In a move that underscores its strategy to become the central nervous system for self-driving cars, Nvidia has signed deals with Chinese automotive giants BYD and Geely. Announced at the company's GTC conference, the agreements will see the two manufacturers adopt Nvidia's Drive Hyperion platform—a full-stack hardware and software suite designed for Level 4 autonomous vehicles.

BYD, already a customer for Nvidia's chips in conventional cars, will now use Hyperion for its next-generation robotaxis. Geely plans to integrate Nvidia's powerful Thor system-on-a-chip into vehicles for its Zeekr brand, which already supplies cars to Alphabet's Waymo. Nvidia's automotive vice president, Ali Kani, confirmed Waymo also uses Nvidia technology "in the car and the cloud."

The partnerships arrive amid persistent U.S.-China trade tensions. While Washington recently allowed sales of Nvidia's advanced H200 data center chips to Chinese firms, supplying the core technology for autonomous vehicles to China's leading EV makers represents a significant step. It could accelerate China's robotaxi development, a field where it competes closely with American efforts from companies like Waymo.

For Nvidia, the automotive unit remains a small fraction of its staggering AI-driven revenue, which hit $51.2 billion last quarter. Yet the company is weaving a broad web in autonomy. Beyond the Chinese deals, Nissan, Isuzu, and ride-hailing firm Lyft will also use the Hyperion platform. Lyft aims to leverage the tech for machine learning and scaling its AV operations. Nvidia's existing partnership with Uber targets a global fleet of 100,000 robotaxis by 2028.

Kani argued Nvidia's progress stems not from road miles, but from simulation and AI models. He pointed to its open-source Alpamayo models and synthetic data generation as key advantages. When asked how many companies are building Level 4 systems on Nvidia, he replied, "I think it's pretty much everyone," listing over a dozen major players.

Addressing public safety concerns, Nvidia introduced Halos OS, described as a "safety guardrail" designed to intervene if an AI model nears an unsafe decision. "The system is still architected to take you to a safe place," Kani said, even if a component fails.

Source: The Verge