Chip design company Arm has announced an edge AI platform to drive performance for IoT applications.

The Armv9 edge AI platform, optimised for IoT with the Cortex-A320 CPU and Ethos-U85 NPU (www.arm.com/products/silicon-ip-cpu/ethos/ethos-u85), enables on-device AI models over one billion parameters. This has earned support from the likes of AWS, Advantech, Siemens and Renesas.

The Cortex-A320 leverages Armv9 to enhance efficiency, performance and security in IoT markets, driving advancements in industrial automation, smart cameras and beyond.

Arm Kleidi for IoT boosts performance by up to 70% and empowers more than 20 million developers to integrate seamlessly with AI frameworks, streamlining edge AI development.

“The AI revolution is no longer confined to the cloud,” said Paul Williamson, Arm senior vice president. “As our world becomes increasingly connected and intelligent, from smart cities to industrial automation, the need to process AI workloads at the edge is not just advantageous, but essential. This marks a significant milestone in this evolution with the introduction of the Armv9 edge AI platform.”

He said there was a need to meet the growing computing demands across IoT applications, such as autonomous vehicles navigating factory floors, smart cameras that could adapt their functionality through software updates, and human-machine interfaces that offered more natural, AI-driven interactions.

“To innovate and scale at pace, they need the flexibility to execute their AI workloads where it makes sense, more robust security and increased software flexibility, and Armv9 technology delivers this at scale,” he said. “The platform delivers an eight-times improvement in machine-learning performance compared with the Cortex-M85-based platform we launched last year.”

The Cortex-A320 (www.arm.com/products/silicon-ip-cpu/cortex-a/cortex-a320) takes advantage of Armv9 architectural features, such as SVE2 for ML performance, and delivers a ten-times ML performance uplift and 30 per cent scalar performance uplift compared with its Cortex-A35predecessor.

The platform’s Armv9.2 architecture also brings security features such as pointer authentication, branch target identification and memory tagging extension to even the smallest Cortex-A devices. This is crucial as edge devices often operate in exposed environments and handle sensitive data.

Arm is extending Arm Kleidi to IoT, a set of compute libraries for developers of AI frameworks designed to optimise AI and ML workloads on Arm-based CPUs with no additional developer work needed. KleidiAI is already integrated into popular IoT AI frameworks, such as Llama.cpp and ExecuTorch or LiteRT via XNNPack, accelerating the performance of key models, including Meta Llama 3 and Phi-3. As an example, KleidiAI brings up to 70% more performance to the Cortex-A320 when running Microsoft’s Tiny Stories dataset on Llama.cpp.

The platform also maintains software compatibility with higher-performance Cortex-A processors. This scalability ensures developers can build products that grow and adapt as requirements change. With access to the Armv9 ecosystem and compatibility with both rich operating systems such as Linux, and real-time operating systems such as Zephyr, developers have flexibility, can leverage existing tools and knowledge, and take advantage of software reuse, reducing time to market and lowering total cost of ownership.

“With over 20 million active Arm developers worldwide, the potential for innovation is immense,” said Williamson.

The platform’s ability to run tuned large language models (LLMs) and small language models (SLMs) for agent-based AI applications opens up edge use cases.

“We’re moving toward a future where intelligent decision-making happens closer to the point of data collection, reducing latency and improving privacy,” said Williamson. “This isn’t just another incremental step forwards, it represents a fundamental shift in how we approach edge computing and AI processing. For the first time, we’re seeing an Armv9 CPU specifically optimised for IoT applications, bringing together ultra-efficiency and advanced AI capabilities in a way that hasn’t been possible until now.”

Miller Chang, president of the embedded sector at Advantech (www.advantech.com), added: “The evolution of edge AI is accelerating, and advancements in Arm’s IoT computing architecture will bring new possibilities for intelligence at the edge. As a leading player in edge computing and edge AI, Advantech sees this innovation as a significant step forward for the broader Arm ecosystem, enabling smarter, more efficient and secure AI-driven applications across industries. This innovation will drive industry growth and technological breakthroughs in the edge computing market.”