An automotive microcontroller (MCU) is a single chip with a processor, on-chip flash and RAM, and peripherals such as ADCs, timers, PWM and CAN or LIN controllers, built to run one control job in a vehicle reliably for fifteen years or more. MCUs sit inside nearly every electronic control unit (ECU), from seat motors and window lifts to brakes, airbags, battery management and motor inverters. What sets them apart from consumer chips is functional safety (ISO 26262), wide temperature grades (AEC-Q100), real-time response and long supply life.
This post explains what goes into an automotive MCU, walks through where they are used domain by domain, and looks at the market, the integration problems and where the architecture is heading.
Key takeaways
- MCUs are the quiet workhorses of the car: one chip per control job, with CPU, flash, RAM and peripherals on a single die, talking over CAN, LIN and increasingly Ethernet.
- Safety-critical MCUs use lockstep cores, ECC memories and built-in self-test to reach the ISO 26262 hardware metrics for ASIL B to ASIL D.
- The heavy computing (camera perception, cockpit graphics) runs on larger SoCs, not MCUs. Many popular articles, including the first version of this one, blur the two.
- Electrification and zonal architectures change the mix: fewer, more capable MCUs per car, with more security hardware and over-the-air update support.
- For engineers, the growth skills are functional safety, security (ISO/SAE 21434), RISC-V and AUTOSAR.
Introduction: the silent revolution of automotive microcontrollers
Microcontrollers are the unsung heroes behind the change from mechanical machines to software-defined vehicles. A modern luxury car is often said to contain 150 or more MCUs, each dedicated to tasks as simple as adjusting a seat or as complex as supporting automated driving. (Published counts vary widely; industry sources usually quote 70 to 150 ECUs in a premium car, and some ECUs contain more than one MCU.) These chips integrate a CPU, memory (flash and RAM) and peripherals (ADCs, PWM modules) on a single die, so they can read sensors, run control algorithms and communicate over CAN or LIN networks in real time.
What is inside an automotive MCU
| Block | Typical content | Why it matters in a car |
|---|---|---|
| CPU cores | Arm Cortex-M or Cortex-R, Infineon TriCore, Renesas RH850, increasingly RISC-V; often in lockstep pairs | Deterministic real-time response; lockstep detects random hardware faults |
| Memory | Embedded flash (hundreds of KB to around 20 MB), SRAM, both with ECC | Code runs in place; ECC corrects single-bit errors |
| Timers and PWM | Capture/compare units, motor-control timers | Injectors, inverters, motors, lamps |
| Analog | Multi-channel ADCs, comparators | Current, voltage, temperature and position sensing |
| Networking | CAN FD, LIN, FlexRay, Ethernet | Talking to other ECUs |
| Security | Hardware security module (HSM), crypto accelerators, secure boot | Firmware authentication, secure OTA, key storage |
| Safety | Built-in self-test, watchdogs, clock and voltage monitors, fault collection unit | Reaching ASIL targets under ISO 26262 |
Automotive parts are also qualified to AEC-Q100, whose temperature grades run from Grade 3 (-40 to 85 °C) up to Grade 0 (-40 to 150 °C). Engine-bay and powertrain MCUs typically need Grade 1 (-40 to 125 °C) or Grade 0.
MCU, SoC or domain controller?
The terms are often mixed up, so it helps to separate them before looking at examples.
| Device | Typical job | Examples of families |
|---|---|---|
| Microcontroller | Real-time control of one function, embedded flash, low power | Infineon AURIX, NXP S32K, Renesas RH850, ST SPC5 and Stellar, TI, Microchip dsPIC33 |
| Application SoC | Perception, infotainment, graphics; external DRAM, runs Linux, Android or QNX | Qualcomm Snapdragon cockpit platforms, NVIDIA DRIVE, Tesla FSD computer, AMD Ryzen embedded |
| Vehicle network processor or domain controller | Gateway, zonal or domain control, mixes real-time and application cores | NXP S32G, Renesas R-Car, Infineon AURIX TC4x in zonal roles |
Some examples below involve SoCs or other ICs rather than MCUs. They are kept, with the device type corrected, because they show where the control and computing work in a car is done.
