Where Microcontrollers Actually Go: Every Industry That Runs on MCUs
The average new car rolling off a 2024 assembly line contains somewhere between 1,400 and 3,000 microcontrollers. A Tesla Model S has more compute spread across its chassis than the entire Apollo Guidance Computer program. A single power window? It’s got its own MCU.
If you’ve been tinkering with an Arduino or ESP32 and wondering whether this hobby has a real career attached to it, the answer is yes, but the map matters. Firmware jobs cluster in specific industries, each with its own tools, certifications, and culture. Pick wrong and you’ll spend two years learning skills that don’t transfer.
This is a tour of where microcontroller applications actually live: 8 industries at a glance, 3 worth a deeper look, and a frame for picking which one to bet on.
The Scale of MCU Deployment
Roughly 30 billion microcontrollers shipped in 2024, generating around $25B in revenue (per IC Insights and Yole estimates, though these numbers shift year to year as the auto cycle and consumer demand swing). That’s about 4 MCUs for every human on Earth, every year. Cumulatively, there are probably 200+ billion MCUs in active use.
Here’s the rough breakdown of where they go:
MCU Shipments by Industry (approximate share)
Automotive ████████████████████ ~40%
Consumer/IoT ██████████████ ~28%
Industrial ████████ ~16%
Appliances ████ ~8%
Medical ██ ~4%
Other ██ ~4%Automotive dominates revenue more than units because car-grade parts cost 10x what a consumer MCU does.
The 8 Major Industries (Quick Tour)
Automotive. Engine control units (ECUs), brake-by-wire, ADAS cameras, infotainment, battery management on EVs. A modern brake controller might run an Infineon AURIX TriCore with redundant cores doing lockstep execution. Safety-critical, AUTOSAR-heavy, slow product cycles.
Consumer Electronics. Earbuds, wearables, gaming peripherals, smart toys. The Apple H2 chip in AirPods Pro is technically an SoC but the firmware discipline is pure MCU work: tight RAM, audio DSP, BLE, battery management. Fast cycles, brutal cost pressure.
Industrial Automation. Picture a PLC, a servo drive, a robotic arm, or a factory sensor node — an ABB robot joint controller might run an STM32H7 with EtherCAT. Long product lifetimes (10-20 years), tons of legacy code, and increasingly cloud-connected.
Medical Devices. Pacemakers, continuous glucose monitors, insulin pumps, hearing aids, hospital infusion equipment. An nRF52840 inside a Dexcom G7 doing sensor fusion and BLE. Heavily regulated, high margins, slow to ship.
Smart Home / IoT. Thermostats, smart locks, lightbulbs, leak sensors, pet feeders. ESP32 or Nordic chips dominate, and the Nest Thermostat’s main board is essentially a glorified MCU with a display. This is also where new connectivity stacks like Matter, Thread, and satellite-backhauled BLE are landing first.
Aerospace & Defense. Avionics, missile guidance, satellite payload control, drone flight controllers. Often radiation-hardened parts (BAE RAD750, Vorago) or qualified versions of commercial MCUs. Highest certification barriers in the industry.
Appliances. Washing machines, microwaves, coffee makers, dishwashers usually run on 8-bit or low-end 32-bit parts (Microchip PIC, STM8, Renesas RL78). Cost is everything: a dishwasher BOM fight over $0.03 is a real meeting that happens.
Agriculture. Soil moisture sensors, irrigation valves, livestock trackers, autonomous tractor subsystems. Low-power MCU + LoRaWAN or cellular is the typical stack. Quietly one of the fastest-growing IoT segments.
Deep Dive: Automotive
Automotive is the largest MCU consumer by both revenue and growth rate. EVs need more silicon than ICE cars (battery management alone can involve 20+ MCUs), and ADAS keeps adding nodes.
Walk through a single power window module. It’s probably an Infineon TriCore or NXP S32K running AUTOSAR Classic. The firmware handles CAN messaging, pinch detection (using motor current sensing), anti-trap algorithms, and diagnostic reporting. All of it has to pass ISO 26262 functional safety review, sometimes at ASIL-B. Every line of C has documented requirements traceability. Every change re-triggers test cases.
The language is C. Some teams are piloting Rust (Volvo, a few Tier 1s) but production is still 95% C. Tooling means Vector CANoe, Lauterbach Trace32, MISRA C checkers, ASPICE process compliance.
Junior roles mostly live at Tier 1 suppliers: Bosch, Continental, Denso, ZF, Aptiv, Magna. The OEMs (Ford, VW, GM) hire too but lean more on integration. Pay is solid, work-life balance is generally good outside of crunch, and the work feels consequential because brakes have to work.
Downsides: process overhead is heavy, and shipping a feature can take 2-3 years.
Deep Dive: Medical Devices
Medical is the highest-margin firmware sector and one of the most recession-resistant. People don’t stop needing insulin pumps because the Fed raised rates.
