NVIDIA Jetson Thor T5000/T4000 Blackwell · 128GB LPDDR5 · 4× Native 10GbE · JetPack 7.2 · Industrial-Grade · Launched July 2026
NVIDIA Jetson Thor
T5000 / T4000
Blackwell
Up to 2,070 TFLOPS
FP4 Inference
14-core Arm
Neoverse V3AE
Up to 128 GB
LPDDR5
4× 10GbE RJ45
Native Direct-Connect
HDMI 2.1
DisplayPort 1.4a
2× USB-C 3.2 Gen 2
2× USB-A 3.2 Gen 1
M.2 B Key (5G/LTE)
2× M.2 E Key (NVMe/WiFi)
19-36V DC Wide
-10°C to 50°C
176 × 112 × 66mm
~1.6 kg
JetPack 7.2
Yocto / Ubuntu
WEDA · SUSI · Edge AI SDK
Inference Kit · Metropolis
July 7, 2026
The Advantech AIR-075 is a next-generation Edge AI Agent System powered by the NVIDIA Jetson Thor platform, delivering a massive 2,070 TFLOPS of FP4 AI inference performance. Launched in July 2026, it represents one of the industry's first edge AI systems purpose-built for Vision and Agentic AI workloads.
Built on the NVIDIA Blackwell architecture, the AIR-075 features four native 10GbE RJ45 ports, providing direct-connect high-bandwidth for multi-camera deployments without external network aggregation switches. This simplifies system topology while enabling real-time, multi-stream video ingestion at full resolution. With up to 128GB LPDDR5 memory and a 14-core Arm Neoverse V3AE CPU, the AIR-075 handles multi-modal AI inference, sensor fusion, and visual AI agent pipelines at the edge.
Fully supported by NVIDIA JetPack 7.2 with both Yocto and Ubuntu options, the AIR-075 integrates seamlessly with the Advantech Edge AI SDK, NVIDIA Metropolis, Triton Inference Server, and Cosmos Reason. Target applications include smart warehouse safety, intelligent traffic management, multi-camera factory vision, and industrial agentic AI deployments. The industrial-grade chassis ensures reliable operation in production environments at scale.
Need Advantech AIR-075?
Contact QS Compute for pricing, availability, and deployment consultation.
Request Quote