Nvidia showcased its new point-to-point Level 2 driver-assist system in a Mercedes-Benz CLA sedan during a demonstration in San Francisco, positioning it as a competitor to Tesla's Full Self-Driving. The system handled complex urban driving confidently, navigating traffic and pedestrians without issues. Company executives outlined an ambitious roadmap for broader rollout starting in 2026.
On a clear day in San Francisco, a Mercedes-Benz CLA sedan equipped with Nvidia's upcoming autonomous driving technology cruised through busy streets for about 40 minutes. The vehicle, using Nvidia's AI-powered Level 2 (L2) system alongside Mercedes' cameras and radar, managed traffic signals, four-way stops, double-parked cars, and unprotected left turns. It even executed a wide right turn to avoid a blocking truck while allowing pedestrians to cross.
The demonstration highlights Nvidia's push into automotive autonomy, aiming to expand its small automotive division, which generated $592 million in the third quarter—1.2 percent of the company's $51.2 billion total revenue. Xinzhou Wu, head of Nvidia's automotive division, emphasized the company's decade-long investment in a full-stack solution, including the Drive AGX system-on-a-chip running on the Blackwell GPU architecture, capable of 1,000 trillion operations per second. "Jensen always says, the mission for me and for my team is really to make everything that moves autonomous," Wu stated.
Nvidia's roadmap includes releasing L2 highway and urban capabilities, such as automated lane changes and traffic signal recognition, in the first half of 2026. Urban features will expand to autonomous parking in the second half, covering the entire United States by year-end. A small-scale Level 4 trial is planned for 2026, followed by partner robotaxi deployments in 2027 and personal autonomous vehicles in 2028. Level 3 highway driving, allowing hands-off and eyes-off operation under conditions, is targeted for 2028.
Partners like Mercedes, Jaguar Land Rover, and Lucid Motors will integrate the customizable system, which uses reinforcement learning to improve over time. Ali Kani, Nvidia's VP and general manager for automotive, noted its comparability to Tesla's FSD in city driving tests, with similar driver takeover rates. "We’re coming fast," Kani said. "I’d say [we’re] very close [to FSD]." The system prioritizes safety with redundancy, though experts remain skeptical of Level 3 viability, and adoption depends on automakers' confidence and regulatory clarity.
