Nvidia has announced a strategic push into the autonomous vehicle market, revealing plans to provide its AI chips and self-driving software to companies operating robotic taxi fleets, with services slated to begin in 2027. The initiative was detailed during demonstrations and statements from company executives, including Vice President of Automotive Xinzhou Wu and CEO Jensen Huang.
The company's automotive and robotics division currently generates $592 million in quarterly revenue, representing about 1% of Nvidia's total sales. Despite its small size, Huang has identified robotics as the company's second most important category after artificial intelligence. "We imagine that someday, a billion cars on the road will all be autonomous," Huang stated at CES in Las Vegas, envisioning these vehicles functioning as either rental robotaxis or personal cars.
Nvidia's hardware cornerstone for this market is the Drive AGX Thor automotive computer, priced at approximately $3,500 per chip. The company markets this system as a way for automakers to reduce research costs and accelerate development timelines. Nvidia also provides access to its AI chips and simulation software for training self-driving models, working closely with manufacturers to customize the technology for specific vehicle models, including adjustments to acceleration patterns.
A key partnership is with Mercedes-Benz, which will integrate Nvidia's autonomous driving system into vehicles launching in late 2026. This system, demonstrated in San Francisco in December, enabled a 2026 Mercedes-Benz CLA to drive autonomously for 90% of an hour-long test. A safety driver intervened only once in a complex traffic scenario involving two buses and a Waymo vehicle. Nvidia classifies this technology as "Level 2 Plus Plus," placing responsibility on the driver, similar to Tesla's Full Self-Driving feature. Mercedes-Benz CEO Ola Källenius promised future "park-to-park" capabilities, stating, "Any parking situation that you feel is intimidating, that car will solve for you."
For the robotaxi segment, Nvidia is targeting Level 4 autonomy—where vehicles can operate without human intervention in specific geographic areas. The company has a previously announced partnership with Uber, formed in October 2024, though the specific launch partner for the 2027 robotaxi service was not named. Nvidia's strategy includes a dual-computer safety system in Drive-equipped cars: a primary vision-language model and a backup rule-based system that takes over when the primary system is uncertain.
Concurrently, Nvidia's CEO Jensen Huang announced that its next-generation Vera Rubin AI chip platform is in full production, claiming it delivers five times better AI computing performance than previous chips. This advancement in generative AI is seen as crucial for improving self-driving capabilities. "With transformers and generative AI, we can do much more," said Wu, outlining a vision for point-to-point self-driving for consumers by 2028 and eventually, voice-command responsive vehicles.
