On August 25, 2025, NVIDIA unveiled Jetson AGX Thor, a groundbreaking AI compute module described as the “ultimate engine for physical AI.” With up to 2070 TFLOPS of AI performance, this next-generation platform sets a new benchmark for robotics, humanoid machines, autonomous driving, and industrial applications.

Pricing That Disrupts the Industry

NVIDIA surprised the market with aggressive pricing. The Jetson AGX Thor Developer Kit starts at $3,499, while the Thor T5000 module for mass production drops to $2,999 per unit when ordered in volumes above 1,000. Combined with NVIDIA’s promise of next-month shipping, developers can move from prototype to production faster than ever before.

Powered by the Blackwell GPU Architecture

The leap in performance comes from the Blackwell GPU architecture, already known for driving AI supercomputing. Jetson AGX Thor integrates:

* 14-core Arm Neoverse-V3AE CPU

* 2,560 CUDA cores

* 96 fifth-generation Tensor Cores

* 2070 TFLOPS peak AI performance (FP4 precision)

Compared to the previous Jetson Orin, Thor delivers 7.5x more performance and up to 3.5x better energy efficiency within a 40–130W power range. Robots can now achieve real-time perception, decision-making, and interaction with extended battery life.

Full-Stack Software and Model Compatibility

Jetson AGX Thor is not just about hardware. It comes with NVIDIA’s full software ecosystem:

* Isaac platform for robot training and simulation

* Metropolis AI for vision and video analytics

* Holoscan for sensor data processing

* CUDA acceleration for continuous software-driven performance gains

It also supports leading AI models including Google Gemini, Meta Llama, OpenAI, Mistral AI, Alibaba Qwen, and ByteDance frameworks — ensuring developers face no compatibility barriers.

Applications Across Industries

With unmatched compute and software integration, Jetson AGX Thor enables new levels of intelligence across industries:

* Humanoid robots: Ubtech and Unitree are integrating Thor into next-generation humanoid robots for industrial and service tasks.

* Autonomous driving: BYD, Li Auto, and Xiaomi have already lined up to adopt DRIVE AGX Thor, optimized for automotive safety and real-time decision-making.

* Medical robotics: United Imaging is exploring Thor for surgical precision and real-time diagnosis.

* Smart transportation: Companies like Wanji Technology leverage Thor for intelligent traffic monitoring.

NVIDIA's Expanding Robotics Ecosystem

Jetson AGX Thor benefits from NVIDIA's 200,000+ developers, 150+ ecosystem partners, and 7,000+ customers built over the past decade. From Boston Dynamics to Amazon and Meta, Jetson platforms have powered leading innovations. Thor now extends this legacy with supercharged compute power and ready-to-deploy software.

A New Era for General-Purpose Robots

With 72% year-over-year revenue growth in its robotics and automotive division (Q2 2025 revenue: $567M), NVIDIA has clearly positioned robotics as its next growth curve. The launch of Jetson AGX Thor accelerates the path toward general-purpose humanoid robots and physical AI, moving the industry from limited automation to human-like adaptability.

Conclusion

The NVIDIA Jetson AGX Thor marks a turning point in robotics and autonomous systems. By combining record-breaking 2070 TFLOPS compute, breakthrough affordability, and full software ecosystem, Thor is not just another chip — it is the supercomputer brain for the next generation of robots.