Nordic Semiconductor has announced two wireless chips that combine Wi-Fi with a 256-MHz application processor, aiming to cut standby power by up to 20 times compared with its previous Wi-Fi parts. One version adds an on-chip neural processor for running machine-learning models locally, a practical step towards smaller connected devices that need not send every sensor reading to the cloud.
Watch Desk analysis
What happened
The nRF7120 and nRF7120E combine Wi-Fi 6, Bluetooth LE and Thread with the application processor on a single chip. The nRF7120E adds Nordic’s Axon neural processing unit; the nRF7120 has no NPU, but can still run edge-AI models using Nordic’s Neuton tooling.
All About Circuits reports that Nordic claims the NPU can run machine-learning models 10 to 17 times faster than the Cortex-M33 processor alone. The company names keyword spotting, anomaly detection and image classification as example uses. These are Nordic’s performance claims, not independent test results.
The design also combines two radios, allowing Wi-Fi and 802.15.4 traffic to be received at the same time. Nordic says devices could process microphone or sensor data locally and send only the result over Wi-Fi, keeping the radio asleep for longer. The article says the development kit and module are still listed as in preparation.
Why it matters
Putting connectivity, processing and, in one model, AI acceleration together could reduce the number of components a device maker needs. Local processing may also help battery-powered devices avoid transmitting raw data continuously. That is a useful engineering direction for connected sensors and appliances, even if a chip announcement alone cannot tell us how much battery life a finished product will gain.
Our read
This is more than an AI label stuck on a radio chip: the nRF7120E has a dedicated NPU, while both parts bring the host processor and wireless connections together. Nordic’s 20-times standby-power comparison is the headline claim; the real test will be how much power complete devices use while doing useful work. For designers, the combination is worth a look. For everyone else, the interesting question is whether edge AI can become small and frugal enough to fade into the background, where most good infrastructure belongs.
What to watch
- When the development kit and module become available, and at what price.
- Independent measurements of standby power and NPU performance in real workloads.
- Which products use the chips, and whether local processing delivers measurable battery or connectivity benefits.
Discussion spark: Would you trust a tiny device to process sensor data locally if it meant longer battery life and less data sent over Wi-Fi, or would you rather keep that work in the cloud?
Sources and evidence
- Source update (16 November 2026, 17:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.