news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Nvidia's Orin-X intelligent driving chip has become standard for mid-to-high-end Chinese new energy vehicle products; in the eyes of users, this chip is also equated to a certain extent with advanced intelligent driving systems.

Recently, a car company's model upgrade caused a stir among users. The Intel Mobileye chip solution of its old models was upgraded to the Nvidia Orin-X solution with higher computing power. While the intelligent driving computing power was improved, the functional iteration potential was also greatly increased, which caused dissatisfaction among many old car owners. This reflects to some extent the importance and "high-end" feeling of Nvidia Orin-X intelligent driving chip in the eyes of users.

Currently, among domestic brands’ new energy vehicle products priced above 200,000 yuan, except for Tesla and products using Huawei’s ADS solutions, almost all products with advanced intelligent driving features use NVIDIA’s Orin-X intelligent driving chip.

Data from Gasgoo Automotive Research Institute shows that in the ranking of installed capacity of intelligent driving domain control chips in the Chinese market in 2023, Tesla's FSD chip ranked first, with shipments of approximately 1.208 million units, accounting for 37%; Nvidia's Orin-X chip ranked second, with shipments of 1.095 million units, accounting for 33.5%; the third-ranked Horizon Journey 5 chip has a large gap with the above two, with shipments of 200,000 units and a market share of 6.1%.

Since the FSD chip is only used by Tesla, Nvidia's competitors in the field of intelligent driving domain control chips are mostly Chinese suppliers such as Horizon Robotics and Black Sesame. The super computing power of a single Orin-X chip of 254TOPS makes it difficult for Nvidia to meet rivals. In 2025, Nvidia will debut the Thor chip with a single computing power of more than 2000TOPS on Zeekr, which seems to further consolidate Nvidia's leading advantage.

But Nvidia's competitors and customers don't seem to want Nvidia to take the big piece of the smart driving chip pie alone.

OEMs and suppliers continue to focus on smart driving chips

In July 2024, at the NIO IN 2024 Innovation Technology Day, Li Bin, Chairman and CEO of NIO, pulled out a chip half the size of his palm from his pocket and announced that the NIO Shenji NX9031 chip had been successfully taped out. This self-developed smart driving chip is manufactured using a 5-nanometer automotive-grade process and has more than 50 billion transistors. A single self-developed chip can achieve the performance of four industry flagship chips.

After NIO launched its smart driving chip, its competitors Xpeng Motors and Li Auto were also reported to be developing their own smart driving chips. Among them, Xpeng Motors has publicly stated that it will invest in four AI directions, including chips. It is reported that its first smart driving chip has been sent for tape-out and is expected to be returned in August; Li Auto's self-developed smart driving chip "Schumacher" is expected to be completed within the year.

Compared with traditional car companies, new car companies mostly use "intelligence" as their brand characteristics. Car companies such as Weilai and Xiaopeng, which emphasize intelligent driving capabilities, are the largest customer groups of NVIDIA Orin series chips in the Chinese market. Monitoring data from Gaogong Intelligent Automobile Research Institute shows that in 2023, Weilai, Xiaopeng and Li Auto alone will contribute nearly 90% of NVIDIA Orin's pre-installed market share in China.

Take NIO as an example. All its products are equipped with four Nvidia Orin-X chips as standard. Based on the annual sales of 160,000 vehicles in 2023, NIO purchased more than 640,000 Orin-X chips. The large-scale switching of major customers such as NIO, Xiaoli, and Li Auto to self-developed chips will have a huge potential negative impact on Nvidia.

According to Frost & Sullivan, the global automotive chip market will be worth about 310 billion yuan in 2022. With continued development and growing demand, the global automotive chip market is expected to exceed 600 billion yuan by 2030. McKinsey predicts that sales of fully autonomous vehicles will account for 15% of global passenger car sales in 2030.

With good market expectations, Nvidia's old rivals in the technology field are also eyeing the smart driving chip market. In January last year, Qualcomm launched the Snapdragon Ride Flex SoC, which supports both digital cockpit and smart driving functions with a single chip. Suppliers such as Bosch and Momenta, as well as car companies such as Nezha, have announced that they will adopt Qualcomm's solutions.

Domestic automotive chip manufacturers such as Horizon Robotics and Black Sesame are also thriving in the capital market. On August 8, Black Sesame Intelligent went public on the Hong Kong Stock Exchange. The company said that the next-generation SoC Huashan A2000 is under development and is expected to be launched in 2024. At the same time, the company is also expanding its automotive-grade chip capabilities, including further development and commercialization of the Wudang series cross-domain SoC, with the goal of achieving mass production by 2025.

Horizon Robotics is currently the smart driving technology company with the largest shipment volume of intelligent driving chips in China. On August 9, the company passed the IPO filing of the China Securities Regulatory Commission and obtained a pass for listing in Hong Kong. Previously, Horizon Robotics' products focused on entry-level intelligent driving assistance systems, and the computing power level was somewhat lower than that of NVIDIA's products of the same period. This year, Horizon Robotics released the Journey 6 series of chips, of which the computing power of the Journey 6P will reach 560TOPS. Its designated customers include SAIC Group, Volkswagen Group, BYD, Ideal Auto, GAC Group, Deep Blue Auto and many other well-known domestic auto companies.

