Why it matters: Nvidia has spent over a decade developing its Jetson platform for robotics and other AI workloads, but the recent surge in generative AI adoption has convinced the company that a robot renaissance is on the horizon. Nvidia claims that significant performance boosts from the new Jetson Thor development kit will enable dramatic leaps forward in reasoning and real-time performance, and numerous robotics companies have already invested in the next-generation module.
Production modules and development kits for the next generation of Nvidia's Jetson AI and robotics platform are now available. Early adopters include Amazon, Meta, John Deere, OpenAI, Boston Dynamics, and other companies hoping to start a boom in the robotics market similar to ChatGPT's impact on generative AI.
Based on Nvidia's Blackwell GPU architecture, Jetson Thor can achieve up to 2,070 TFLOPs, outperforming its predecessor, AGX Orin, by a factor of more than 7.5 on an identically sized 100mm x 87mm die. Thor accomplishes this by more than doubling Orin's power draw, swelling from a 15-60W range to 40W-130W, improving energy efficiency 3.5-fold.
The module's most crucial new feature is the ability to run multiple AI models simultaneously, improving real-time processing and navigation of its physical environment. Nvidia began developing Jetson in 2014 for robotics, driverless cars, and other AI applications, and clients are already planning to integrate Thor into upcoming robots.
For example, Agility Robotics, which is helping automate Amazon warehouses, will employ Thor for its sixth-generation Digit humanoid robot. Boston Dynamics is also set to debut a new variant of its well-known Atlas robot running on Thor.
If Nvidia's promised robot boom indeed arrives, it could present risks alongside opportunities. Although the labor market is a primary concern, studies indicate that even workers who keep their jobs amid increasing automation can experience a loss of purpose and autonomy as tasks left to humans become increasingly monotonous.
Furthermore, recent events have proven that Nvidia doesn't have total control over who accesses its AI computers. Fully autonomous Russian prototype attack drones were recently discovered undergoing field testing in Ukraine using Orin computers. According to the Ukrainian military, they can detect, analyze, and attack targets without human intervention. Further advances like those promised by Thor could make such technology even more dangerous.
Jetson Thor T5000 modules start at $2,999 for 1,000 units, and AGX Thor development kits start at $3,499.
Nvidia launches Jetson Thor: $3,499 "robot brain" AI computer with real-time decision-making
