Edge AI for IoT is rapidly becoming a defining requirement for the next generation of connected devices, and Nordic Semiconductor is positioning itself at the center of this shift. The company has unveiled an industry-leading, ultra-low-power edge AI solution designed to bring real intelligence to even the smallest battery-powered IoT devices. The move promises faster decisions, stronger privacy, and longer battery life at massive scale.
At the heart of Nordic’s announcement is a tightly integrated platform that combines new silicon, AI models, and developer tools. This includes the nRF54LM20B ultra-low-power wireless SoC with an integrated Axon neural processing unit, custom ultra-tiny Neuton AI models, and the Nordic Edge AI Lab. Together, these components aim to simplify and accelerate edge AI development for embedded engineers.
According to Vegard Wollan, CEO of Nordic Semiconductor, the shift toward edge intelligence is unavoidable. While cloud-based “AI factories” train models, Nordic focuses on deploying intelligence where real-world events occur.
By processing data directly on the device, edge AI eliminates round-trip latency, improves safety, strengthens privacy, and significantly reduces power consumption for connected products.
This approach is especially critical for battery-powered IoT devices that operate in constrained environments. Instead of sending raw data to the cloud, devices can now make millisecond-level decisions locally. This not only improves responsiveness but also helps manufacturers meet growing regulatory and compliance requirements around data handling and sovereignty.
A key highlight of the platform is the next-generation nRF54LM20B SoC. This chip is the first large-memory member of Nordic’s nRF54L Series and integrates the Axon NPU, an ultra-efficient AI hardware accelerator. Nordic acquired Axon technology through its acquisition of Atlazo in 2023, and this SoC marks its first large-scale commercial deployment.
The Axon NPU delivers up to seven times faster performance and up to eight times higher energy efficiency compared to competing solutions. It is optimized for demanding edge AI tasks such as sound classification, keyword spotting, gesture recognition, and image-based detection. These capabilities allow developers to run sophisticated AI workloads without sacrificing battery life.
Beyond the NPU, the nRF54LM20B offers a robust hardware foundation. It features 2 MB of non-volatile memory, 512 KB of RAM, a 128 MHz Arm Cortex-M33 processor paired with a RISC-V coprocessor, high-speed USB, and up to 66 GPIOs. The SoC also supports Nordic’s fourth-generation ultra-low-power 2.4 GHz radio, enabling Bluetooth LE, Bluetooth Channel Sounding, Matter over Thread, and other wireless protocols.
Complementing the hardware are Neuton edge AI models, which are designed for extreme efficiency. These models typically measure under 5 KB in size and can be up to ten times smaller, faster, and more energy-efficient than traditional CPU-based AI models. Despite their size, they support use cases such as anomaly detection, biometric monitoring, activity tracking, and gesture recognition.
Developers can create and customize these models using Nordic Edge AI Lab, a development environment built to reduce complexity. The tool allows engineers to generate AI models tailored to their specific datasets and devices, without requiring deep expertise in machine learning. As a result, AI becomes a natural part of product design rather than an added burden.
One real-world deployment highlights the impact of this approach. A global supply chain company upgraded its smart tracking devices using AI models created in Nordic Edge AI Lab.
The devices can now detect real handling events such as shocks, shaking, and transport conditions directly on the device. These insights were rolled out across an entire fleet without operational disruption, supported by Nordic’s nRF Cloud lifecycle services.
While intelligence is moving to the edge, cloud services remain essential. Nordic emphasizes that over-the-air updates, device observability, and lifecycle management are critical as AI-enabled fleets scale. Manufacturers need continuous visibility into device performance, both to improve products and to meet customer and regulatory expectations.
By combining edge intelligence with cloud-based management, Nordic aims to ensure connected products can evolve securely and efficiently over time. Data collected from deployed devices can be used to refine features and optimize performance without disrupting the end-user experience.
With its integrated hardware, ultra-tiny AI models, and developer-friendly tools, Nordic Semiconductor is making edge AI practical for a broad range of applications. From wearables and smart sensors to industrial and supply chain systems, the company is setting a new standard for ultra-low-power edge intelligence.
As Edge AI for IoT becomes a baseline expectation rather than a premium feature, Nordic’s approach could shape how billions of connected devices are designed and deployed in the years ahead.
For more updates on edge AI, IoT innovation, and the future of intelligent devices, visit ainewstoday.org and stay ahead of what’s next in AI.
