IN Brief:
- Nordic's nRF7120 supports dual-band Wi-Fi 6, while the nRF7120E adds Wi-Fi 6E and AI acceleration.
- Independent Wi-Fi and multiprotocol radios support simultaneous reception of different wireless traffic.
- Nordic claims up to 20 times lower standby power consumption compared with its nRF70 generation.
Nordic Semiconductor has introduced the nRF71 series of wireless system-on-chips, combining Wi-Fi 6, Bluetooth Low Energy, Thread, and application processing in devices intended for battery-powered connected equipment. The first two products are the nRF7120, which supports dual-band Wi-Fi 6, and the nRF7120E, which adds tri-band Wi-Fi 6E and an Axon neural processing unit for local machine learning inference.
Nordic claims up to 20 times lower standby power consumption than its previous nRF70 Wi-Fi generation. The comparison addresses a significant constraint for battery-powered Wi-Fi equipment: a device must remain capable of maintaining or restoring connectivity even when its application processor and sensors spend long periods inactive. The energy consumed during those intervals can be as important to operating life as the power needed for an occasional data transmission.
The nRF71 architecture includes two independent radios with separate RF and processing resources. One handles Wi-Fi communication, while the other supports 2.4GHz multiprotocol operation using technology derived from Nordic’s nRF54 family. The radios can receive simultaneously, allowing a product to retain a Wi-Fi connection while listening for Bluetooth LE or Thread traffic. This removes the need to time-share all receive activity through a single radio path.
That separation is useful in systems combining several communications roles. A Thread border router may need to maintain contact with low-power devices on a local mesh while providing connectivity to a wider IP network. An access controller may use Bluetooth LE for local interaction alongside Wi-Fi for back-end communications. The two radios give the system hardware for those parallel requirements, although application firmware must still coordinate traffic, security, and power management.
Wi-Fi 6 introduces mechanisms such as scheduled communication that can help supported devices manage activity and sleep intervals. Their practical effect depends on access-point support, network configuration, and traffic patterns. The headline standby-power improvement is therefore a comparison between Nordic product generations, not a guarantee that an industrial sensor will operate for a particular number of months or years on one battery.
The nRF7120 supports communication in the 2.4GHz and 5GHz Wi-Fi bands. The nRF7120E extends the option to 6GHz through Wi-Fi 6E, providing access to another band where the local regulatory regime and network equipment permit it. Adding 6GHz capability changes antenna, coexistence, and regional approval considerations, so the suitable variant depends on the end product’s markets and radio requirements.
Both initial parts contain 4MB of on-chip non-volatile memory and 1MB of user-accessible RAM. These resources allow the wireless stack and application code to be brought together within one SoC, potentially avoiding a separate application processor in selected designs. The available memory must still accommodate firmware, communications buffers, application data, and any local inference model used by the product.
The nRF7120E integrates Nordic’s Axon neural processing unit. Nordic reports machine learning execution speeds between 10 and 17 times those of the device’s 256MHz Arm Cortex-M33 processor when running the workloads used for its comparison. That figure concerns suitable inference tasks and does not apply to ordinary embedded software or establish a comparable reduction in the energy required by a complete system.
Example inference workloads include keyword recognition, anomaly detection, and image classification. A condition-monitoring device might use a compact trained model to identify an unusual sequence of vibration measurements before sending an event record across Wi-Fi. Such processing can reduce the amount of raw data sent to another computer, but the usefulness of the classification depends on the sensor, training data, model accuracy, and acceptable rate of false detections.
Nordic also supports its Neuton tools for creating small machine learning models that can execute on the main processor, where a dedicated accelerator is unnecessary or unavailable. The company links that local software approach with nRF Cloud services for delivering model updates. Deployed products will still need suitable security controls and reliable update handling, particularly if an updated model changes how an industrial device interprets measurements.
Integrating multiple radio protocols and application processing on the same silicon may reduce the number of chips and interconnections needed on a circuit board. It does not remove the supporting engineering requirements: power supply behaviour, RF layout, antenna matching, electromagnetic compatibility, firmware maintenance, and application-specific testing still determine the suitability of the completed design.
The initial devices are sampling to selected customers, with developers directed to Nordic’s sales channels for access and technical information. Nordic is also presenting the family, including a Nordic-branded module, at electronica 2026. Broader volume availability and final product power budgets will depend on the device variant, supporting software, and results of customer integration work.
