Nvidia Launches Full-Stack Autonomous Vehicle Software Platform in Full Production

Nvidia has announced that its full-stack autonomous vehicle software platform, Nvidia Drive, is now in full production, aiming to accelerate large-scale deployment of self-driving transport globally across passenger cars, trucks, and robotaxis. The platform leverages deep learning and foundation models trained on extensive human driving data to enable human-like decision-making and supports a modular, scalable approach for automakers to adopt either the entire stack or subsets, including features for Level 3 automation such as surround perception and automated lane changes. Nvidia Drive is supported by a comprehensive three-computer solution encompassing DGX systems and GPUs for AI model training, the Omniverse and Cosmos platforms for simulation and synthetic data generation, and the in-vehicle DRIVE AGX computer for real-time sensor processing. Additionally, Nvidia has won the Autonomous Driving Grand Challenge at the 2025 Computer Vision and Pattern Recognition conference for the third consecutive year, recognized for its Generalized Trajectory Scoring method that improves trajectory prediction safety and compliance in dynamic simulations. These advancements underscore Nvidia's commitment to developing smarter, safer autonomous vehicles and facilitating the transition to higher levels of vehicle automation as regulations evolve. The company also emphasizes safety through its Nvidia Halos system, which integrates hardware, software, AI models, and tools to ensure safe AV development and deployment.