High-Performance In-Vehicle Computing for Autonomous Vehicles

• Two DRIVE AGX Thor in-vehicle computers based on Blackwell architecture.

• Features 14 high-definition cameras, 9 radars, 1 lidar, and 12 ultrasonic sensors and a microphone array.

• Includes DRIVE™ AV software, purpose-built for L4 autonomy.

• Delivers more than 1,000 INT8 TOPS (2,000 FP4 FLOPs).

• Offers a scalable architecture supporting Level 2+ to fully autonomous driving.

• Provides superior safety, with ASIL-D compliance and redundancy.

• Delivers up to 254 TOPS of AI performance.

• Offers a scalable architecture supporting Level 2+ to fully autonomous driving.

• Provides superior safety, with ASIL-D compliance and redundancy.

• Open and Transparent AI: A portfolio of open AI models, simulation frameworks, and AI datasets that teams can inspect and adapt for transparent, reproducible AV development.
• Explainable Long‑Tail Autonomy: Vision‑language‑action models that reason through complex scenarios and verbalize decisions, making driving behavior interpretable and auditable.
• An Accelerated Path to Level 4: High‑fidelity, closed‑loop simulation that shifts miles from road to lab, reducing test time and accelerating validation.

• ASIL-D Safety Foundation: A production-proven, certified architecture that provides the base for the most advanced, mission-critical applications.

• Unified AI Acceleration: Seamlessly powers both high-performance autonomous driving and an immersive cabin AI experience through dedicated acceleration for LLMs and VLMs.

• Unparalleled Versatility for Software Deployment: Cloud-to-car deployment for both QNX and Linux while providing a holistic security framework for the era of Software Defined Vehicles.

• Features full-stack software that includes active safety, parking, and hands-off and eyes-off driving.

• Provides state-of-the-art software architecture using an end-to-end model for self driving that ensures natural driving behavior, even in complex scenarios.

• Features the NVIDIA Halos safety guardrail system that prioritizes both vehicle and occupant protection, from the cloud to car.

Advanced transformer for multi-frame, bird’s-eye-view 3D perception in autonomous driving.

End-to-end autonomous driving stack integrating perception, prediction, and planning with sparse scene representations for efficiency and safety.

Efficient 3D object detection for autonomous driving by propagating sparse object queries temporally.

NVIDIA is enabling a new era of AI-defined driving, bringing its level 2++ capabilities to U.S. roads this year, starting with Mercedes-Benz, a long-standing partner in advancing safe, intelligent mobility.

Toyota, the world’s largest automaker, will build its next-generation vehicles on NVIDIA DRIVE AGX Orin^TM, running the safety-certified NVIDIA DriveOS operating system.

GM, the largest U.S. automaker, will use NVIDIA AI, Omniverse and DRIVE AGX to develop and build next-gen vehicles, factories, and robots.

Starting in 2026, JLR vehicles will be built on the NVIDIA DRIVE AI-defined platform—from the cloud, built on NVIDIA data center solutions, to the car, powered by NVIDIA DRIVE AGX Orin.

The Volvo EX90 is built on the NVIDIA DRIVE AGX Orin system-on-a-chip (SoC), with plans to migrate to the NVIDIA DRIVE Thor computer, running DriveOS for its future fleet of cars.

Hyundai Motor Group will use the NVIDIA DRIVE Hyperion platform to accelerate the development of data-driven autonomous driving systems.