Nvidia has unveiled its new $3,499 robot “brain,” Jetson Thor, designed to bring AI decision-making directly to industries ranging from manufacturing to healthcare, aligning with CEO Jensen Huang’s vision of “physical AI.”
During his keynote address at Nvidia’s annual GTC conference in March, Huang outlined the progression of AI through four distinct phases. The first wave, perception AI, which began roughly a decade ago, focused on the ability to recognize speech and classify images. The second wave, generative AI, characterized by large-language models like ChatGPT, has dominated the last five years, producing text and images based on predictive patterns. The current wave, agentic AI, allows AI models to reason and perform tasks independently. Huang posits that physical AI represents the culmination of these advancements, integrating AI into tangible, real-world applications.
The Jetson Thor developer kit, priced at $3,499, is slated to begin shipping next month. The announcement saw a slight increase in Nvidia’s stock, which further rose on Tuesday morning. Powered by Nvidia’s Blackwell chips, Jetson Thor promises “unmatched performance and scalability” to handle the computational demands of generative AI models. According to Nvidia, Jetson Thor offers “up to 7.5x higher AI compute and 3.5x better energy efficiency” compared to its predecessor. The system is designed to enable robots to perceive and respond to their surroundings in real time, a crucial capability for expanding AI adoption in the physical world.
Jetson Thor is a compact computer intended to be installed within a robot, enabling it to simultaneously run multiple AI models for perception, understanding, and action without relying on cloud connectivity. Nvidia likens it to integrating a foreman, safety officer, and navigator into a single system, allowing a robot to identify hazards, reroute around obstacles, and maintain operational efficiency. The combination of on-board power and software allows robots to manage multiple tasks concurrently, similar to a driver checking mirrors, listening for sirens, and changing lanes, resulting in smoother movements, faster decision-making, and reliable performance.
Nvidia is actively expanding its AI capabilities beyond training chatbots to include applications in the physical world. This expansion broadens Nvidia’s target market to include sectors such as logistics, manufacturing, healthcare, construction, retail, and autonomous systems, all of which prioritize uptime and safety. Nvidia’s announcement highlighted early adopters and evaluators across e-commerce, industrials, and tech, suggesting potential near-term pilot programs that could lead to larger orders.
For robotics engineers, Jetson Thor offers reduced latency and fewer errors. It enables robots to react in milliseconds, making the difference between successfully catching a falling object and dropping it, or navigating around an obstacle instead of colliding with it. This responsiveness is crucial because every second saved accumulates across numerous tasks, while safety incidents are costly and disruptive. Nvidia emphasizes the system’s power efficiency and ability to handle multiple AI tasks simultaneously, reducing the number of computers required per robot and simplifying system design.
Companies are currently adopting Jetson Thor, supported by a developer ecosystem built around Nvidia’s Isaac tools to accelerate prototyping and deployment. Business leaders can initiate trials using the developer kit and then scale via production modules if pilot programs demonstrate improved efficiency or safety. Nvidia’s long-term goal is to make robots as commonplace and dependable as other capital equipment, with AI handled locally to ensure speed, capability, and controllability.
Fortune utilized generative AI to assist with the initial draft of the article, which was subsequently verified for accuracy by an editor before publication.
