"After two years of development, Jidu's first robocar concept car will be at the Beijing auto show in April this year. This joint venture between Baidu and Geely has joined the automatic driving industry represented by Weilai, Xiaopeng and Chinese express, becoming a competitor of Tesla, an American company, which shows the pace of innovation in the world's largest auto market." On February 21, the Financial Times wrote in a report. However, the British media does not want to advertise Chinese car companies or praise China's auto industry. It has a turn of the tongue: Jidu robocar may have technical risks because its "brain" comes from American company NVIDIA. Jidu automobile is not an isolated case. Like the domestic mobile phone industry, chip is also the "soft rib" of China's most self driving automobile companies. Whether it is the Wei Lu, Xiaopeng, ideal, or the two echelon members of Na Zha and real horse, all rely on the automatic driving chip provided by American enterprises such as NVIDIA, Qualcomm and Intel. So will China's rapidly developing smart car industry be stuck with key chips by the United States like Huawei's mobile phones? Do you have your own "spare tire"? In the current rapid development of various cutting-edge technologies, automatic driving has become a strategic commanding point for all countries. This technology is the product of the deep integration of the automobile industry with the new generation of information technologies such as artificial intelligence, Internet of things and high-performance computing. It is also the main direction of the development of intelligence and networking in the field of global automobile and transportation. Chinese autopilot is transitioning from L2 to L3 source: screenshot of Southwest Securities Research Report in March 2021 As an important part of the decision layer of the automatic driving system, the automatic driving chip is the key hardware support for the realization of automatic driving. With the improvement of automatic driving level, the increase of data processing demand and the update and iteration of software, the demand of the automotive industry for the number of intelligent driving chips, various performance parameters, especially computing power, continues to increase. Soochow securities estimates that the value of AI chip bicycle will increase from $100 in 2019 to $1000 + in 2025. According to the data of cheetah Research Institute, the scale of China's autopilot chip market has grown rapidly since 2020. It is estimated that the scale will reach 4.6 billion yuan in 2025, with a compound growth rate of 55.9% in five years; From 2025 to 2030, the growth rate of market scale is expected to slow down, reaching 10.4 billion yuan in 2030, with a five-year compound growth rate of 17.7%. Mackenzie: by 2040, self driving cars will account for 40% of China's new car sales. But like smart phones, the key component of autopilot chips is still dominated by American enterprises. The Research Report of cheetah Research Institute in July 2021 pointed out that NVIDIA and Intel Mobileye are the first echelon of autopilot chips in the world. Mobileye mainly focuses on L1 and L2 autopilot, with an overall market share of more than 70%; NVIDIA focuses on L3 and above markets and is in a leading position in this market; The existence of Qualcomm communication giants cannot be ignored; Tesla occupies an independent echelon, and its self-developed FSD chip is the most mature and mass-produced chip in the current market. Chinese enterprises also attach great importance to this field. Horizon, black sesame intelligence, Xinchi technology, Huawei Hisilicon, Zero run and other companies have launched layout. Among them, horizon has launched the third generation of vehicle specification products, journey 5; The newly launched Huashan No. 2 a1000 pro of black sesame intelligence was successfully released in July last year; Xinchi Technology launched a new generation of v9t autopilot chip in April last year; Zero run automobile cooperates with Dahua to launch Lingxin 01 with 28nm process; Weilai is also reported to be developing its own chip; Huawei has built the mdc810 autopilot platform. It is reported that the platform is equipped with shengteng 610 chip. However, there is still a big gap between Chinese enterprises and American competitors in this field. "China's autonomous driving chip industry is still in the early stage of development. Chinese enterprises are currently in the stage of imitation, learning and catching up. Chip enterprises buy some IP, design independently, and then find a OEM for production. Chinese local enterprises have negligible market share at this stage because there is almost no mass production of autonomous driving front loading." According to the Research Report of the cheetah Research Institute last July. Source: screenshot of Research Report of cheetah Research Institute in July 2021 For car companies eager to mass produce and list, American chips are still the first choice. According to the financial times, there are at least 18 