Automotive industry recently has been more in the public eye than ever before. Roland Berger indicates that global light vehicle production volume had kept rising since 2012 but started to weaken in the past two years. However, in this still favorable environment, global supplier industry is expected to maintain its profitability level.

The four automotive megatrends Mobility, Autonomous driving, Digitization and Electrification will continue to change the automotive industry. For instance, new mobility business models are poised to disrupt car ownership, personal mobility and goods logistics. The share of new vehicle sales for application in the field of new mobility (e.g. ride hailing, car sharing) may range between 10%-15% in the US and Europe and up to 35% in China by 2025. Moreover, due to increasing regulatory pressure and accelerating technology advancement, Roland Berger and Lazard jointly estimate that scenarios for the share of EV cars in 2025 range from 8-20% in the US, 20-32% in Europe and 29-47% in China.

By means of artificial intelligence that is a globally hot topic nowadays, autonomous driving and digitization are offered limitless possibilities. In digitization, within the next 10 years almost all cars in mature markets will have some form of connectivity. Notably, the timeline for level 4/5 autonomous keeps accelerating as economics, policy and technology are put into practice: Penetration rates for autonomous cars (SAE level 4/5) may reach a level between 5% and 26% in 15 – 20 years. This number can be higher in China as the government has enacted favorable policies so as to promote development of autonomous driving sector.

Chinese autonomous driving industry has been drastically growing since 2015 when China’s state council published the "Made in China 2025" document that indicates development of the autonomous driving-based intelligent vehicle as one of the most significant strategies in the automotive industry. A large number of startups got devoted to autonomous driving domain. In the next year many car makers released autonomous driving development plans. Since then tremendous hot cash has been poured into the new technology area.

General Investment Trend

The autonomous driving investment displays an upward trend in both volume and frequency from 2011 to 2017. Due to policy effects, sharp growth of both volume and frequency appears in 2015 and remained in the next two years. The amount of cash flowing into self-driving sector reached the peak (CNY 29.24 billion) in 2017.

After three-year rapid growth, the market became cautious about the industry as investment amount declines in 2018, despite an increasing number of investment activities. Up to the mid of 2019, the amount of investment has already surpassed the amount in 2018 but investment activities are much less than the same period last year. This signals a more concentrated market is forming.

Hardware and Software

Investment in autonomous driving can be categorized into two aspects: hardware investment and software investment.

The hardware basically refers to sensors such as radar and camera that are used to achieve connectivity of vehicle to external environment and monitoring system. So far, an autonomous vehicle is generally equipped with five short-distance laser radars, eight millimeter-wave radars, sixteen cameras and one or two inertial measurement units (IMU).

Statistics from Yole Developpement show that around half of investment in autonomous driving field is used for R&D of sensors which are a core part of intelligent driving. With consecutive technology advances, the cost of sensors will be eventually much lower. For instance, the price of a laser radar product of Velodyne decreased from over USD 10,000 to USD 3,999. Hence, along with maturing industrial chain and cheaper core parts, the cost of autonomous driving system will have a notable cut in the future.

The software covers broad self-driving algorithms, high-precision positioning system and v2x services (vehicle to everything). Baidu, one of five companies granted licenses by China MIIT to develop artificial intelligence platforms, announced the Apollo program in April 2017 aiming to establish an open-source platform for autonomous driving software. Based on artificial intelligence technology, the Apollo platform could provide various functions for any part related, such as software, hardware, statistics set and computing tool, funds, developer society, testing ground and so on. The high-precision positioning system is a necessity for self-driving technology to be landed and commercialized. According to the China Industrial Information website, the global high-precision positioning market is expected to grow up to USD 2.1 billion in 2020 and USD 9.4 billion in 2025. By 2020, China will launch 35 satellites in total and provide positioning services for global users.

The wide application of self-driving vehicle entails synergy of cars and roads. From the technology perspective, a single intelligent vehicle has limits facing various transport scenarios. For instance, most laser radars can effectively cover the range of tens of meters, but the range is not long enough for a single vehicle to quickly respond to a stationary entity in front of it when keeping up a slightly high speed. However, such problems are solved if the road tells the car what is happening around. Moreover, information exchange between roads and cars makes sensors less important and necessary, so it helps reduce sensors equipped on an intelligent vehicle to some extent, and further lower the cost of manufacturing. Hence, cooperation of vehicles and roads is an essential condition to realize mass production of autonomous driving vehicles. The foremost thing is: v2x service is the basis of such synergy as it works as a bridge between cars and external environment including roads, and any form of information exchange between cars and the outside must be conducted using v2x technology.

So far, Chinese internet giants including Baidu, Alibaba and Tencent consecutively launched v2x application platforms which convey complicated algorithms in the past few years.

