often not all forward compatible due to high requirements for hardware configuration. To some extent, this also leads to the phenomenon that the replacement cycle of smart phones will be shortened when the technology has made significant progress (for example, 5G has shortened the replacement cycle of smart phones). With the improvement of smart car configuration and computing power, as well as the gradual evolution of intelligent driving and intelligent cockpit, we expect that smart cars are also expected to usher in a shorter replacement cycle in the future.

The importance of "intelligence" on the demand side is increasing, and it will eventually usher in "jump". A car is a product with many needs. Traditional needs include handling, power, price, appearance, space, ride comfort, etc. In the age of intelligence, people's energy invested in "driving" is released, the need for control, power decline, and the requirements for assisted driving/autonomous driving and cockpit intelligence rise. Although the functions of L2.5+ assisted driving products provided by manufacturers are mainly concentrated in high-speed closed roads and automatic parking scenes, so it is not "just needed" for consumers, but we believe that with the development and maturity of this technology, reliability and safety improvement will bring qualitative changes to the car experience. The importance of autonomous driving on the demand side will "jump." Once the functions of autonomous driving and smart cabins are greatly improved, consumer demand is also expected to jump. The current average replacement cycle for passenger cars in China is 14 years, and intelligence may shorten this replacement cycle at some point in the future.

Why will the smart car pattern appear first decentralized and then centralized trend? The redefinition of automotive products by intelligent electric vehicles is a challenge to the original R&D organization form of traditional automobile companies, which requires close cooperation between software, hardware, mechanical engineers and product planning to jointly complete product definition and R&D design. Companies with stronger full-stack R&D capabilities and more flexible R&D organizational forms are expected to define better products and achieve higher sales and market share. Larger vehicle sales and longer driving range will also generate more data to optimize autonomous driving algorithms, helping Oems achieve a better experience on autonomous driving. Therefore, from the perspective of the final idea, the concentration of intelligent electric vehicles must ultimately be higher than that of traditional fuel vehicles.

However, at present, it is still in the "early spring and Autumn" of intelligent electric vehicles, and there are a large number of people who have not entered the game. In addition, the production capacity investment, production, and production models in the automotive industry exit cycle is much longer than consumer electronics, in recent years, Changan Peugeot Citroen (DS), Dongfeng Renault and Borgward Automobile, for example, production capacity and brand exit cycle is about 6-8 years. Under the wave of smart electrification, we see three forces converging in the automotive industry, which is expected to usher in a period of intensive new brands and new capacity increase in the next five years, and there may be a crowded scene in 2021-2025: