Lanner EAI-I351 Delivers Server-Class Edge AI with NVIDIA Jetson Thor for Autonomous Systems
January 20, 2026
News
Lanner Electronics launched a robotic AI platform designed for next-generation physical AI. The EAI-I351 is designed around the NVIDIA Jetson Thor system-on-module (SoM) to deliver server-class performance for autonomous mobile robots (AMRs), heavy-duty industrial vehicles, and edge-native generative AI applications.
The NVIDIA Jetson Thor leverages the NVIDIA Blackwell GPU architecture that supports up to 2,070 FP4 TFLOPS of AI compute utilizing an efficient 130W power.
When juxtaposed with its predecessor, NVIDIA Jetson AGX Orin, the NVIDIA Jetson Thor series delivers seven and a half times greater AI compute performance, three and a half times improved energy efficiency, and has a dedicated transformer engine enhanced for running multi-modal generative AI models and Large Language Models (LLMs) locally.
Incorporated is a group of accelerators, including a third-generation Programmable Vision Accelerator (PVA), optical flow accelerators, and dual encoders/decoders, ensuring low-latency processing for complex visual tasks.
The EAI-I351is engineered for the bandwidth needs of autonomous environments with high-speed networking and numerous sensor interfaces. It is available in two core configurations based on the compute module:
NVIDIA Jetson T5000:
- 128GB Unified Memory
- 2070 FP4 TFLOPS
NVIDIA Jetson T4000:
- 64GB Unified Memory
- 1200 FP4 TFLOPS
Flexible network I/O configurations include one QSFP28 (four 25GbE or three 25GbE, depending on SKU) combined with one 5GbE RJ45. For visual perception and sensor fusion, eight GMSL2 deserializers enable direct connectivity for high-resolution automotive cameras. Additional I/O includes four USB 3.2 Gen1 and four Digital I/O (DIO) for legacy device integration.
For rugged environments, the EAI-I351 supports an operating temperature range of -25°C to 70°C. Seamless cloud-to-edge communication is delivered via two M.2 expansion slots accommodating Wi-Fi and 5G/LTE connectivity modules.
Fully optimized for NVIDIA AI software stack, the platform includies NVIDIA Isaac for robotics simulation and deployment, NVIDIA Metropolis for vision AI, and NVIDIA Holoscan for real-time sensor processing. It supports agentic AI workflows, such as Video Search and Summarization (VSS) using NVIDIA Cosmos Reason.
For more information, visit lannerinc.com/products/edge-ai-appliance/edge-inference-appliances/eai-i351.
Chad Cox. Production Editor, Embedded Computing Design, has responsibilities that include handling the news cycle, newsletters, social media, and advertising. Chad graduated from the University of Cincinnati with a B.A. in Cultural and Analytical Literature.
