At its GTC conference in Washington, D.C., NVIDIA announced a deep collaboration with Uber to expand the world's largest Level 4 (L4) autonomous driving-ready mobile network. This collaboration will combine Uber's future next-generation Robotaxi and autonomous delivery fleet with NVIDIA's latest DRIVE AGX Hyperion 10 autonomous driving development platform and DRIVE AV software stack designed specifically for L4 autonomous driving.
NVIDIA helps Uber integrate human-robot fleet, aiming for 10 vehicles
NVIDIA's goal is to accelerate the growth of the Level 4 ecosystem through its platform, supporting Uber in gradually expanding its global autonomous vehicle fleet to 10 vehicles starting in 2027. These vehicles will be jointly developed by Uber, NVIDIA, and other ecosystem partners, and will utilize NVIDIA DRIVE technology.
In addition, NVIDIA and Uber will also work together to develop a data factory based on the NVIDIA Cosmos world-based model development platform to process the large amounts of data required for self-driving car development.
Uber's strategy is to integrate human drivers and autonomous vehicles into a single operating network, providing a unified ride-hailing service that includes both human and robotic drivers. NVIDIA emphasizes that this network, powered by DRIVE AGX Hyperion-ready vehicles and an AI ecosystem, will seamlessly connect today's human drivers with future autonomous vehicle fleets.
NVIDIA founder and CEO Jensen Huang stated, "Robotaxi symbolizes the beginning of a global transportation transformation, and together we are creating a framework for the entire industry to deploy autonomous vehicle fleets powered by NVIDIA AI infrastructure at scale."
Uber CEO Dara Khosrowshahi responded: "NVIDIA is the backbone of the AI era. Self-driving cars will improve our cities, and we are excited to partner with NVIDIA to realize this vision."
The DRIVE L4 ecosystem expands: Stellantis, Lucid, and Mercedes-Benz join the passenger transport sector, and there are also gains in freight transport.
Several global automakers, Robotaxi operators, and Tier 1 suppliers have joined NVIDIA and Uber's L4 ecosystem:
• Stellantis:They are developing AV-Ready Platforms, which will integrate NVIDIA's full-stack AI technology and collaborate with Foxconn on hardware and system integration.
• Lucid:It is promoting Level 4 autonomous driving capabilities for its next-generation passenger cars and will introduce the NVIDIA AV software stack based on the DRIVE Hyperion platform into its US models.
• Mercedes-Benz: Based on its MB.OS operating system and DRIVE AGX Hyperion, it is testing future collaborations with industry-leading partners. Its new S-Class already offers an L4 driver assistance experience.
NVIDIA and Uber will also continue to support existing L4 ecosystem partners, including Avride, May Mobility, Momenta, Nuro, Pony.ai, Wayve, WeRide, and others.
In the freight truck sector, Aurora, Volvo Autonomous Solutions, and Waabi are also using the NVIDIA DRIVE platform to develop L4 autonomous trucks. Their next-generation system based on NVIDIA DRIVE AGX Thor will accelerate Volvo's future L4 fleet deployment.
DRIVE AGX Hyperion 10: A universal platform for L4-Ready vehicles
The core hardware announced this time is the NVIDIA DRIVE AGX Hyperion 10 production-grade reference platform. Its positioning is to enable any vehicle to reach L4-ready status, providing a unified hardware and sensor foundation, allowing automakers to build vehicles that can be equipped with any compatible autonomous driving software.
• Core operations:It is equipped with two DRIVE AGX Thor automotive platforms based on the NVIDIA Blackwell architecture, each providing more than 2000 FP4 TFLOPS (1000 INT8 TOPS) of real-time computing power.
• Sensor kit:It includes 14 high-resolution cameras, 9 radars, 1 LiDAR, and 12 ultrasonic sensors.
• Software:Includes the security-certified NVIDIA DriveOS operating system.
• design:Modular, customizable design and pre-verified sensor architecture accelerate development and reduce costs.
The DRIVE AGX Thor is optimized to run Transformer, Visual Language Action (VLA) models, and generative AI workloads, supporting the high-complexity computations required for L4 autonomous driving.
Generative AI and foundational models are reshaping autonomous driving, releasing the largest dataset of autonomous vehicles.
NVIDIA emphasizes that its autonomous driving solution is being trained on trillions of miles of real and synthetic driving data, utilizing basic AI models, large language models, and generative AI.
• Inference VLA Model:By combining visual understanding, natural language reasoning, and action generation, autonomous vehicles (AVs) can acquire human-like understanding capabilities to cope with complex real-world road conditions. Foretellix is collaborating to integrate its Foretify Physical AI toolchain with NVIDIA DRIVE for testing and validating VLA models.
• Largest dataset of autonomous vehicles released:NVIDIA has simultaneously released the world's largest multimodal autonomous vehicle dataset, which includes 1700 hours of real-world camera, radar, and LiDAR data from 25 countries, which will accelerate the development and validation of basic autonomous driving models.
NVIDIA Halos: Establishing a New Standard for Entity AI Safety Certification
To ensure the safety of autonomous vehicles, NVIDIA announced the launch of NVIDIA Halos, providing a holistic security framework from the cloud to the vehicle.
• Halos AI System Testing Laboratory:Dedicated to AI safety and cybersecurity in the automotive and robotics sectors, it performs independent assessments and oversees the new Halos Certified Program.
• First batch of members:AUMOVIO, Bosch, Nuro, Wayve, and others became the first batch of laboratory members to receive ANSI accreditation.
This plan aims to accelerate the safe and large-scale deployment of L4 autonomous driving and other AI systems.
