How domestic solutions for intelligent automobiles reshape the industry landscape from Huawei's perspective

5. Potential challenges facing Huawei as a Tier 1 company

1. Market space: Going overseas is worth watching

Going overseas has become an increasingly important new growth curve for domestic OEMs. In recent years, domestic OEMs have taken the initiative to go global in various ways, such as exporting technology, services, and brands overseas. In addition, China's automotive industry has achieved breakthroughs and leadership in many professional segments, and has extended to overseas markets. According to Wind data, my country's domestically manufactured automobile exports reached 3.92 million units from January to October 2023, a year-on-year increase of 59.7%. From the perspective of downstream export destinations, Asia, Europe, and South America are the top three export regions, with export sales accounting for 36%, 26%, and 14% in 2022, respectively (Source: China Passenger Car Association, e-Europe Think Tank). In recent years, as car companies such as NIO, Xpeng, and BYD have stepped up their efforts to go overseas, Europe and Southeast Asia have become popular areas for going overseas.

From the perspective of exporters, the main OEMs in my country's passenger car exports in recent years are SAIC, Chery, Tesla and other OEMs. In 2022, the top five passenger car exporters in my country are SAIC, Chery, Tesla, Changan and Dongfeng, with export volumes of 480,000, 450,000, 270,000, 250,000 and 240,000 respectively. From January to August 2023, the top five vehicle exports in my country are SAIC, Chery, Tesla, Changan and Geely, with export volumes of 670,000, 560,000, 240,000, 230,000 and 210,000 respectively.

There are still practical obstacles for the export of local autonomous driving technology as a whole. In the early days, my country's automobile export was mostly based on low-price competition strategies. At present, domestic automobile export is more dependent on electrification and intelligence to support product advantages. From the perspective of the regulatory environment: my country's intelligent driving industry technology faces very strict safety supervision when entering overseas mature markets, such as personal information protection and data security. From the perspective of the trade environment: In October 2023, the European Commission issued an announcement, deciding to launch an anti-subsidy investigation on pure electric passenger cars imported from China. Therefore, there is still a certain resistance to the overall export of China's autonomous driving industry chain (especially localization solutions such as Huawei) in the short term. The core variables of subsequent export progress still need to observe the gap between the progress of the domestic and foreign automotive intelligent industries, the geopolitical environment and other multiple factors. From a practical perspective, industry chain players may adopt a step-by-step strategy. In the early stage of exporting automotive intelligent products, they can focus more on cockpits and low-level intelligent driving solutions.

(II) Product empowerment: short-term product costs are relatively high

The scale effect is one of the core links of the competitiveness of the automotive Tier 1. Based on the industrial law of the automotive manufacturing industry, costs generally continue to decline with the expansion of scale, thus forming a strong scale effect moat, which is reflected in the financial statements as higher gross profit margins and lower period expense rates. Considering the heterogeneity of products of different companies, the gross profit margin alone is not convincing. We observe from the period expense rate (period expense rate = management expense rate + sales expense rate + research and development expense rate + financial expense rate). Two typical tier 1s were selected. From the analysis of long-term financial data, the core variable examined is the relationship between the company's revenue growth rate and the change in the period expense rate. That is, if the company's revenue growth rate is positive, the change in the period expense rate is negative (that is, the period expense rate is downward), indicating that the scale effect exists. From the financial data disclosed in the financial reports of Desay SV and Huayang Group (covered by GF Auto Group), the revenue growth rate and the change in the period expense rate show a significant negative correlation, and the statistical significance is significant. The correlation between the revenue growth rate and the change in the period expense rate of the two companies is -73% and -76%, respectively.

Huawei's automotive BU has a large number of R&D personnel, but the current revenue generation is relatively low. The lack of economies of scale has led to high short-term product costs. According to the financial reports of relevant companies, based on the comparison of Huawei's automotive BU and the typical tier 1 operating indicators, the current revenue generation of each R&D personnel in Huawei's automotive BU is about 300,000 yuan, while that of Desay SV and Jingwei Hirain is 4.26 million yuan and 1.65 million yuan respectively. Huawei's automotive BU's current R&D investment is also relatively large, with an investment of more than 20 billion yuan in 4 years, while that of Desay SV and Jingwei Hirain is 1.7 billion yuan and 700 million yuan respectively.

