NVIDIA Unveils DRIVE AV Full-Stack Autonomous Driving System, Launching an "Android-Style" Open Ecosystem

Last Thursday, NVIDIA released a 22-minute autonomous driving demonstration video. CEO Jensen Huang rode alongside Wu Xinzhou, head of NVIDIA’s China team, in a vehicle equipped with the DRIVE AV system, which handled the entire journey autonomously—navigating complex scenarios such as construction zones, illegally parked vehicles, and narrow lanes marked by traffic cones—without any human intervention. The system is built on NVIDIA’s Thor chip and features a four-layer “perception-decision-control” architecture. It innovatively introduces a “dual-brain” mechanism, running an end-to-end AI model in parallel with a traditional rule-based safety stack to ensure both safety and flexibility. DRIVE AV offers a complete ecosystem—from chips and sensors to algorithms, simulation testing, and over-the-air (OTA) updates—enabling rapid deployment of Level 2 to Level 4 autonomous driving solutions. For overseas automakers, this represents a turnkey solution that significantly lowers the cost and technical barriers of in-house development. Amid Tesla's dominance and restrictions on Chinese autonomous driving solutions going global, NVIDIA is positioning itself as the “Android platform” for intelligent driving worldwide, providing traditional automakers with a critical enabler for their smart transformation.