Nvidia announced at its annual GTC conference that two of China's biggest carmakers, BYD and Geely, will adopt its Drive Hyperion self-driving taxi platform, marking the chip giant's attempt to solidify its key position in the global self-driving market. In addition to BYD and Geely, Nvidia also confirmed that Nissan and Isuzu will also use the platform to develop Level 4 autonomous vehicles.
Currently, BYD already uses NVIDIA chips in its fuel vehicles and electric vehicles, and this cooperation will be extended to the development of a new generation of Level 4 autonomous driving models based on the Drive Hyperion platform. Geely, on the other hand, is said to be using NVIDIA Thor chips in its new JK brand models and supplying JK vehicles to the US driverless travel company Waymo for the latter's self-driving taxi service. Ali Kani, head of Nvidia's automotive business and the company's vice president, said at a media briefing that Waymo uses Nvidia's products not only in the car but also in the cloud. "Waymo is using us both in the car and in the cloud."
The news that Nvidia is providing self-driving chips and software to China's two major car companies comes at a time when tensions between China and the United States on trade and tariff issues continue to rise. In particular, high-performance chips used for training artificial intelligence models in data centers have long been the focus of competition between the two parties. The Trump administration has recently approved the export of Nvidia H200 chips to Chinese customers. In terms of electric vehicle production, China has a clear advantage over the United States, but in the field of self-driving taxis, the current competition between the two countries is closer. For example, Baidu has launched commercial self-driving taxi operations in more than ten cities in China, while Waymo has approximately 3,000 vehicles providing commercial services in 10 cities in the United States.
Nvidia's cooperation with BYD and Geely is seen as expected to significantly accelerate the implementation of autonomous driving products by Chinese car companies, thereby increasing China's chances of surpassing the United States in this field. Some members of the U.S. Congress have once again called for the advancement of the long-stalled autonomous driving legislation. One of the reasons is to maintain its technological lead over China.
Although NVIDIA has long been a chip and software supplier for advanced assisted driving systems of many mainstream car companies, the scale of its automotive business is still far lower than that of AI-related business. In the third quarter of 2025, Nvidia's total revenue was US$51.2 billion, of which the automotive segment contributed US$592 million, accounting for only about 1.2%. In order to further strengthen its voice in the field of autonomous driving, in addition to cooperating with Chinese car companies, Nvidia will also provide Nissan with its Hyperion platform, which will also use self-driving taxi software developed by the British company Wayve. NVIDIA is also working with Isuzu and Tier IV to build Level 4 autonomous buses using next-generation Drive AGX Thor-based SoCs.
In terms of travel platform, Lyft announced that it will use the NVIDIA Hyperion platform to build its own self-driving taxi fleet, with vehicles and software coming from multiple partners. Lyft said it will use Nvidia technology to enhance its machine learning capabilities and accelerate enterprise-level operations for its ride-hailing and autonomous driving businesses. Nvidia has previously partnered with Uber, Lyft’s main competitor, and the two parties plan to deploy up to 100,000 self-driving taxis globally by 2027. Nvidia’s latest disclosure states that this plan is expected to cover 28 markets on four continents by 2028, with Los Angeles and San Francisco taking the lead in early 2027. Uber is working with Lucid, Volkswagen, Stellantis and other car companies, which are all using Nvidia products to build self-driving vehicles.
Facing the first-mover advantage accumulated by Waymo and Tesla in terms of real road mileage, Nvidia regards its virtual testing capabilities and its open source AI model combination Alpamayo as the key to its breakthrough. Ali Kani said that Nvidia's autonomous driving technology stack took more than ten years to build, and the new generation leap benefits from general inference models such as Alpamayo, as well as synthetic data generation and testing capabilities based on Omniverse, NuRec and Cosmos. He said that almost all autonomous driving companies with development level 4 are currently using NVIDIA products, and named companies such as Waymo, Nuro, Waabi, Zoox, Wayve, Momenta, Pony.ai, WeRide, Baidu, DeepRoute and ZYT.
As more self-driving taxis hit the road, public concerns about safety are growing. Tesla vehicles equipped with Level 2 Autopilot Assist have been involved in hundreds of accidents, including 23 injuries and at least two fatalities. Waymo's vehicles have been photographed violating traffic rules around school buses, and have also been "stuck" at intersections, causing serious congestion. In response to the risks posed by these so-called "edge scenarios," NVIDIA has launched a new product called Halos OS to help its Level 4 autonomous driving partners build safer systems. Kani describes Halos OS as "safety guardrails" for self-driving systems, intervening when an AI model is about to make an unsafe decision.
He said that NVIDIA helps partners build a safety architecture that can be decomposed into each function to ensure that even if any computing unit or sensor fails, the entire system can still bring the vehicle to a safe area. Driven by the joint efforts of Chinese car companies, U.S. technology companies and global travel platforms, NVIDIA is trying to transform from an "autonomous driving solution provider" to a core infrastructure supplier in the era of self-driving taxis. Its full-stack layout in chips, software, virtual testing and safety systems will continue to undergo double inspections by the market and regulation in the next few years.