NVIDIA GTC 2026 Focuses on Scaling Autonomous Driving, Unveils Three Major System Upgrades

At the GTC 2026 conference, Wu Xinzhao, Vice President of NVIDIA Automotive, stated that autonomous driving is transitioning from technological validation to large-scale deployment. The company is no longer focusing solely on computing power; instead, it unveiled three key updates centered around its DRIVE Hyperion platform: the inference model Alpamayo 1.5, which can understand natural language commands and explain its decisions; the unified safety architecture Halos OS, certified to automotive-grade ASIL D standards; and simulation tools NuRec and Cosmos, which enhance scenario generation and testing efficiency, enabling up to 2 million simulations per day. Meanwhile, NVIDIA is accelerating collaborations with automakers such as BYD, Geely, and Mercedes-Benz, as well as mobility platforms like Uber, to advance mass production of Level 4 (L4) projects and global fleet deployment. Overall, this strategy underscores a shift in the autonomous driving race toward engineering execution, standardization, and continuous iteration capabilities.