news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

after entering the second half of the competition for automobile intelligence, autonomous driving has become the "decisive factor" for car companies. in the field of autonomous driving, tesla has always been a benchmark. in the early days, it adopted mobileye's integrated software and hardware solution, and then switched to nvidia's chip + self-developed algorithm software and hardware decoupling solution. now it has returned to the "emphasis on software and hardware integration" strategy based on self-developed chips and self-developed intelligent driving algorithms.

this has also led to the hardware-software integration solution becoming a new trend in the autonomous driving industry. recently, chentao capital and three other parties jointly released the 2024 "autonomous driving hardware-software integration evolution trend research report" (hereinafter referred to as the "research report"), which shows that due to the successful case of tesla, the current general view in the industry is that it is feasible for vehicle manufacturers to adopt a "heavy hardware-software integration" solution.

"focus on integration of hardware and software" means that the same company completes the full stack development of chips, algorithms, operating systems/middleware, and an ecological cooperation model is derived based on this. this model includes overseas companies such as mobileye, tesla, and nvidia (under development) and domestic companies such as huawei, horizon, and momenta (under development).

in terms of domestic vehicle manufacturers, wei, xiaoli, byd, and geely are also working towards the direction of software and hardware integration, and have entered the field of self-developed chips from the early self-developed algorithms. on august 27, xiaopeng motors announced that the company's self-developed turing chip was successfully taped out; and a month ago, on july 27, weilai released its self-developed 5-nanometer smart driving chip "shenji nx9031"; geely's new generation of smart driving chip ad1000 was officially unveiled at the beginning of this year. in addition, it is reported that ideal, byd, etc. have formed teams to conduct research on autonomous driving chips.

in addition to the "heavy hardware and software integration" solution, the "light hardware and software integration" autonomous driving solution is also very popular among car companies. "light hardware and software integration" means that the autonomous driving solution company uses third-party chips, has optimization capabilities and product delivery experience on a specific chip, and can maximize the potential of the chip. typical cases in this regard include zhuoyu (dji) and momenta.

the above research report states that the integrated hardware and software autonomous driving solution can become a trend, on the one hand because it can achieve higher performance, lower power consumption, lower latency and closer integration, and more importantly, it can bring obvious cost advantages to enterprises. taking tesla's fsd chip as an example, the one-time tape-out cost of chip 1 (hw3) is estimated to be us$10, and chip 2 (hw4) is us$30, while the corresponding value of nvidia orin chip is us$30, and thor is as high as us$100.

however, on the other hand, the investment of car companies in self-developed chips is very large. the input-output ratio will be a heavy topic that these car companies have to face. research reports show that taking a 7nm process and 100+tops high-performance soc as an example, the r&d cost is higher than 100 million us dollars (including labor costs, tape-out costs, packaging and testing costs, ip licensing fees, etc.). liu yudong, executive general manager of chentao capital, publicly stated that from the economic considerations of car companies, we believe that if the shipment volume of self-developed chips is less than 1 million pieces, it may be difficult to achieve a balance in the input-output ratio.

wang xiaohui, the person in charge of domain control products at tianjue technology, said at the 2024 autonomous driving software and hardware collaborative development forum that automakers cannot cover the cost of self-developed chips by selling cars well in the short term. only by occupying a certain market share in the long term can they achieve the input-output ratio required by chip manufacturing that conforms to business logic. in the future, the model of automakers making their own software and hardware integrated solutions may exist, but there will not be too many, at most 1-2.

regarding the future development trend of software and hardware integration, the above research report believes that, in general, software and hardware integration and software and hardware decoupling are two sides of the same coin, and the market will eventually form a situation where the two coexist, but in the short term, companies that adopt software and hardware integration solutions will show stronger competitiveness in the market. in the autonomous driving industry, the trend of software and hardware integration will vary depending on the level of autonomous driving solutions: for low-level intelligent driving, car companies often directly adopt the supplier's software and hardware integration solutions and develop in the direction of standardization; for key capabilities such as high-level intelligent driving algorithms, the proportion of self-developed by car companies will increase.