
Photo/VCG
On the evening of January 7, the major U.S. stock indices opened higher. Among them, NVIDIA started with a gain of over 2%, reaching a new all-time high. However, the good times were short-lived as NVIDIA's stock began to plummet, and closed 6% lower on the day.
On the same day, NVIDIA CEO Jensen Huang announced a series of significant technological innovations at the CES 2025 event in Las Vegas. These included the new RTX 50 series GPUs, AI supercomputers, the next-generation automotive smart driving chip Thor, and the newly launched Cosmos world foundation model. Huang stated, "GeForce brought AI into millions of homes, and now AI is coming back to GeForce."
On the first day of CES2025, Huang revealed the automotive partners NVIDIA is working with, including Tesla, BYD, Jaguar Land Rover, Li Auto, Mercedes-Benz, Toyota, Rivian, Xiaomi Auto, Volvo, Lucid, among others. Notably, NIO and XPeng Motors, which announced their own smart driving chip development last year, were not on this list. As early customers of NVIDIA's Orin chip, models like XPeng's P7 and NIO's ES8 currently use NVIDIA chips. Some speculate that these automakers may have reached the stage of deploying their self-developed chips, gradually reducing their reliance on NVIDIA.
NVIDIA has also established partnerships with several Chinese companies. During his keynote speech at CES 2025, Huang introduced the new RTX series GPUs and the next-generation automotive processor "Thor." Additionally, NVIDIA provided a full-range solution from virtual to real through its Omniverse platform and Cosmos simulation engine, accelerating the development and testing of autonomous driving systems.
In terms of cooperation, NVIDIA announced that Toyota will build its next-generation vehicles on the NVIDIA DRIVE AGX Orin platform, running the safety-certified NVIDIA DriveOS operating system. Furthermore, Aurora, Continental, and NVIDIA announced a long-term strategic partnership to deploy autonomous trucks powered by NVIDIA DRIVE on a large scale. The DriveOS system will be integrated into Aurora Driver, which is part of Continental's plan to produce SAE Level 4 autonomous driving systems by 2027.
As one of the most important suppliers in the automotive chip field, NVIDIA holds a significant market share. According to data from the Global Automotive Research Institute, in the 2023 China market for intelligent driving domain controller chip installations, Tesla's FSD chip ranked first with about 1.208 million units, accounting for 37%; NVIDIA's Orin-X chip ranked second with 1.095 million units, accounting for 33.5%. Since Tesla's FSD chip is exclusive to Tesla, most domestic automakers primarily use NVIDIA chips.
In addition to BYD, Li Auto, Xiaomi EV, XPeng, and NIO, NVIDIA has also partnered with many other Chinese automakers and autonomous driving companies. For example, in January 2024, NVIDIA partnered with Great Wall Motor to develop a high-end intelligent driving system based on the DRIVE Orin centralized computing platform. The Coffee Pilot system can support parking, highway, and urban scenarios, achieving full-scenario intelligent navigation and assisted driving functions without the need for high-precision maps. On the autonomous driving enterprise side, domestic companies like Yuanrong Qixing are also NVIDIA partners. It is reported that the two parties plan to use NVIDIA's DRIVE Thor chip to adapt to the company's end-to-end intelligent driving model by 2025.
Regarding the absence of XPeng and NIO from NVIDIA's automotive partner circle, some believe it is due to the significant progress in their self-developed chips, as well as issues with the NVIDIA Thor chip's multiple delays and export restrictions.
According to Ji Xuehong, director of the Automotive Industry Innovation Research Center at North China University of Technology, companies like NVIDIA have a chip profit margin of up to 90%, which means automakers have to bear high costs when purchasing. If automakers can develop their own chips, they can control costs to some extent.
File photo/Zheng Yuhang (NBD)
Public information shows that the Thor chip was released by NVIDIA at the GTC conference in 2022. The chip has a computing power of 2000 TFLOPS, double that of Altan and eight times that of Orin. It was originally planned to be mass-produced in 2024 and put into use in 2025. Li Auto, BYD, and XPeng were customers of NVIDIA's Thor chip.
It is worth noting that after the launch of XPeng P7+ this year, the product manager of XPeng's P series and G series stated that the P7+ was initially planned to use the NVIDIA Thor chip directly during its development. However, due to the chip's repeated delays, it ultimately adopted a dual Orin-X configuration.
In fact, the Thor chip has been rumored to be delayed multiple times. It was initially planned to be mass-produced in mid-2024 but was soon changed to large-scale production and delivery between April and May 2025. According to the latest news, NVIDIA's Thor chip will be mass-produced in 2027.
"If there is a shortage of intelligent driving chips, it will affect the research and development and production plans of automobile manufacturers," said Jiang Han, a senior researcher at Think Tank Pangoal. This situation will promote the acceleration of the automotive intelligent driving chip industry and push the industry towards a more independent and controllable direction. In fact, under the background of the delay of the Thor chip, the launch rhythm of many automakers' new products has also been affected, and they have accelerated their self-research and development efforts.
On November 6 last year, He Xiaopeng, chairman and CEO of XPeng Motors, showcased the company's self-developed "Turing AI chip". The chip has a 40-core processor, customized for AI large models, and has the potential for applications in AI cars, AI robots, flying cars, and other fields. In September 2023, NIO also announced its self-developed intelligent driving chip and declared successful tape-out in July of the same year. Li Auto is also advancing its self-developed chip project.