Arm unveils Armv9 edge AI platform to lower barriers for on-device intelligence
Compute platform firm Arm has launched the world’s first Armv9 edge AI platform, designed for IoT and edge AI SoC for machine learning. It’s low-cost with easy access to advanced AI performance and security for edge devices through Arm Flexible Access.
The platform features Arm Cortex-A320 CPU and Arm Ethos-U85 NPU for on-device AI models of more than one billion parameters. Advanced security measures such as Pointer Authentication Code (PAC), Branch Target Identification (BTI) and Memory Tagging Extension (MTE) are also incorporated.
The platform is designed to be used for next-gen edge AI applications such as smart cameras, robotics and smart home products.
Arm Flexible Access accelerates innovation and time-to-market for startups and OEMs by reducing barriers to entry. Cortex-A320 launches in November 2025, with Ethos-U85 arriving Q1 2026.
Arm Flexible Access has seen over 400 tape-outs, with up to 300 members at any one time, representing both startups and major OEMs. Over half of the workloads on Arm Flexible Access are aimed at embedded and IoT usage.
The Armv9 architecture also pairs machine learning prowess with performance through the SVE2 and efficiency from Ethos-U85. It is a platform to make human-machine interactions smarter and more natural in vision, voice, and gesture recognition.
Arm is working to make advanced technology and computing more accessible to innovators throughout its ecosystem. AI, by 2028, will take center stage when kicking off a new IoT project with Arm technology.
Arm Flexible Access provides low-cost or no-cost access (for eligible startups) to a large number of Arm technologies and tools. Companies are charged a licensing fee for the technology used to produce final designs, encouraging experimentation and innovation.
Partners such as Raspberry Pi, Weeteq, Hailo and SiMa.ai have pioneered leading edge AI technologies built on Arm Flexible Access.
Arm is working to develop its ecosystem in order to address increasing requirements for performance, efficiency and security in edge AI.
AI Infra Summit 2026
Article Topics
ARM | Arm flexible access | armv9 | edge AI | hardware | IoT
Comments