A continuous glucose monitor (CGM) like the Abbott Libre or Dexcom G7 is a clean example:
[Glucose Sensor] --> [nRF52 MCU] --> [BLE Radio] --> [Phone App]
|
v
[Flash Storage]
[Battery Mgmt]The MCU runs maybe 14 days on a coin cell. Firmware handles ADC sampling, calibration algorithms, sensor drift compensation, BLE pairing, encrypted data logging, and over-the-air updates. All of it falls under IEC 62304 software lifecycle and FDA Class II (or III for implantables) regulation.
You’ll spend more time writing tests, traceability matrices, and risk analyses than writing code. A 2,000-line module might have a 200-page verification report.
Employers: Medtronic, Abbott, Dexcom, Stryker, Boston Scientific, plus a long tail of startups doing wearables and digital therapeutics. Pay is strong, hiring is steady, and the work has obvious meaning.
The catch: cycles are slow (18-36 months to ship), and if you hate paperwork, you’ll be miserable.
Deep Dive: Industrial IoT
Industrial IoT is where traditional embedded meets modern connectivity. It’s also where generalists thrive because you need firmware + radio + a bit of cloud + sometimes ML, all in one head.
A concrete example: a predictive maintenance vibration sensor strapped to a factory pump. It might pair an STM32L4 (sipping power) with a MEMS accelerometer like the ST LIS2DW12. A tiny TensorFlow Lite Micro model runs anomaly detection on-device, and a LoRaWAN or BLE radio kicks only the interesting events upstream. Battery target: 5+ years on a single D-cell.
The interesting trend: more inference is moving to the MCU itself because backhaul is expensive and latency matters. ARM’s Cortex-M55 with Helium extensions and Ethos-U55 NPUs are showing up in real designs.
Connectivity is the other dominant theme. Factories want TSN (time-sensitive networking) for in-plant determinism. Cloud side runs on MQTT or proprietary protocols. Remote assets increasingly use low-power wide-area options. For a fleet reporting back through a BLE-to-satellite link, duty-cycle tradeoffs and terrestrial transmission patterns matter as much as the sensing code.
Employers: Siemens, ABB, Rockwell, Honeywell, plus the entire startup ecosystem (Augury, Samsara, Particle, Tulip). Pay sits a notch below auto/medical but barrier to entry is lower and the work moves faster.
Which Industries Should You Learn For?
Here’s the cheat sheet:
Industry | Growth | Pay | Barrier | Best If You Like...
-----------------|--------|-------|---------|--------------------
Automotive | High | $$$$ | High | Safety, scale, C
Medical | High | $$$$ | High | Rigor, low-power, BLE
Industrial IoT | High | $$$ | Medium | Connectivity, ML, breadth
Consumer/IoT | Med | $$$ | Low | Fast cycles, UX
Aerospace | Med | $$$$ | Highest | Determinism, certs
Appliances | Low | $$ | Low | Cost optimization
Agriculture | High | $$$ | Low | Remote systems, powerSalaries date fast and vary wildly by region. Check Levels.fyi or Glassdoor for current numbers in your market before betting on any of this.
A few honest recommendations:
If you’re starting from Arduino/ESP32 hobby projects: consumer IoT or smart home is the natural on-ramp. Your skills already transfer. Build 2 or 3 portfolio projects with proper BLE provisioning, OTA updates, and a real cloud backend, and you’re hireable.
If you want the highest-paying long-term bet: automotive or medical. Both reward depth, both have hiring momentum, both are hard to outsource. Be ready for slow culture and heavy process.
If you’re a generalist who likes touching everything: industrial IoT. You’ll do firmware, radio, a bit of cloud, sometimes ML, sometimes mechanical integration. It’s the closest thing embedded has to full-stack.
Skills that travel across all of them: C, an RTOS (FreeRTOS or Zephyr is fine), comfort with a logic analyzer and oscilloscope, peripheral drivers (SPI/I2C/UART/CAN), and the ability to read a datasheet without flinching. Anything you build on top of the Zephyr reference application or similar open codebases will sharpen all of those at once.
Pick One Industry in the Next 6 Months
MCUs are everywhere, but firmware careers aren’t. They cluster in industries with sharply different cultures, certifications, and pay bands. The worst move is to stay generic for 3 years and end up competing against people who specialized.
Pick one in the next 6 months. Reshape your next 2 or 3 hobby projects to look like work in that industry: if it’s medical, build something low-power with BLE and proper logging; if it’s automotive, learn CAN and AUTOSAR basics; if it’s industrial IoT, ship something that runs for a year on a battery and reports to a cloud dashboard.
The hobbyists who get hired are the ones whose projects look like the job.
Hubble Network gives MCU projects global satellite connectivity on standard Bluetooth radios, so battery-powered devices can report from anywhere without cellular modems or gateways. See how it works →