NVIDIA Jetson Thor Kicks Off a New Age for Humanoid Robots
NVIDIA has launched the next-generation Jetson Thor system for the Blackwell architecture, with the intention of enormous leverage in the robotics field, especially in humanoid robot development. This launch falls in line with the company's focus on "Physical AI", aiming to imbue robots with reasoning capabilities suitable for performing complex tasks.
Big Ahead in Robotics Compute
At the heart of the system is the new Jetson T5000 module based on Blackwell. According to NVIDIA, this platform provides considerable generational performance advantage over the last Jetson Orin, namely, 7.5 times more AI compute, 3.1 times more CPU performance, and double the memory. This power lets Jetson Thor run generative AI frameworks and reasoning models directly on the device, endowing the robot with a more competent "brain".
The system is designed to provide up to 2,070 teraflops of AI computation using FP4 precision, within a power envelope of 130W to enable a much-dened application to join a demanding robotics project.
NVIDIA Jetson Thor Specifications
The following are key technical specifications relating to the Jetson Thor platform:
| Category | Specification |
|---|---|
| Module | Jetson T5000 |
| GPU | 2560-core NVIDIA Blackwell architecture GPU with 96 fifth-gen Tensor Cores |
| CPU | 14-core Arm Neoverse V3AE 64-bit CPU |
| Memory | 128 GB 256-bit LPDDR5X at 273 GB/s |
| Networking | 1x 5GbE RJ45 connector, 1x QSFP28 (4x 25 GbE) |
| I/O | QSFP connector, HDMI port, DisplayPort, 2x USB-A 3.2, 2x USB-C 3.1, Gigabit Ethernet, 2x 13-pin CAN header, Microfit power jack |
| Storage | 1TB NVMe M.2 Key M slot |
| Power | 40W – 130W |
Early Adopters and Price
Agility Robotics and Boston Dynamics are other top robotics companies that have started implementing Jetson Thor systems in their products. The early adoption augurs well for these firms to produce more capable robots in the near future.
But this computing power comes at quite a cost. The NVIDIA Jetson AGX Thor developer kit comes with a bill of $3,499-a hefty amount for teams building humanoid robots.
The case for NVIDIA's commitment to this space was strengthened by Jensen Huang himself signing off many of the initial units sent to customers. With partner Foxconn widely expected to launch humanoid robots before the end of the year, AI with Physicality seems to be a big strategic thrust for the company.