Investment Structure Layout

We calculate the accumulated amount of investment activities that occurred in China between 2011 and mid of 2019 in subsidiary fields of self-driving industry. The chart below indicates more than half of total investment flows to vehicle manufacturing, which is reasonable. Barriers to entering the automotive industry are impenetrable compared with other industries. The primary reason behind is the high cost of car platform making, which covers the purchase of materials and equipment, vehicle design, manufacturing, quality control and so on. Even big traditional carmakers like Volkswagen attempt to use the same vehicle platform so as to reduce production cost.

However, as the new generation of vehicles, self-driving cars entail the different platform from traditional ones. For startups, building platforms on their own or partnering with the traditional manufacturer are two options to have suitable autonomous driving vehicle platform, but either option is at a high cost.

Algorithm gains the second largest portion of investment over recent years. Algorithm in autonomous driving industry basically refers to the deep learning algorithm, a class of machine learning algorithms. Deep learning uses multiple layers to progressively extract higher level features from raw input and produces results comparable to and in some cases superior to human experts. Learning can be based on different learning framework and can be supervised, semi-supervised or unsupervised, depending on the way of training. However, despite various learning frameworks, the quality of the algorithm essentially relies on both quality and quantity of input data. There is no short cut to quickly obtain massive driving data needed for the system to make "human" decision. For companies providing algorithm services, road test is an essential way to have the raw input, the quality of which will be consecutively improving as more and more tests are conducted, but along comes a researching cost.

Compared with manufacturer and algorithm, sensor, v2x, position and platform/chip have much smaller shares of investment in autonomous driving domain. Reasons can be varied. By calculating the average investment amount in these fields, we further analyze the investment trend in autonomous driving industry.

Analysis Dissection

The average investment amount was calculated by total volume dividing total frequency and could represent barrier level of entering an industry to some extent. Results show the amount of manufacture is dramatically higher than other sectors, which is in accordance with the statement mentioned above that car making needs massive capital input.

Algorithm has a lower average amount than sensor but its total amount is much higher. In other words, lower barriers of algorithm sector attract more investment opportunities. Therefore, it has been the main entrance for global startups to enter the self-driving industry. Chinese market data provides evidence that the total amount of algorithm investment is actually promoted by investment frequency rather than volume, but it also indicates strong competitiveness in algorithm market. As algorithm heavily relies on quality and quantity of test data, firms that enter the market earlier and conduct more tests could be more dominant, whereas laggard algorithm is meant to be eliminated.

In contrast to algorithm, apart from manufacture, v2x has the highest average investment amount but low accumulated volume. So far, v2x applications still remain undeveloped as it involves many problems to be settled. For example, the realization of v2x services entails full commercialization of 5G technology in order to satisfy bandwidth needs of data transfers. V2x service providers also seek to acquire assistance from the government when they are engaged in integration of vehicle, internet and transportation infrastructure. Moreover, v2x faces problems with technology, such as inaccurate communication.

Both technological difficulties and unsound infrastructure system jointly form a high entry barrier that keeps many startups and investment away from v2x market. The complexity of deploying v2x services suggests that only players with a mass of bets in the pocket could join this table. Chinese tech giants Baidu, Alibaba and Tencent illustrate superiority in the development of v2x services, while other v2x startups are mixed.

Computing platform and chip and position system have small shares of investment in terms of average and accumulated amount, which indicates the market's investment preference. For the position system side, the HD map features plenty of inputs and long commercialization cycle. There is still space left for Chinese map companies to improve positioning technology and the market are still on its way to get the technology commercialized. Despite remaining immature, position system market has already hatched some magnates that take many shares of the market. According to iResearch, three map giants Navinfo, Amap and EMG occupy around 99% of market share in China. For investors, throwing money into the map industry might not bring them proportionate returns.

In terms of computing platform and chip, the traditional ECU computing platform cannot satisfy requirements of deep learning algorithm, so upgraded computing platforms are necessary to produce self-driving vehicles. However, chips, the core part of computing platform, directly determines platform performance. In other words, the priority should be given to the development of chip industry, but it is a long journey as the domestic chip market is almost monopolized by foreign companies like Intel and NVIDIA. To break through the monopoly, some Chinese chip companies were founded in recent years. For example, Cambricon, a Chinese chip company representative, closed Series B round of funding in 2018, two years after its founding, and now becomes a unicorn company in Chinese chip market with a valuation at USD 2.5 billion.

Investment in Chinese autonomous driving industry over past years heavily flowed to algorithm field, making algorithm a comparative advantage compared with European and American markets, but the weakness of hardware is visible: demands of core parts of intelligent vehicles still heavily rely on foreign suppliers. Though some Chinese companies have been hatched out, investment in the hardware domain is far from adequate.

The investment trend shows that after three-year development a more concentrated market is forming and those who are left behind in autonomous driving industry will be gradually eliminated.