NVIDIA introduced the Jetson Thor T3000 and T2000 AI modules on July 15 to support robotics and Edge AI applications. The new NVIDIA Blackwell-powered modules help developers build smarter robots and AI machines with less power and smaller hardware. The company plans to launch both modules in the first quarter of 2027.
The launch comes as more industries adopt robots for real-world work. NVIDIA created the new Jetson Thor modules to run powerful AI models directly on robots and edge devices. This approach reduces cloud dependence while improving speed, safety, and overall performance.
The Jetson Thor T3000 is the more powerful model in the lineup. It delivers up to 865 FP4 teraflops of AI performance and includes an eight-core Arm Neoverse CPU with 32GB LPDDR5X memory.
NVIDIA said that the module is about half the size and uses nearly half the power of the larger T5000 while offering similar AI performance for language, vision, and robotics tasks.
The Jetson Thor T2000 targets developers looking for a more affordable option. It delivers up to 400 FP4 teraflops with 16GB memory and supports visual AI, autonomous mobile robots, and industrial machines. NVIDIA also introduced new software tools that reduce memory use and make advanced AI models easier to run on smaller systems.
NVIDIA stated that companies like Amazon Robotics, Boston Dynamics, Agile Robots, Hitachi, 1X, and Techman Robot already use the Jetson Thor platform.
Developers can test the T3000 through JetPack 7.2.1 emulation today, while T2000 support will arrive later. Both modules are expected to become available in the first quarter of 2027.