At GTC 2026, NVIDIA CEO Jensen Huang once again demonstrated his strong appeal for the "mobility revolution," announcing that the world's four major automakers—BYD, Geely, Isuzu, and Nissan—have officially adopted the NVIDIA DRIVE Hyperion platform to build a new generation of Level 4 autonomous vehicles. At the same time, NVIDIA also announced an expanded partnership with Uber, planning to launch fully autonomous robotaxi services in Los Angeles and San Francisco in the first half of 2027, with plans to expand to 28 cities worldwide by 2028, signifying that the commercialization of autonomous vehicles has moved from proof-of-concept to large-scale deployment.
Automakers Embrace It Fully: From Passenger Cars to Buses, NVIDIA Platform Becomes Standard for Self-Driving Vehicles
In his keynote speech, Jensen Huang stated that BYD, Geely, and Nissan, three leading automakers, will adopt NVIDIA DRIVE Hyperion's production-ready computing and sensor architecture to develop next-generation Level 4 autonomous driving plans. Nissan, in particular, will combine Wayve's software technology to further enhance the learning capabilities of its autonomous driving system.
Isuzu and TIER IV will collaborate to develop Level 4 autonomous buses, using the NVIDIA DRIVE AGX Thor system-on-a-chip, extending autonomous driving technology to the commercial vehicle and public transportation sectors.
Hyundai Motor Group also announced an expanded strategic partnership with NVIDIA, integrating the NVIDIA DRIVE Hyperion platform to comprehensively deploy advanced driver assistance systems from Level 2+ to Level 4 Robotaxi design. Simultaneously, Hyundai will combine its Software-Defined Vehicle (SDV) capabilities and large-scale fleet data with NVIDIA's AI computing infrastructure, and explore cooperation with its autonomous driving joint venture, Motional, to accelerate innovation in next-generation autonomous mobility services.
Uber Robotaxi timeline released: Launch in two US cities in 2027, expanding to 28 cities by 2028.
In addition, NVIDIA and Uber have expanded their collaboration, announcing that they will build a fleet of autonomous vehicles powered entirely by the NVIDIA DRIVE Hyperion platform, creating one of the world's largest autonomous ride-hailing networks. The network will officially launch in Los Angeles and the San Francisco Bay Area in the first half of 2027, and will expand to 28 cities across four continents by 2028.
This end-to-end autonomous driving solution will integrate the NVIDIA DRIVE AV software stack, the NVIDIA Alpamayo open model, and the NVIDIA Halos operating system, accelerating the development and deployment of secure and scalable Robotaxi services globally. Other mobility service leaders such as Bolt, Grab, and Lyft also plan to accelerate their autonomous mobility initiatives with NVIDIA DRIVE Hyperion, demonstrating that the entire industry is moving towards a software-defined Robotaxi fleet.
A complete upgrade to the underlying technology: Alpamayo 1.5, Halos OS, and Omniverse NuRec
To support large-scale autonomous driving deployments from Level 2+ to Level 4, NVIDIA simultaneously released several key technology updates:
• NVIDIA Halos OS:This is a unified safety architecture designed specifically for AI-driven vehicles. Built on the ASIL D certified DriveOS, it integrates a safety intermediary layer and deployable safety applications through a three-layer safety architecture, providing a production-ready safety foundation for the Level 4 autonomous system on DRIVE Hyperion.
• NVIDIA Alpamayo 1.5:As a major upgrade to the open suite of autonomous driving AI models, Alpamayo 1.5 is an interactive and guided inference model. It can take driving videos, vehicle trajectory history, navigation guidance, and natural language prompts as input, and output a driving path with inferred trajectories. Developers can directly guide vehicle behavior through navigation and text prompts, allowing the vehicle to learn more effectively from rare or unpredictable events.
• NVIDIA Omniverse NuRec:This is a technology based on 3D Gaussian Splatting that can reconstruct interactive simulation environments from real-world data. It assists autonomous driving developers in stress testing reasoning behavior within highly realistic simulations, significantly reducing the time and cost required to manually construct worlds. Leading autonomous driving toolchain providers, including dSPACE and Foretellix, have already integrated NuRec.
