The Five MCUs We’d Be Thrilled to Unwrap for Christmas in 2025

Microcontrollers have changed. Once, “best” meant the lowest quiescent current you could afford, a familiar toolchain, and just enough RAM to avoid a re-spin. Today, “best” spans everything from tiny always-on controllers for sensor fusion, wireless SoCs that manage complex radios, and feature rich AI MCUs that can run vision/audio interfaces.

There is no single winner, the ideal part depends on your power budget, your memory and I/O needs, what models you plan to run, and more.

Still, we at ipXchange have tried our best to distill the field down into five forward-looking MCU series that feel “next-generation”, silicon that we would love to find in a stocking on a Christmas morning.

Disclaimer: We are biased. We love microcontrollers that punch above their weight, sip power, and still have the headroom for modern AI/ML. These are our favourites, from benches, booths, and late night chats with engineers, in no particular order.

1. Alif Semiconductor Ensemble, for battery-friendly AI that scales

If you want an MCU family that grows with your ambition, Alif Ensemble is the family that spans from simple object recognition, to “baby transformers” models. The latest Ensemble devices, the E4, E6 and E8, add an Arm Ethos-U85 NPU and official ExecuTorch runtime support, so you can build small language models and transformer operators without moving to a power-hungry SoC.

Alif’s aiPM technology powers only parts of the chip that are in immediate use, allowing their devices to have microprocessor-level performance with low-power MCU energy budgets.

The flagship E8 is their latest fusion device. Dual Cortex-M55 cores, Ethos-U85 + dual Ethos-U55 NPUs for heavy AI workloads and dual Cortex-A32 for higher-level tasks when you need them. You can keep an MCU-first workflow for sensing and control, then fire up heavier perception or UX when the job demands it.

At the other end of the range, the E3 charms us with practical vision. OpenMV’s AE3 camera runs on E3’s dual M55 + dual U55 cores, giving you MicroPython computer vision that fits on a one-inch square. Check out this article on the OpenMV AE3 & N6 cameras for a feel of Alif-powered vision.

ipXchange deep dives: The Microcontroller With 3 NPUs Inside?,How to Boost AI Model (with Edge Impulse),Alif Ensemble DevKit Blinky Tutorial,The Engineer’s Guide to Vision AI: How to Build and Deploy Custom Models (with Alif Semiconductor & Roboflow)

Why we like it: Real AI/ML throughput in an MCU power profile, a credible development path, and growing board support that blurs the line between proof-of-concept and production builds.

2. Nordic Semiconductor nRF54L, for wireless communication that does more with less

Not everything is about AI. Nordic’s nRF54L family is the natural successor to the nRF52, boasting more compute per microwatt, improved radio efficiency, and a new low-power multiprotocol radio for wireless connectivity. We’re already seeing developers and teams move prototypes across.

The RoyalBlue54L Feather showcases an open, Zephyr-ready board with an integrated debugger. Together with the maker debug board, it gives you two quick routes to get an nRF54 on the bench fast. For a real-world build, look at Sentinel-Fall, a privacy-focused fall detection system fusing a 60 GHz radar and PIR with an ESP32 host and uses an nRF54L15 for BLE connectivity.

One of our very own in-house design engineers, Elliott has even been working on a tiny USB-C connected nRF54L15 receiver, complete with an Ignion Virtual Antenna®. Here’s a sneak peek:

Elliott says: “Developing with the nRF54 series is dead simple. The technical documentation is great, clean reference schematics and layouts are easy to find. The nRF SDK for VSCode makes uploading firmware and accessing examples easy as! Overall, one of the best experiences developing with a microcontroller I've had in years.”

ipXchange deep dives: Ultra-low power wireless connectivity with Nordic,nRF54L vs nRF52 Comparison

Why we like it: Nordic’s toolchain and connectivity expertise remain a safe bet, and the nRF54 extends that with modern computers and IO that makes tiny products feel responsive rather than constrained. Subscribe to ipXchange's YouTube channel to follow our upcoming series designing with the nRF54.

3. Ambiq Apollo510, for wearable AI without charger anxiety

Ambiq’s specialty is wringing real work out of microwatts. Their Subthreshold Power Optimized Technology (SPOT®) platform makes Apollo510 sip power, and with Arm Cortex-M55, the Apollo510 delivers a staggering 10x performance improvement and up to a 30x jump in power efficiency for AI workloads compared to previous generations.

