Edge AI on Arm Cortex-M and Ethos-U

Running machine learning models on low-power, resource-constrained devices is difficult because traditional cloud AI workflows don’t scale efficiently to the edge. Developers must balance power, performance, and memory limits while ensuring real-time inference. Explore how Arm solves these challenges in the launchpad.

Embedded developers can upskill in areas like model optimization, quantization, and deployment using frameworks such as LiteRT and ExecuTorch. Arm provides tutorials and tools to make this transition easier. Explore Resources.

Arm’s Cortex-M processors with Helium enable high-performance vector processing for AI, DSP, and vision tasks, while Ethos-U NPUs deliver efficient on-device AI acceleration for real-time inference. Explore Arm’s Edge AI Resources.

Developers can use the Vela Compiler to optimize LiteRT models for Ethos-U NPUs. Arm’s ecosystem also includes Keil MDK, VS Code extensions, ExecuTorch, LiteRT, and the Arm Model Zoo for quick experimentation and deployment. Explore Arm’s Edge AI Tools and the launchpad above.

Start with Arm’s Edge AI quick start guide to set up your environment. Take your next step by exploring all Edge AI resources – guides, learning paths, tools and libraries, and example applications – to support your build.