#Germany #Edge AI #Nuremberg #Nordic Semiconductor #nRF54LM20B

Nordic Semiconductor's Bold Move in Edge AI

Nordic Semiconductor, a global leader in energy-efficient wireless connectivity solutions, has announced significant advancements in its edge artificial intelligence (AI) strategy. The company is expanding its next-generation ultra-low-power solutions portfolio by introducing innovative features, including the broad availability of its Neural Processing Unit (NPU)-capable nRF54LM20B System-on-Chip (SoC). With this step, Nordic aims to make energy-efficient edge intelligence accessible to developers and battery-operated devices alike.

High-Performance Acceleration for Real-Time Intelligence

The integrated NPU within the nRF54LM20B comes equipped with substantial memory, enhancing TensorFlow Lite models by up to 15 times while consuming considerably less power compared to the Arm Cortex CPU execution. The NPU also delivers up to seven times greater performance and eight times better energy efficiency than its nearest competitor in the edge AI space. This enables the handling of high data rates from sensors, audio, and event-driven edge AI workloads, even with a minuscule battery.

Øyvind Strøm, Executive Vice President of Short-Range at Nordic Semiconductor, emphasizes this new generation of edge-AI capabilities as a transformative force for small, battery-operated devices. By integrating powerful intelligence directly into the device without compromising on energy efficiency, latency, or system complexity, Nordic is paving the way for a new class of products that are smarter, faster, and significantly more efficient than ever before.

Streamlined Creation of Edge AI Models

Nordic has not only enhanced its ultra-low-power edge AI solution with advanced hardware acceleration but has also introduced features that simplify common hurdles faced by developers when integrating intelligence into various battery-operated IoT products.

Simplified Model Creation and Optimization

Developers can now build edge AI models for on-device execution by uploading custom datasets and defining the model architecture for generating classification or regression models. This fundamental shift simplifies the design of edge AI architectures, making them ideal for wearables, industrial sensor applications, and smart home devices.

Moreover, the ability to generate custom activation word models from a single text input (e.g., "Hello Nordic") allows developers to bypass the need for collecting or labeling thousands of audio samples. This alleviates the costs and complexities associated with assembling speech datasets, enabling voice-activated control for devices such as smart speakers, wearables, remote controls, and battery-operated audio accessories.

Comprehensive Ecosystem Support

With the rollout of its new online tool, the Nordic Edge AI Lab, developers can create models that support accelerated inference on the NPU or efficient execution on the CPU. This flexibility addresses integration challenges between models, tools, and embedded firmware, leading to functionalities such as audio classification, anomaly detection, predictive maintenance, and data processing, without the need for cloud execution.

Overall, these enhancements dramatically reduce the typical effort needed to embed AI capabilities within devices and substantially cut down the time required from concept to a functional prototype.

Availability and Future Prospects

The nRF54LM20 DK is now available through Nordic's distribution partners. For samples of the nRF54LM20B SoC, interested parties should contact Nordic's sales team. Mass production is anticipated to begin in the second quarter of 2026.

Experience Nordic Semiconductor at Embedded World 2026

Nordic Semiconductor will showcase its latest innovations during the Embedded World 2026 event in Nuremberg, Germany, from March 10-12. Attendees can visit their booth at Hall 4A, Stand 310 for product demonstrations and discussions.