Nvidia has launched the Jetson AGX Thor, a new robotics computing module described as a "robot brain," designed to advance the capabilities of humanoid and general robots across various industries including manufacturing, logistics, healthcare, and construction. Powered by Nvidia's Blackwell GPU architecture, the Jetson Thor delivers up to 2,070 FP4 teraflops of AI compute, representing a 7.5-fold increase in AI processing power and 3.5 times greater energy efficiency compared to its predecessor, the Jetson Orin. The module also offers 3.1 times higher CPU performance and twice the memory capacity, with 128GB of memory onboard. Priced at $3,499, the Jetson Thor is available as a developer kit and production modules, with shipments expected to begin next month. Nvidia emphasizes the device's real-time reasoning and inference capabilities, enabling robots to process high-speed sensor data and perform autonomous decision-making on-device without reliance on cloud computing. Early adopters include major technology firms such as Amazon, Meta, Agility Robotics, and Boston Dynamics. The Jetson Thor aims to accelerate the era of general robotics and physical AI, supporting applications in humanoid robots, industrial automation, surgical machines, and autonomous vehicles.