Nordic Semiconductor introduced its nRF71 series chips on 3 November 2026, launching a family of single-chip wireless systems that pair low-power Wi-Fi and edge machine learning. The components are built for battery-powered hardware, cutting connected standby power consumption up to twentyfold compared with the company’s previous generation of standalone Wi-Fi companion silicon.
Key points
- The new systems combine Wi-Fi, Bluetooth Low Energy, Thread and application processing on one chip.
- The higher-end variant includes an on-chip neural processing unit that runs machine learning tasks up to 17 times faster than a standard processor.
- Nordic reports connected standby current as low as 8 microamps on 5 GHz networks.
- Hardware is currently sampling with selected commercial partners rather than shipping for general purchase.
A single-chip design for battery power
Nordic designed the silicon to replace multi-chip assemblies inside connected hardware. In an official release, the company announced the nRF71 Series to address the physical space and energy limits of small connected devices. Earlier systems based on the nRF70 line required a separate Bluetooth or microcontroller chip to act as a host, alongside a companion chip handling Wi-Fi.
The initial product range consists of two systems on chip: the dual-band nRF7120 and the tri-band nRF7120E. Both models integrate a 256 MHz Arm Cortex-M33 application core alongside a 256 MHz RISC-V coprocessor that handles peripheral emulation and time-critical background routines.
To handle embedded software without separate flash components, both chips include 1 MB of user-accessible SRAM and 4 MB of non-volatile memory. That capacity accommodates communication stacks, Matter network layers, customer application code and system updates.
Edge AI and tri-band radios
The flagship nRF7120E adds hardware-accelerated machine learning alongside Wi-Fi 6E connectivity across the 2.4, 5 and 6 GHz spectrum bands. Central to the design is Nordic’s Axon neural processing unit, a dedicated machine learning block built into the silicon.
Nordic stated that the Axon accelerator executes inference workloads between 10 and 17 times faster than an unassisted Cortex-M33 processor. The dedicated block targets tasks including keyword spotting, visual anomaly detection, acoustic surveillance and sensor fusion. Running these neural networks directly on the device allows hardware to transmit processed alerts instead of streaming raw telemetry, conserving radio bandwidth and battery life.
For hardware using the base nRF7120, which lacks the Axon core, developers can execute smaller custom models directly on the primary Arm processor using Nordic’s Neuton software tool. The vendor’s cloud service manages remote over-the-air model updates after deployment.
Simultaneous radio operations and low standby drain
Both silicon variants house separate radio paths for Wi-Fi and 2.4 GHz multiprotocol communications, each backed by dedicated processing circuitry. This design permits simultaneous reception across protocols, allowing a device to communicate over Wi-Fi with cloud infrastructure while actively scanning for local Bluetooth Low Energy or Thread packets.
In battery-powered smart home devices using Matter over Wi-Fi, the Bluetooth radio handles setup and commissioning while the Wi-Fi transceiver maintains network connectivity. Both chips also incorporate Near Field Communication support.
Power measurements published by the manufacturer report average standby current under DTIM 10 network cycles at 14 microamps when connected on 2.4 GHz frequencies, and 8 microamps on 5 GHz bands. Both figures rely on Target Wake Time mechanisms and a dedicated low-power Wi-Fi receiver, enabled as standard configuration within the company’s nRF Connect software development kit.
Sampling programmes and module packaging
Alongside the discrete silicon, Nordic developed the pre-certified nRF7120 module measuring 10 by 14 millimetres. The package incorporates the system on chip, radio matching circuits, filters, a diplexer, a transmit and receive switch, an internal PCB antenna, physical shielding and a 64 MHz reference crystal.
The module line uses identical physical pin layouts across dual-band and tri-band editions, enabling manufacturers to move between standard configurations and AI-accelerated Wi-Fi 6E without changing physical circuit boards. Nordic is pursuing regional certifications for the module across North America, Europe and Japan.
Nordic is restricting distribution of both chips and modules to commercial clients enrolled in project-based early sampling. The company has not set a date for broad retail availability.
