At the NVIDIA GTC 2026 conference held on March 17, NVIDIA announced partnerships with automakers including BYD (比亚迪), Geely (吉利), Nissan, and Isuzu to accelerate the development of L4 autonomous driving models and projects using the NVIDIA DRIVE Hyperion platform.
DRIVE Hyperion is not a single chip but a complete reference solution encompassing the vehicle-side computing platform, sensor architecture, network connectivity, and security systems. This platform acts as a mature intelligent driving foundation, allowing automakers to develop L4 solutions without starting from scratch.
To ensure the absolute safety of L4 systems, NVIDIA introduced the Halos OS unified security architecture, which utilizes a three-layer structure to integrate security middleware with deployable applications and includes an active safety stack compliant with NCAP five-star standards. Several ecosystem partners, including Quanta (广达), Hesai (禾赛), and Valeo, have joined the NVIDIA Halos AI system testing laboratory to co-develop autonomous driving safety standards.
Furthermore, NVIDIA released the open-source model AIpamayo 1.5, which processes driving videos, navigation commands, and natural language prompts to output driving trajectories with reasoning logic. This model is designed to help developers better handle rare and complex road conditions, enabling vehicles to learn more efficiently. For the testing and verification of L4 autonomous driving, NVIDIA (英伟达) launched Omniverse NuRec based on 3D Gaussian Splatting technology, which can import real-world data to rapidly reconstruct and render 3D simulation scenes, thereby reducing modeling costs and allowing for faster stress testing of edge scenarios.
While L4 autonomous driving is still some distance away from large-scale entry into the general passenger car market, it is more likely to first operate in specific areas and limited scenarios. NVIDIA's collaboration with major automakers like BYD and Geely signals that the race for high-level intelligent driving continues to advance.