Dongxing Securities Co., Ltd.: Smart driving prices gradually declining, industry expected to accelerate development

Zhitong
2024.08.28 12:46
portai
I'm PortAI, I can summarize articles.

Dongxing Securities released a research report stating that the price of autonomous driving is gradually decreasing, and the industry is expected to accelerate its development. With XPeng's chip "XPeng Turing" successfully mass-produced for multiple vehicle models, and Huawei's smart car division turning losses into profits in the first half of the year, industry differentiation is accelerating, and the strength of leading companies is increasing. In addition, domestic autonomous driving chip manufacturers such as Black Sesame Intelligence, Horizon Robotics, and Desay SV are expected to benefit from industry development. The main point of view is that the successful mass production of autonomous driving chips by Nio and XPeng marks the industry entering a stage of diversified competition

According to the report released by Dongxing Securities on the Zhitong Finance APP, low-cost autonomous driving solutions are gradually emerging, and the industry landscape is facing accelerated differentiation. The strength of leading enterprises is gradually increasing. This can be seen from the expectations of Tesla's FSD entering China, the release of Huawei-related vehicle models and solutions, and the choices of self-developed chips by XPeng and Momenta. It is evident that the pressure to reduce costs for OEMs and the core position of computing chips remain relatively certain. In this context, domestic autonomous driving chip manufacturers such as Black Sesame Intelligence (02533), Horizon Robotics (unlisted), and Desay SV (002920.SZ) which closely cooperate with overseas chip leaders are expected to benefit from industry development. Huawei's ecosystem partner Softcom Power (301236.SZ) and related automotive companies are also expected to benefit.

Events:

  1. On the evening of August 27th, XPeng Motors announced at the MONA M03 listing event that its global first chip "XPeng Turing", which can be simultaneously applied in robots, AI vehicles, and flying cars, successfully taped out on August 23rd. At the same time, the XPeng MONA M03Max version has advanced autonomous driving functions, and this version will start deliveries after the 2025 Chinese New Year.

  2. On August 25th, Sailus announced its plan to purchase 10% of Shenzhen Yingwang Intelligence from Huawei for 11.5 billion yuan in cash. According to the "Major Asset Purchase Report (Draft)" disclosed, Yingwang Intelligence achieved revenue of 10.435 billion yuan and a net profit of 2.231 billion yuan in the first half of this year, with gross and net profit margins reaching 55.36% and 21.38% respectively, turning losses into profits.

Key Points from Dongxing Securities:

Nio and XPeng successively tape out chips, entering a diversified competition stage for autonomous driving chips.

After Nio's self-developed autonomous driving chip "Nio Adam NX9031" taped out, XPeng announced that "XPeng Turing" has also successfully taped out, gradually entering the testing and verification stage. This chip is an end-to-end large model customized chip, integrating 2 self-developed NPUs (neural network processors), 2 independent ISPs (image signal processors), and adopting a domain-specific architecture for neural networks (DSA); with 40 core processors, it provides strong computing support for large models, supporting local operation of models with up to 30 billion parameters.

The advantages of self-developed computing chips mainly lie in the following aspects: in terms of the supply chain, relying on overseas manufacturers poses significant risks amid increased external uncertainties. Currently, high-end autonomous driving mainly uses Nvidia's Orin chip, with weak bargaining power for automakers and production constrained by Nvidia's supply; in terms of performance improvement, dedicated self-developed chips can customize chip specifications and requirements based on their own algorithm architecture, design chip architecture based on relevant neural networks, cut off irrelevant general-purpose modules to achieve the optimal balance between performance and power consumption; in terms of cost, short-term expenses can be exchanged for long-term gross profit. Although the R&D investment required for self-developed chips is significant, once successfully developed, the scale effect is expected to significantly reduce the cost of single-chip compared to outsourcing, leading to gross profit improvement; in terms of branding and differentiation, it can demonstrate the technological strength of automakers and achieve differentiation competition through software and hardware synergy.

However, self-developed chips still face challenges such as high investment, long implementation cycles, and difficulty in realizing scale effects dependent on their own shipment volume. The industry may enter a stage of diversified competition, with overseas suppliers still holding advantages. Suppliers such as Huawei and Horizon Robotics continue to penetrate high-end autonomous driving through ecological cooperation. Some tech-savvy automakers will partially break away from external suppliers through self-developed strategies, presenting development opportunities for the domestic computing ecosystem Smart driving prices gradually decline, industry expected to accelerate development.

XPeng MONA M03 offers three models, with the entry-level 515 Long Range version priced at 119,800 yuan, the 620 Extended Range version at 129,800 yuan, and the 580 Extended Range Max version at 155,800 yuan. According to XPeng's promotion, the Max version does not have a lidar, equipped with advanced intelligent driving assistance, not limited to cities, routes, or road conditions. This version will start delivery after the 2025 Chinese New Year. Currently, the prices of models equipped with L2-level assisted driving functions and even advanced intelligent driving are gradually declining. With the penetration in low-priced models, car companies are expected to accelerate in terms of smart driving product shipments, data accumulation, consumer education, etc. In addition to China's policy direction of "single vehicle intelligence + vehicle-road coordination," reducing smart driving costs in the medium to long term through infrastructure construction will help drive industry development, and low-cost smart driving solutions may be gradually implemented.

Huawei's automotive business turns losses into profits, deepening ties with suppliers and ecosystem partners.

Huawei's smart car brand, Seres, was established earlier this year, relying on Huawei's related resources. Through three sales models: traditional parts sales, HI mode (Huawei Inside), and smart selection mode, it has achieved a leading advantage in the field of smart driving and smart cabins in China. It has stable cooperation with major OEMs such as Seres, Changan, Dongfeng, Chery, BAIC, and JAC. In the first half of this year, Seres has turned losses into profits, with revenue of 10.435 billion yuan, with hardware accounting for 58.39% (60.93 billion yuan, gross margin of 33.41%) and software accounting for 41.61% (43.42 billion yuan, gross margin of 86.17%). Its profitability is at the forefront of the industry, and economies of scale are gradually being realized.

The strength of Huawei's automotive business lies not only in its technical capabilities but also in its strategic direction. Its "platform + ecosystem" strategy provides digital bases and development tools for smart cars. By the end of 2023, it has developed more than 300 upstream and downstream partners in the industry chain. Its flexible cooperation model and ecosystem brand effect have gradually strengthened its industry position. The downstream car manufacturers' investment this time is expected to further enhance the stickiness of Huawei's automotive ecosystem.

Risk Warning:

Risks include slower-than-expected development of the smart driving industry, increased investment in self-developed by car companies, slower-than-expected end-to-end technology development, less-than-expected industry policy support, and slower-than-expected company listing speed