The path to next-generation autonomous vehicles accelerates with a new announcement from NVIDIA. The tech company has launched pre-orders for its highly anticipated DRIVE AGX Thor Developer Kit, a platform designed for developers and researchers worldwide to create, test, and deploy intelligent mobility and autonomous driving systems adhering to automotive safety standards.
Initial units are set to begin shipping in September, with automakers, Tier 1 suppliers, and mobility startups already confirming their interest in this development kit.
A platform built for the new era of AI in automotive technology
The DRIVE AGX Thor Developer Kit combines the power of the NVIDIA Blackwell architecture, next-generation Arm Neoverse V3AE CPUs, and the DriveOS 7 operating system. It’s designed to handle increasingly complex AI models—from vision and language systems to generative reasoning models.
Key technical features include:
– Computing Power: up to 2,000 FP4 TOPS (equivalent to 1,000 INT8 TOPS) to run multiple AI inference loads in parallel.
– Memory and Bandwidth: 64 GB of LPDDR5X memory with a bandwidth of up to 273 GB/s.
– Integration Capabilities: inputs and outputs prepared for 360° cameras, radars, lidars, and industry-standard communication buses (GbE/10GbE and PCIe).
– Security Certifications: compliant with ISO 26262 (ASIL-D) and ISO/SAE 21434 standards crucial for safety and cybersecurity in automotive systems.
The kit will be available in two configurations: one for bench tests (SKU 10) and another tailored for deployment within actual vehicles (SKU 12).
An expanding ecosystem
The announcement is complemented by industry validation, with NVIDIA confirming that major automotive giants such as BYD, GAC, IM Motors, Li Auto, Volvo Cars, Xiaomi, and Zeekr are building their future systems on this platform.
In freight transportation, autonomous truck companies like Aurora, Gatik, PlusAI, and Waabi have also adopted DRIVE AGX Thor. Furthermore, software players including DeepRoute.ai, Nuro, WeRide, and ZYT are integrating the kit into their autonomous driving platforms.
Level 1 suppliers like Continental Automotive, Desay SV, Lenovo, Magna, and Quanta are already offering production systems based on DRIVE AGX Thor.
End-to-end security: from cloud to vehicle
A key aspect of this platform is its integration with NVIDIA Halos, an all-encompassing security system combining hardware, software, and AI models to ensure reliability from the cloud to the vehicle.
Halos’ approach cascades across three levels:
1. Technological: platform-level security, algorithms, and ecosystem.
2. Development: rules for design, deployment, and validation to ensure security compliance at every stage.
3. Computational: seamless connection of training, simulation, and deployment workflows—using NVIDIA DGX for training, Omniverse and Cosmos in OVX for simulation, and DRIVE AGX for final deployment.
Comparison with the previous generation: Thor versus Orin
The leap from NVIDIA’s previous platform, DRIVE AGX Orin, is significant.
| Feature | DRIVE AGX Orin | DRIVE AGX Thor |
| ——– | ————– | ————– |
| GPU Architecture | Ampere | Blackwell |
| INT8 Power | 254 TOPS | 1,000 TOPS |
| FP4 Power | – | 2,000 TFLOPS |
| CPU | 12-core Cortex-A78A | Arm Neoverse V3AE |
| RAM | 32 GB LPDDR5 | 64 GB LPDDR5X |
| Bandwidth | 200 GB/s | 273 GB/s |
| Cameras | 16 GMSL 2 | 16 GMSL 2 + 2 GMSL 3 |
| Ethernet | 30 Gb/s | 76 Gb/s |
This progression is especially evident in computational capacity, memory, and connectivity—all critical for executing next-gen AI models and processing vast high-resolution sensor data streams in real time.
Beyond automotive: intelligent mobility and robotics
While its primary focus is automotive, NVIDIA emphasizes that the DRIVE AGX Thor Developer Kit is also being used as a foundational platform for advanced robotics and intelligent mobility applications. The company has linked its unveiling to its Jetson AGX Thor product family and Jetson T5000 modules, already employed in next-generation robots and physical AI applications.
The future of autonomous driving
The introduction of the DRIVE AGX Thor Developer Kit affirms NVIDIA’s strategic commitment to becoming the backbone of autonomous driving technology. It’s not just about providing powerful hardware but fostering a comprehensive ecosystem—from simulation to validation and deployment.
With this move, NVIDIA positions itself ahead of competitors in an industry keen to accelerate the transition to safe, certified autonomous systems.
Pre-orders are now open, with the first units expected in September—a milestone likely signaling a new chapter in the race toward autonomous vehicles.
Frequently Asked Questions
– What is the NVIDIA DRIVE AGX Thor Developer Kit?
It’s a hardware and software development kit for creating, testing, and deploying autonomous driving and intelligent mobility systems.
– How does DRIVE AGX Thor differ from DRIVE AGX Orin?
Thor offers up to four times the computational power of Orin, more memory and bandwidth, and new security and connectivity certifications.
– When will it be available?
Pre-orders are open now, with initial units shipping in September 2025.
– Who are NVIDIA’s main partners in this project?
Automakers like Volvo, BYD, Xiaomi; autonomous truck companies like Aurora and Waabi; and system suppliers such as Continental and Magna.
via: https://blogs.nvidia.com/blog/drive-agx-developer-kit-general-availability/ and https://developer.nvidia.com/drive/agx#section-where-to-buy