different Chinese companies that use the chip of NVIDIA to be used in the key links of automatic driving business. These companies include Alibaba's auto driving start-up company AutoX and Wen Yuan know line jointly invested by Renault, Nissan and MITSUBISHI, and Apple's investment in China. Randy Abrams, head of semiconductor research at Credit Suisse Asia, pointed out that Chinese electric vehicle enterprises still need to use the best products of their kind in key technologies. "NVIDIA is still more advanced in vehicle processing platforms and artificial intelligence training and simulation required to develop these platforms". In January this year, Geely holdings announced that the company's smart electric brand krypton and Intel's Mobileye will further deepen cooperation. The two sides plan to jointly launch the world's first consumer electric vehicle with L4 level intelligent driving ability in 2024; At the end of 2020, Great Wall Motors and Qualcomm reached cooperation in the field of automatic driving and will adopt Qualcomm snapdragon ride? In the high-end models of Great Wall Motors produced in 2022? Platform. An anonymous Chinese semiconductor source said that the development of autonomous driving chips in China is still plagued by a series of "fundamental problems". "China's largest auto company produces about 1 million cars a year," the semiconductor person pointed out. This scale is far from enough to provide sustained funding for advanced autonomous driving chips, and it is very difficult to achieve the scale required to be competitive. The Financial Times reported that in recent years, the high cost of developing autonomous driving chips has prevented dozens of companies from entering the market. At the same time, industry giants such as NVIDIA are repurposing chip technologies such as GPU, which are the basis of the explosive development of mobile phones, desktop computers and data centers over the past 20 years. Chinese mainland chip industry chain in the global market share of data sources: Financial Times In addition to technology and cost, how to persuade customers to verify has also become a difficulty for domestic autopilot chip players. Xiao Jianxiong, founder and CEO of autox, said his company was worried about delays in working with new players. "We hope to take action as soon as possible. We want to expand the scale of autopilot cars. Productivity is really important to us. The maturity of this ecosystem is really very helpful." He said of working with NVIDIA and its engineers and supplier ecosystems. In addition to chip design, British media also mentioned the risk that domestic chips could not be manufactured. The Financial Times reported that the capital expenditure of TSMC, the world's largest chip foundry, reached US $44 billion this year, compared with about US $30 billion last year. Chinese mainland's largest foundries, by contrast, spent $5 billion on capital expenditures this year. Most industry experts believe that the gap between SMIC and TSMC is still five years away. Mass production time node between SMIC and each wafer factory But the "worry" of the financial times is not entirely accurate. Chinese auto driving chip enterprises have emerged, and domestic car enterprises are also eager to try. Although the chips of American companies such as Mobileye have adopted the most advanced 5nm process, horizon, black sesame and Zero run also have products with 16nm and 28nm processes, which can be OEM by SMIC. Moreover, although more advanced processes can improve chip performance, the power consumption and cost brought by higher computing power are also factors that chip enterprises must consider. In addition, the solutions provided by Mobileye under Intel cannot customize the "soul" for car enterprises. At present, domestic automatic driving chip manufacturers have adopted an open cooperation mode, which can feed back a large amount of accumulated data to the development process, which has a greater attraction to domestic car enterprises. For example, before, the first model of ideal car, ideal one, used Mobileye's eyeq4 chip. However, because the algorithm of Mobileye chip is fixed in the chip, it is unable to iterate in the visual perception layer. It is an "algorithm black box". Affected by this, the 2021 ideal one uses horizon journey 3 chip to replace the chip of Mobileye. Another reason why US companies will not easily cut off supply is that China has the world's largest new energy vehicle market, accounting for half of the world, and the popularity of automatic driving is accelerating. On January 12, China's Ministry of industry and information technology issued a document saying that China's new energy vehicle sales in 2021 were 3.521 million, of which the new passenger vehicle market equipped with combined driving assistance system accounted for 20%. Two years ago, the penetration rate of L2 assisted driving was only 3.3%. If American enterprises cut off supply, while losing a huge cake, it will undoubtedly bring rare opportunities to domestic chips. (Xinhua News Agency)