NVIDIA partners with Uber, Foxconn, and Stellantis to build the world's largest Level 4 autonomous vehicle ecosystem.

NVIDIA ( announced on 10/28 that it will collaborate with Uber, Stellantis, and Foxconn ) to jointly create the world's first mass-producible Level 4 autonomous vehicle fleet. Subsequently, Stellantis announced the collaboration details in Amsterdam, with testing expected to start in 2027 and mass production scheduled for 2028.

NVIDIA announces a global autonomous driving alliance, with Uber, Foxconn, and Stellantis joining first.

NVIDIA CEO Jensen Huang ( announced at the latest GTC conference that they will establish a global Level 4 autonomous driving ecosystem centered around their latest generation DRIVE AGX Hyperion 10 platform.

This project involves multiple global car manufacturers and ride-hailing platforms, such as Lucid, Mercedes-Benz, Aurora, Volvo, Pony.ai, etc., which are also developing self-driving solutions compatible with the Hyperion architecture. Below are the technical and practical responsibilities of each main partner:

Uber is laying out a global self-driving network, aiming to deploy 100,000 self-driving cars by 2027.

According to the cooperation content, Uber will use the Nvidia Hyperion 10 architecture as the core to establish the world's largest self-driving ride-hailing network, integrating human drivers and self-driving cars on the same platform, and plans to officially deploy the self-driving fleet starting in 2027.

The target number of vehicles built using NVIDIA technology in the first batch is 100,000, and an AI data factory will be established to conduct autonomous driving training and data processing using the NVIDIA Cosmos platform.

Stellantis joins the alliance to develop the first Level 4 Robotaxi platform.

Dutch automotive giant Stellantis announced the signing of a memorandum of cooperation with NVIDIA, Uber, and Hon Hai on the same day, aiming to jointly develop a mass-producible Level 4 autonomous vehicle platform, initially focusing on commercial vehicles and small passenger cars. These vehicles will be equipped with NVIDIA DRIVE AGX Hyperion 10 architecture, integrating the DriveOS operating system and DRIVE AV full-stack software, supporting automated parking and fully driverless modes.

The first batch of the fleet is planned to be operated by Uber, with an expected deployment of 5,000 vehicles in U.S. cities, and mass production starting in 2028.

) Note: Fully autonomous driving Level 4 refers to highly automated driving, which is fully self-driving within designated areas and does not require human intervention (.

Hon Hai plays the role of hardware hub, integrating NVIDIA HPC and sensing systems.

In this alliance, Hon Hai will assist Stellantis in integrating vehicle hardware and electronic systems, providing high-performance computing )HPC( modules and sensor layouts to ensure production consistency and safety standards during global deployment.

NVIDIA emphasizes that this collaboration is not just about supplying chips, but providing a complete technical backbone ranging from AI computing, sensing architecture to generative model training, creating an integrated platform for autonomous vehicles that includes an “AI intelligent core, environmental sensing system, and decision-making hub.”

NVIDIA Technology Core: Hyperion 10 and DRIVE Thor

NVIDIA's next-generation DRIVE AGX Hyperion 10 is the technological backbone of this global autonomous driving alliance, utilizing the Blackwell architecture chip DRIVE AGX Thor. Each vehicle is equipped with two sets of processors, achieving a computing power of up to 2,000 FP4 teraflops ), approximately 1,000 TOPS (. Hyperion 10 will integrate:

14 sets of high-definition cameras

9 sets of millimeter wave radar

1 Optical Radar )LiDAR(

12 sets of ultrasonic sensors

Hyperion 10 features a modular design that allows different manufacturers to quickly integrate and can be updated online )OTA(.

In addition, NVIDIA has introduced the visual language action model )VLA( in generative AI, allowing vehicles to instantly understand road conditions, semantics, and human behavior. For example, during sudden traffic changes or unstructured intersections, the autonomous driving system can respond in a “human logic” manner.

NVIDIA establishes security certification standards, Halos project goes live.

NVIDIA also announced the launch of the “NVIDIA Halos” AI Safety Certification Program, becoming the industry's first AI system standard specifically for assessing the safety of autonomous driving and robotics. The first batch of partners includes several AI autonomous driving companies such as AUMOVIO, Bosch, Nuro, Wayve, and others.

This project is verified by the NVIDIA Halos AI System Testing Laboratory to ensure that all self-driving cars and robotic products comply with physical AI safety and cybersecurity standards.

The public AI training database is released, providing over 1,700 hours of multi-national traffic footage.

To accelerate the training of autonomous systems in the industry, NVIDIA has simultaneously released the world's largest multimodal autonomous driving dataset, covering 25 countries, 1,700 hours of images, radar, and optical radar data.

This data will be used for AI model training and autonomous driving system validation, simulated and tested by the Foretellix platform, assisting partners in accelerating development within a safety framework.

This article discusses how NVIDIA partners with Uber, Hon Hai, and Stellantis to build the world's largest Level 4 autonomous vehicle ecosystem, first appearing in Chain News ABMedia.