Arm Unveils Armv9 Edge AI Computing Platform to Transform the Future of IoT
February 27, 2025 – Arm, the global leader in computing IP, has launched the world’s first Armv9 Edge AI Computing Platform, bringing groundbreaking advancements to IoT (Internet of Things) applications. Featuring the Cortex-A320 CPU and Ethos-U85 NPU, this new platform enhances ML (Machine Learning) computing power by 8x compared to previous Cortex-M85-based solutions and enables edge devices to run large language models (LLMs) with over 1 billion parameters. This marks a significant milestone in edge AI, introducing a new era of intelligence, security, and energy efficiency.
February 27, 2025 – Arm, the global leader in computing IP, has launched the world’s first Armv9 Edge AI Computing Platform, bringing groundbreaking advancements to IoT (Internet of Things) applications. Featuring the Cortex-A320 CPU and Ethos-U85 NPU, this new platform enhances ML (Machine Learning) computing power by 8x compared to previous Cortex-M85-based solutions and enables edge devices to run large language models (LLMs) with over 1 billion parameters. This marks a significant milestone in edge AI, introducing a new era of intelligence, security, and energy efficiency.
A Breakthrough in Edge AI Performance and Efficiency
Key Components of the Armv9 Edge AI Platform
Cortex-A320 CPU – The most compact and energy-efficient Armv9-A processor designed for IoT devices.
30% higher scalar performance vs. Cortex-A35
10x ML inference improvement
Supports BF16 and INT8 AI data types for accelerated neural network processing
Scalable Vector Extension (SVE2) enhances AI and ML computations
Ethos-U85 NPU – 4x performance boost over its predecessor with a 20% increase in power efficiency.
MAC units scale from 128 to 2048 (delivering up to 4 TOPs at 1GHz)
Supports both Transformer architectures and CNNs for AI inference
Optimized for industrial automation, smart cameras, and autonomous applications
By combining Cortex-A320 with Ethos-U85, the platform delivers real-time AI processing with enhanced security, enabling intelligent edge solutions for industrial automation, smart surveillance, and autonomous systems.
Transforming IoT with High-Performance Edge AI
Paul Williamson, SVP and GM of Arm’s IoT division, emphasized:
“AI innovation is no longer confined to the cloud. As the world becomes increasingly connected and intelligent, processing AI workloads at the edge offers unparalleled advantages in real-time decision-making, privacy, and efficiency. The Armv9 Edge AI Computing Platform is a milestone in this evolution.”
Real-time AI Processing for Smarter IoT
Traditional edge AI chips lack processing power, often requiring video feeds to be sent to the cloud for analysis, which introduces latency.
The Cortex-A320 + Ethos-U85 combination enables on-device AI inference, reducing bandwidth dependency and enhancing real-time responsiveness.
In multimodal AI workloads, pairing Ethos-U85 with Cortex-X925 further enhances performance for video processing and AI-powered automation.
Reliability & Security for Industrial Applications
Proven Arm architecture with 3,000+ design wins across industrial and embedded applications.
Built-in security features including Memory Tagging Extension (MTE), Pointer Authentication (PAC), and Secure EL2 virtualization to enhance system integrity.
Armv9 Architecture: Redefining Edge Computing
The Armv9 Edge AI platform introduces three core innovations to the IoT landscape:
Ultimate Security
Memory Tagging Extension (MTE) – Reduces memory vulnerabilities.
Pointer Authentication (PAC) & Branch Target Identification (BTI) – Prevents control flow attacks.
Secure EL2 Virtualization – Enables multi-tenant edge applications with hardware-level security.
Performance Leap
Enhanced Neon & Scalable Vector Extensions (SVE2) for ML acceleration.
New Matrix Multiplication Instructions optimize AI and ML workloads.
Energy Efficiency Revolution
50% power efficiency improvement over Cortex-A520.
Optimized microarchitecture with dense L1 cache storage and streamlined data pathways.
Delivers up to 256 GOPS (Giga Operations Per Second) at 2GHz – enabling ML execution directly on the CPU without external accelerators.
KleidiAI: 70% Faster AI Model Deployment
To address AI development challenges, Arm is integrating KleidiAI into its IoT ecosystem, accelerating AI model execution by up to 70%.
AI Ecosystem Integration
KleidiAI is now embedded into Llama.cpp, ExecuTorch, and Meta Llama 3
Pre-optimized for Microsoft’s Tiny Stories dataset, delivering unprecedented AI performance gains
Seamless compatibility with FreeRTOS, Zephyr, Linux, and Android, allowing 20M+ developers to migrate existing tools effortlessly
Industry Endorsements & Adoption
The Armv9 Edge AI Computing Platform has garnered widespread support from global tech leaders, reinforcing its role in shaping the future of edge computing.
Global Industry Adoption
Amazon Web Services (AWS):"Arm's Edge AI platform enables AWS IoT Greengrass on lightweight devices via Nucleus Lite, reducing memory footprint while maintaining high performance."
Siemens:"Armv9 architecture is a game-changer for industrial automation, expanding our AI product portfolio across manufacturing, smart infrastructure, and mobility applications."
Renesas Electronics:"The Cortex-A320 excels in AI/ML processing, security, and power efficiency, accelerating Renesas’ innovation in scalable, high-efficiency IoT solutions."
The Future of Edge AI & IoT
With generative AI expanding into edge computing, the IoT industry is witnessing an unprecedented transformation.
Arm’s latest platform is not just an incremental upgrade—it is a paradigm shift in edge AI, offering unparalleled performance, security, and efficiency.
As AI becomes embedded in billions of IoT devices, Arm’s Edge AI Computing Platform is set to redefine how smart devices process AI workloads in real time.
Conclusion:
Arm’s cutting-edge AI innovations will pave the way for the next generation of intelligent, secure, and energy-efficient IoT applications across industrial, automotive, and consumer sectors.
