Seeed Studio Announces Supports for NVIDIA’s Next-Generation Jetson T2000 & T3000 Modules for Scalable Edge AI and Robotics
SHENZHEN, China, July 13, 2026 — Seeed Studio, today shares with all developers that we’ll support the upcoming NVIDIA Jetson T2000 and T3000 modules, the latest additions to the Jetson AGX Thor family, purpose-built to power the next wave of scalable edge AI and autonomous robotics.
The NVIDIA Jetson AGX series stands as the world’s most adopted robotics computer platform, already deployed across factories, warehouses, industrial equipment, and the rapidly emerging field of humanoid robotics. Humanoid robotics represents one of the largest technological opportunities of our time, with the potential to become a multi-trillion-dollar industry.
As this sector moves toward high-volume production and mass-market adoption, developers and manufacturers urgently need compact, power-efficient AI supercomputers optimized for scalable deployment. The new T2000 and T3000 modules bridge the gap between the proven Orin generation and the flagship T5000, delivering a generational leap in AI compute performance and energy efficiency through the NVIDIA Blackwell GPU architecture and Arm Neoverse CPU cores.
The Jetson T3000 module directly addresses the demands of next-generation autonomous systems.
- built around an NVIDIA Blackwell GPU delivering 865 FP4 TFLOPS and an 8-core Arm Neoverse CPU at 2.6GHz.
- includes 32 GB of LPDDR5X memory with 273 GB/s bandwidth, 25GbE high-speed connectivity, and full-stack safety features—all packaged at roughly half the size and power of the flagship T5000.
- achieves approximately similar T5000’s inference performance for LLMs, VLMs, VLAs, and WFMs, while offering the same high memory bandwidth and optimized tokens per watt.
While the Jetson T2000 module extends the Thor architecture to a broader range of edge AI applications.
- delivering 400 FP4 TFLOPS of AI compute from its NVIDIA Blackwell GPU, paired with a 6-core Neoverse CPU and 16 GB of LPDDR5X memory at 137 GB/s bandwidth.
- serves as the ideal entry point to the NVIDIA Jetson Thor product family within a 40W power envelope, enabling visual AI agents, autonomous mobile robots, manipulators, and other intelligent edge systems.
Developers can begin designing with the Jetson AGX Thor Developer Kit, which will emulate the T3000 later this month with NVIDIA JetPack 7.2.1, the T2000 emulation will be added soon. The kit supports the complete NVIDIA software stack, along with open models including NVIDIA Nemotron, NVIDIA Cosmos 3, and NVIDIA Isaac GR00T, plus new Jetson agent skills.
Discover the tutorial on how we deploy Nemotron-Nano-30b model via NVIDIA NemoClaw on NVIDIA Jetson Thor developer kit.
A significant milestone for edge AI developers is the upcoming NVIDIA Cosmos 3 Edge, the first fully open omnimodal world model for physical AI coming to NVIDIA edge computers, including Jetson Thor devices. This lightweight 4B parameter model enables real-time livestream and multi-view vision understanding for smart infrastructure, robots, and autonomous systems. It can also be customized into task, embodiment, or sensor-specific world-action models in a few hours.
The Jetson T2000 and T3000 modules are scheduled for availability in Q1 2027. As an NVIDIA Elite Partner, Seeed Studio remains deeply committed to building a rich developer ecosystem around Jetson edge computing products. We will provide emulation and migration services to help developers transition seamlessly to the new modules.
Seeed also previewed its upcoming J601 carrier board based on the NVIDIA Jetson T5000 module, offering developers and enterprises a powerful platform for prototyping and deploying the most demanding edge AI applications, launching soon in late July.
- Up to 2070 TOPS for Physical AI inference and processing;
- Compact as 168x153mm for a space-constrained robotics system;
- 19-48V DC input supported for battery-powered autonomous machines;
- rich interfaces including 10GbE, USB 3.2, CAN-FD, RS-485, I2S for rapid system integration;
- 8x GMSL2 ports for high-performance perception, SLAM, BEV, and autonomous navigation
The broader developer community is encouraged to begin exploring Thor architecture capabilities today using the available developer resources and open models. Subscribe to the Seeed newsletter for the latest updates.
About Author