The newly announced Apollo510B variant brings Bluetooth 5.4 on a dedicated network processor so the application core can remain asleep until there is something worth doing.

On the AI side, the M55 + Helium combo gives credible acceleration for keyword spotting, beamforming, sensor fusion and vision whilst keeping cost and power consumption low. The wireless Apollo510B pushes that into connected designs without destroying your battery life. This write-up walks through the dev experience.

A fun project we have seen using Apollo4, is the Xiaomi Mi Band 8 community hack, where developers flashed custom firmware due to the band’s Apollo4 Blue Lite MCU and exposed test pins. Previous generation chip, but same design philosophy.

ipXchange deep dives: Ambiq Apollo510 MCU takes on next-level endpoint AI,How to build power-efficient GUIs with Ambiq MCUs,Apollo4 vs Apollo5: Ambiq's MCUs in action

Why we like it: Weeks on a charge with enough acceleration for useful edge AI, KWS, audio analytics and vision.

4. Infineon PSoC™ Edge E84, for configurable multi-purpose control that now speaks ML

Projects are about graphics, HMI and security as they are about raw processing power. Infineon’s PSoC™ Edge E84 series pairs an Arm Cortex-M55, Cortex-M33, Ethos-U55 and Infineon’s NNLite. The MCUs also include a low-power 2.5D GPU, as well as Edge Protect Category 4 providing security features like pre-configured credentials and a hardware secure enclave. Infineon has been pushing ModusToolbox™ software, and now NVIDIA TAO support for those wanting a template route to vision and audio models without building a toolchain from scratch.

We love seeing real builds, and there are plenty using the PSoC™ Edge E84. Start with Edgi-Talk, an open source AI-powered handheld assistant complete with a MIPI-DSI capacitive touchscreen. Or how about a sensor robot for radar surface detection, combining motor control with an HMI layer for visualisation and remote access.

ipXchange deep dives: Why Infineon's PSoC Edge Microcontroller Is Great For AI,Infineon + Thistle Technologies on Securing the Edge with OPTIGA™ Trust M

Why we like it: A jack-of-all-trades workhorse with a sensible split between low-power control and accelerated tasks. ML is there when needed, but the platform still earns its keep when only robust control, HMI and IO is required.

5. STMicroelectronics STM32N6, for vision MCU design with big RAM headroom

Everybody knows ST. The toolchains are familiar (we love STM32CubeIDE here at ipXchange), the ecosystem is huge, and parts are easy to source. STM32N6 sits right in that comfort zone, while pushing the envelope for on-device vision processing, pairing an Arm Cortex-M55 + Helium with ST’s Neural-ART accelerator and a very generous amount of SRAM - 4.2MB!

It’s not difficult to find projects using STM32 MCUs. The CamThink AI Camera NeoEyes NE301 is a low-power edge AI camera built on the STM32N6, providing 0.6 TOPS of compute - enough to run local person detection and gesture recognition without sending any data to the cloud. OpenMV’s N6 camera sits alongside the AE3 as its larger and more expandable companion. On the tools side, Edge Impulse shipped early support for the N6, providing an intuitive path from data collection to working firmware on the STM32N6570-DK dev kit.

ipXchange deep dives: STM32 Edge AI Microcontroller: Low Power Computer Vision,How to Handle Real-Time AI Decision-Making on Embedded Hardware

Why we like it: A vision-enabled MCU from a vendor everyone already trusts, with serious on-chip RAM and first-class community support & tooling.

Honourable mentions

Narrowly missing the list is the NXP i.MX RT700, which delivers a Swiss Army knife of capabilities including Tensilica® HiFi 4 DSP for audio processing and a 30-70% reduction in power consumption compared to previous generations, as well as the Renesas RA8P1 powered by beefy Arm Cortex-M85 and companion Cortex-M33 with 1 Mb embedded on-chip MRAM.

Lists are fun, prototypes are better. Our five picks cover a spread of MCUs built for a variety of applications, with a clear trend toward smarter, lower power devices that keep intelligence on the board.

Ultimately, the right choice is the one that matches your project’s specific requirements the best, so if your favourite is not here, tell us what you are building and why it earns a spot. Then grab a dev kit, clone a project, measure, tweak and share your results with the ipXchange community so the rest of us can learn, build alongside you.