The new AIR-075 is a next-generation edge AI agent system powered by NVIDIA Jetson Thor. Designed by Advantech for next-generation vision and agentic AI, the AIR-075 delivers a massive 2,070 TFLOPS of AI compute and features four native 10GbE ports optimised for high-bandwidth, multi-camera deployments.
Fully supporting NVIDIA JetPack 7.2 for Yocto/Ubuntu projects, the system integrates WISE-Edge Developer Architecture (WEDA) and NVIDIA NemoClaw to enable advanced multimodal reasoning and autonomous decision-making for robotics, factory automation and agentic AI applications. Its on-module validated DRAM secures critical memory supply, eliminating market sourcing risks to meet the surging edge AI demand.
Accelerating multi-camera vision AI
Powered by NVIDIA Jetson Thor with the NVIDIA Blackwell architecture, the AIR-075 delivers high-performance edge AI inference in a compact, thermally optimised, industrial-grade chassis. Its four native 10GbE RJ45 interfaces provide the massive bandwidth required for multiple IP cameras, enabling real-time video ingestion and AI processing without the need for external network aggregation switches.
This direct-connect capability slashes system complexity and perfectly aligns with existing IP-based factory vision infrastructure. Complemented by a wide 19–36V DC power input and flexible M.2 expansion for NVMe storage, 5G and Wi-Fi, the AIR-075 provides a highly scalable foundation for autonomous machines and distributed edge AI deployments.
Faster agentic AI deployment
The AIR-075 is backed by Advantech’s edge AI software ecosystem and NVIDIA’s open models including NVIDIA Cosmos, NVIDIA Nemotron, NVIDIA GR00T and NVIDIA JetPack 7.2 to simplify development, deployment and fleet operation. WEDA serves as the AI deployment infrastructure, enabling centralised lifecycle management of AI models, agent workflows, software updates and operational policies across distributed edge environments.
Combined with NVIDIA NemoClaw-ready runtime support, the AIR-075 enables multimodal AI agents to interact with cameras, sensors, enterprise systems, and physical devices through natural language and automated workflows. Edge AI SDK provides pre-validated AI frameworks, development tools, and reference models, while NVIDIA JetPack 7.2 brings agent-friendly Jetson software and official Yocto support for custom Linux distributions that optimise system performance, while bit-for-bit reproducibility ensures deterministic builds.
The AIR-075 significantly reduces the effort required to move from proof-of-concept to large-scale production deployment for physical AI across industrial environments.