In addition, Huawei is currently expanding its circle of friends. Car brands equipped with Huawei's intelligent driving system are evolving from smart car brands such as Wenjie and Zhijie to traditional car companies such as Changan Deep Blue and Dongfeng Lantu. An industry insider told the First Financial reporter that most Chinese car companies have communicated with Huawei on intelligent driving and intelligent cockpits. The strong brand appeal and good performance of Huawei's ADS intelligent driving system are expected to win more customers in the future.

Can Siege topple Nvidia?

Cost and price are considered to be important reasons why many automakers choose domestic chip suppliers or develop their own chips.

Judging from the current competitive situation in the automobile market, "reducing costs and increasing efficiency" has become the sword of Damocles hanging over the heads of automobile companies. However, the contradiction between reducing costs and intelligent driving systems is the high cost and high user attention. Richard Yu, Executive Director of Huawei, Chairman of the Terminal BG, and Chairman of the Intelligent Automotive Solutions BU, said on Weibo, "We have invested a lot in R&D in intelligent driving technology. Huawei's high-end intelligent driving products priced below 300,000 yuan are actually sold at a loss."

Monitoring data from Gaogong Intelligent Automobile Research Institute shows that in 2023, the average delivery price of new cars in the Chinese market equipped with NVIDIA intelligent driving chips as standard will still be as high as 386,600 yuan.

For car companies that develop intelligent driving systems based on Nvidia Orin-X, chips are expected to become the most expensive parts in the car after power batteries. A management of an autonomous driving company told reporters that the current price of bulk purchase of Orin-X is less than 10,000 yuan per chip. NIO uses four Orin-X chips in each car, and the cost may be close to 40,000 yuan; the mainstream solution of most car companies is to use two Orin-X chips, and the chip cost is also in the five figures.

So despite the huge investment in chip development, Li Bin still believes that in the long run, self-developed chips can achieve technological cost reduction.Previously, Li Bin said in an interview with reporters that the company purchased a lot of Nvidia chips last year, which cost the company a lot of money. For this reason, the company turned to self-developed chips. One chip can replace four chips, so it can reduce costs.

However, not all executives of auto companies agree that self-developed chips are an effective way to reduce costs. Zhu Jiangming, chairman and CEO of Leapmotor, once said in an interview with reporters that compared with electronic consumer products with shipments of tens of millions or even hundreds of millions, car sales are too small to form economies of scale, and the investment in chip research and development is too large, so it is not cost-effective to develop chips in-house.

In addition to reducing costs, self-developed intelligent driving chips can better achieve software and hardware integration and achieve better performance. Tesla's FSD and Huawei's ADS are the two leading intelligent driving assistance systems in the industry, both of which use self-developed chips and self-developed algorithms. Taking Tesla FSD as an example, its self-developed Hardware3.0 (HW3.0) intelligent driving computing power is 144TOPS, but the FSD system based on this hardware can successfully realize high-speed and urban NOA functions.

Most of the intelligent driving systems in China that use the Orin-X solution require two or more Orin-X chips to achieve high-speed and urban NOA functions, with a computing power reserve of 508TOPS. Technical personnel from several new car companies said that it is difficult to achieve urban NOA functions using a single Orin-X chip (computing power 254TOPS).

However, several autonomous driving R&D personnel told reporters that compared with other chip solutions on the market, Nvidia Orin-X still has huge advantages in terms of chip computing power and surrounding development tools.

Faced with the "siege" from car companies and friendly competitors, NVIDIA has begun to independently develop intelligent driving systems, changing the current situation of only providing computing power chips.

At present, Nvidia's revenue share in the automotive field is not high. According to the 2024 fiscal year report disclosed by Nvidia, Nvidia's automotive business revenue in that fiscal year was US$1.091 billion, a year-on-year increase of 21%, but the proportion of the company's total business revenue was only 1.79%.

The industry generally believes that the value of software on smart cars will increase significantly. Morgan Stanley believes that software currently accounts for 10% of the value of a car, and will account for about 60% of the value of a car in the future; Volkswagen said that by 2030, the cost of software development will reach about 50% of the cost of vehicle development.

For NVIDIA, the collaborative development of software and hardware can not only bring better new energy, but also open up a new profit track. In addition to a series of products based on smart driving chip hardware, NVIDIA has also begun to directly deploy smart driving systems in the automotive business. Previously, NVIDIA had won Mercedes-Benz's autonomous driving system development project and provided a full-stack solution of software and hardware, but the progress was not as expected; in early 2023, Wu Xinzhou, the former vice president of autonomous driving, joined NVIDIA to lead the research and development and mass production of the full-stack autonomous driving software system. And this layout of NVIDIA has also made itself and many of its former customers potential competitors.