#USA #Edge AI #Nuremberg #Nordic Semiconductor #nRF54L

Nordic Semiconductor Boosts Leadership in Edge AI

Nordic Semiconductor, a prominent name in low-power wireless connectivity solutions, has recently unveiled a significant advancement in its efforts to mainstream low-power edge artificial intelligence (AI). This development introduces the new nRF54LM20B System on Chip (SoC), equipped with a Neural Processing Unit (NPU) that promises to revolutionize how we perceive machine learning capabilities in battery-powered devices.

Enhancing Real-Time Intelligent Processing

The NPU integrated into the nRF54LM20B can accelerate TensorFlow Lite models, delivering performance that is up to 15 times faster while consuming significantly less energy than traditional processing on an Arm Cortex CPU. This leap in technology allows for high-speed sensor data handling, audio processing, and event-driven AI workloads to be effectively managed within the constraints of small battery sizes.

According to Øyvind Strøm, Executive Vice President of Short-Range at Nordic Semiconductor, this generation of edge AI capabilities radically transforms what battery-powered devices can perceive and interact with in real-time environments. He emphasizes that the integration of robust processing capabilities directly into devices not only improves responsiveness but also enhances energy efficiency.

Simplifying AI Model Development

Nordic's new technology aims to simplify the development and deployment of edge AI models. By allowing developers to create and optimize AI models easily, this innovation breaks down common barriers associated with integrating intelligence into various connected and battery-operated IoT products. Among the key features introduced is the streamlined process for generating models tailored for specific use cases, empowering developers to upload custom datasets and define their model architectures easily.

For instance, developers can now create activation words from a single text input, such as “Hello Nordic,” eliminating the need for extensive audio sample collections. This could significantly reduce the time and resources typically required for voice command applications, enabling more efficient development of smart speakers, wearable technology, and remote control devices.

Expanded Ecosystem Compatibility

Moreover, the new Nordic Edge AI Lab's model generation tools offer flexibility, allowing models to be optimized for either the NPU or the CPU, depending on the specific performance and power consumption requirements. This flexibility alleviates integration frictions between models, tools, and embedded firmware while facilitating various functions, such as audio classification, anomaly detection, predictive maintenance, and environmental sensing.

When combined, these advancements notably decrease the effort needed to incorporate AI functionalities into embedded devices, compressing the typical development timeline from concept to functional prototype.

Availability and Future Prospects

The nRF54LM20 DK is already available through Nordic's distribution channels, with samples of the nRF54LM20B SoC accessible for interested developers through Nordic's sales department. Mass production is slated to commence in the second quarter of 2026.

Visitors to Embedded World 2026, taking place in Nuremberg from March 10 to 12, will have the opportunity to see demonstrations of these groundbreaking technologies at Booth 310 in Hall 4A.

Nordic Semiconductor continues to lead the charge in low-power, high-efficiency processing for edge AI, paving the way for the next generation of intelligent, responsive devices that will seamlessly integrate into our daily lives.