Nordic Edge AI 

Ultra-low-power edge AI across the industry's broadest range of applications 

Your data. Your model. Your edge.

From saving bandwidth and energy to more responsive real-time performance, implementing AI in your embedded applications offers massive benefits beyond the buzzwords. Nordic Semiconductor offers two unique technologies, Neuton models and Axon NPU, exclusively to our customers, to cover the industry's broadest range of devices, applications, and customer needs.

Why edge AI​

Improve responsiveness, battery life and security

Benefits of local AI

Longer battery life

Sending raw data wirelessly can easily become the most power-intensive part of your application. By processing locally, your battery can last significantly longer.

Privacy by default

If no data leaves the device, no data breach can ever happen. If data is only processed, not even a lost device will lose your data.

Lower latency

Instead of completing a round-trip to the cloud, data is processed locally, making the latency go from seconds to milliseconds.

Works offline

Local AI means that the feature will continue working, even when there is no connection to the internet available.

No cloud cost

By running AI locally on the device, you can lower or even eliminate cost associated with sending and hosting data in the cloud.

Scales indefinitely

Because the AI resides in every device, there is no need to account for the total number of devices.

Technology options​

CPU-run or NPU-accelerated edge AI for any application

Neuton models

Target:
Arm Cortex application core

Framework:
Proprietary format: Neuton models

Best for:
Classification, regression and anomaly detection in time-series sensor data

Model creation:
Upload your dataset to create custom models in Edge AI Lab

Hardware requirements:
Works on any Nordic SoC or SiP module

Neuton models

Axon NPU

Target:
Dedicated NPU core: Axon NPU

Framework:
TensorFlow Lite/LiteRT

Best for:
Audio, Imaging/Vision, high-rate time-series and advanced sensor fusion

Model creation:
Bring your own model or use no-data workflows in Nordic Edge AI Lab

Hardware requirements:
Integrated into selected products

Axon NPU

Or use both! Models on the CPU and models on the Axon NPU execute independently, meaning the nRF54LM20 can run two models at the same time.

Capabilities​

Ultra-low-power edge AI

Edge AI explanation graphic for webpage

"Edge AI" means different things depending on who's saying it. At Nordic, it means running inference on the same SoC that provides ultra-low-power wireless capabilities, in a battery powered device that can last for months on a single charge.

You will not be able to run a local LLM on the Axon NPU, but you can speed up tasks that were previously unfeasible to run on the Arm Cortex CPU, to make it viable for embedded devices. For example, running local voice activation algorithms, to enable higher levels of privacy. To let a voice-assistant selectively access a cloud-based or edge-server-based LLM, instead of always streaming audio from your device to the LLM.