Microcontroller (MCU)
A microcontroller, or MCU, is a small computer designed to live inside a product and handle a focused job reliably for years. It might read a temperature sensor, control a motor, listen for a wake word, or decide whether a vibration pattern signals a failing machine.
What is on the chip
Unlike a desktop computer, an MCU puts its essential parts on one inexpensive, low-power chip: a processor core, program storage (flash), working memory (SRAM), and hardware for connecting to sensors and electronics. Common families include ARM Cortex-M, STM32, and ESP32. Many run without a full operating system; their software starts at power-on and repeatedly performs a small set of real-time tasks.
Why it changes AI deployment
MCUs are the home of TinyML: machine-learning inference on devices with tight limits. A Cortex-M4 might have only a few hundred kilobytes of RAM and a few megabytes of flash—far less than a phone or Raspberry Pi. The model, its input buffers, intermediate calculations, and application code must all fit. Tools such as TensorFlow Lite for Microcontrollers convert a trained model into compact C/C++ data and code, commonly using 8-bit integer quantization instead of 32-bit floating-point math.
What this enables
- A battery-powered sensor node detects an abnormal motor sound locally and transmits only an alert.
- A wearable recognizes a gesture without sending motion data to the cloud.
- A wake-word detector listens continuously while consuming far less power than a larger processor.
Choosing an MCU means designing around memory, energy, and latency from the beginning. An oversized model can fail to compile, exhaust RAM during inference, drain a battery, or miss a real-time deadline. A well-matched MCU model delivers private, immediate decisions even when the device has no network connection.
A microcontroller (MCU) is a compact, low-power chip integrating a processor, memory, and interfaces for controlling sensors, actuators, and embedded devices. In Edge AI, MCUs run highly optimized TinyML models locally under tight memory, compute, and energy limits. They enable always-on tasks such as wake-word detection, vibration-based fault monitoring, and simple sensor classification without cloud connectivity.
Think of a microcontroller, or MCU, as the tiny, focused brain inside an everyday object. A washing machine uses one to follow its cycle. A thermostat uses one to watch the temperature. Unlike the powerful computer in a phone, an MCU is built to do a small job reliably while using very little electricity.
For on-device AI, this matters because an MCU can let simple AI run right where data is collected. A battery-powered sensor might recognize an unusual machine vibration, or a wearable might detect a gesture, without sending information to the internet. That makes devices cheaper, more private, and able to work even when offline.