For relevant partners and suppliers, there is product homogeneity and gross profit margin pressure. Huawei's automotive BU currently has a small revenue scale because it is difficult to obtain representative financial data from the perspective of industry chain suppliers. However, similar situations have precedents in the server field. Huawei has mastered the core value chain of AI servers, and its suppliers are mostly in the links of whole machine assembly and sales channels. The degree of homogeneity of product added value and bargaining power is relatively high, and the gross profit margin is relatively general. According to data disclosed by Wind, in 2022, the gross profit margins of Digital China's self-branded business and Tuowei Information's domestic self-branded server and PC business were 10.4% and 12.9% respectively. From the perspective of industrial logic, the core components of automotive intelligence are chips, intelligent driving domain control, on-board sensors, core bottom software, algorithms and other chains. Huawei has considerable strength in related fields. Therefore, from the perspective of pricing power in the industrial chain, Huawei's automotive suppliers and OEM partners are likely to have certain gross profit margin pressure and product homogeneity pressure.

Investment opportunities clearly exist, but they are more industry β than individual stock α. The trend of automobile intelligence is clear, and the industry is still in its infancy. In the early stage of industry development, foreign-funded solutions accounted for the majority. From 2019 to 2020, the rise of domestic solutions has been obvious. In the future, the share of domestic intelligent ecosystems may accelerate. Against this background, the entry threshold of relevant software and hardware suppliers may be lowered, and the industry β will be enhanced, but the profitability may be somewhat different from the ecosystem dominated by overseas chip manufacturers, so the alpha of individual companies is challenged.

(III) Challenges of product upgrade: Challenges faced in the field of cabin-pilot integration

From the perspective of the evolution of automobile intelligence, the overall evolution trend of automobile EE structure is distributed ECU architecture, domain controller EE architecture, and centralized EE architecture: (1) Distributed stage: specific functions are controlled by specific ECUs. This structure cannot withstand the trend of increasingly rich automobile functions. Too many ECUs make the EE architecture extremely complicated; (2) Domain controller stage: Ethernet is introduced, and the EE architecture is further optimized based on different domain divisions. Furthermore, the smart cockpit domain and the smart driving domain are integrated, and a single AI chip realizes edge-side computing such as inside and outside the car and integration; (3) Centralized stage: The on-board central computer is formed, covering the body domain, power domain, chassis domain, and safety domain, and computing chips are integrated.

From 2023 to 2025, typical OEMs will focus on central domain control. At present, typical new forces and some independent car companies (Great Wall, BYD, Geely, etc.) have gradually entered the stage of domain concentration architecture, and joint ventures and foreign-funded car companies have followed suit. Some leading companies have begun to explore cross-domain fusion architecture and central computing architecture.

At present, the mainstream players in the field of smart driving chips in the world have mainly released two central domain control chips, both of which use a 5nm process. In the process of layout for cross-domain integration, NVIDIA, the leading company in smart driving chips, and Qualcomm, the leading company in smart cockpit chips, have successively released SoC chips for cross-domain fusion central computing. In 2022, NVIDIA released the Thor SoC for central computing platforms with a computing power of 2000TOPS. Qualcomm also released the Snapdragon Ride Flex SoC, which, together with an external AI accelerator, can achieve 2000TOPSAI computing power. NVIDIA Thor chip is expected to be mass-produced in 2025, and Zeekr has been announced as the first model; Qualcomm Flex chip is expected to be mass-produced in 2024. According to Gasgoo Automotive Information, NVIDIA Thor and Qualcomm Snapdragon Ride Flex both use a 5nm process. Based on the industry rules, we believe that this also reflects the common judgment of the industry leaders on the SOC in the era of cabin-driving integration—that is, under the requirements of stronger computing power and lower power consumption, the main control SOC in the era of cabin-driving integration needs to adopt a more advanced semiconductor process.