Historical context
| Period | Milestone |
|---|---|
| 1970s | Microprocessor engine control arrives. The first version of this post named the Intel 8048 in Ford's EEC-III engine control system, managing fuel injection. That attribution is not confirmed here: many histories credit Ford's EEC family with Toshiba's TLCS-12 for the first generation and Intel's 8061 for EEC-IV (1983). |
| 1986 to 1990s | Bosch introduced CAN in 1986, the first production car with CAN followed in 1991, and ISO 11898 was first published in 1993. |
| 2000s | CAN became standard across ECUs, enabling features such as electronic stability control that need several ECUs to share data. |
| 2020s | Multi-core MCUs, for example with Arm Cortex-R52 cores, are used to meet ASIL D safety requirements. |
Why MCUs matter
Features such as Tesla's Autopilot or BYD's Blade battery packs depend on many MCUs doing control work around the main computers, for example in braking, steering, battery monitoring and power conversion. MCUs balance performance (mainstream automotive parts run at up to about 300 MHz, with newer ones such as AURIX TC4x quoted at up to 500 MHz), power efficiency (sleep currents in the region of 10 µA are commonly quoted) and cost (from around a dollar for a basic body node to $100 or more for high-end parts, as reported; prices vary with volume and memory size).
Microcontroller applications in automotive domains
The examples below were collected for the first version of this post. Where a public source shows a detail is wrong, it has been corrected. Where it could not be checked, the claim is kept and marked "as reported".
Safety systems save lives in milliseconds
Anti-lock braking is the classic MCU job. The first version of this post stated that Bosch's ABS 8.1 system uses Infineon AURIX TC2xx MCUs to prevent wheel lock-up by modulating brake pressure 20 times per second, and that this reduced fatal crashes by 15% in the EU (EEA, 2020). These details are as reported and not re-verified; in particular the 15% figure and its source should be treated with care.
Airbag control units read crash sensors and must decide within tens of milliseconds. MCUs such as Renesas RH850/P1x are reported to analyse crash data within 0.05 seconds to trigger airbags. They rely on safety mechanisms such as lockstep cores and ECC memory so that a hardware fault cannot cause a false deployment.
Dual-core lockstep architectures (for example in NXP S32S processors) provide redundancy, which matters for systems such as electronic parking brakes and steering.
From sensors to steering
Tesla's Hardware 3 Autopilot computer processes data from 8 cameras, 12 ultrasonic sensors and 1 radar (the sensor set of that era). Correction: it is built around Tesla's own FSD chip, an SoC designed by Tesla and manufactured by Samsung, not a Samsung Exynos-based MCU. At its 2019 Autonomy Day, Tesla quoted the computer as processing 2,300 frames per second.
Innoviz's MEMS-based lidar is reported to use AMD (Xilinx) Zynq UltraScale+ devices, which are FPGA-based SoCs, to adjust laser pulse timing dynamically, improving object resolution by 40% (SAE, 2022, as reported).
On ISO 26262: the standard does not set a fixed diagnostic coverage for a named function such as automatic emergency braking. Correction: the first version said ISO 26262 mandates diagnostic coverage of 99% or more for ADAS functions. What the standard does is assign an ASIL to each safety goal through hazard analysis; for ASIL D it targets a single-point fault metric of at least 99% and a latent fault metric of at least 90% (ASIL B: 90% and 60%). AEB safety goals are often rated ASIL B to ASIL D depending on the analysis.
Optimisation at every drop of energy
Toyota states that its 2.5-litre Dynamic Force Engine reaches 40% thermal efficiency, and 41% in hybrid form, against roughly 30% often quoted as typical for petrol engines. The engine controller is reported to use Renesas RH850 MCUs for functions such as precise variable valve timing.
Battery management is another large user. The Tesla Model 3 Long Range pack is reported to contain 4,416 cells. The first version of this post said Texas Instruments BQ7961x MCUs monitor them. Correction on device type: the TI BQ7961x parts are battery cell monitor ICs, not MCUs; they measure cell voltages and temperatures and report to a separate BMS microcontroller. Which monitor chips Tesla actually uses is not confirmed here.
Porsche Taycan's 800 V system is reported to use Infineon AURIX TC3xx MCUs to control its traction inverters, switching the power transistors at around 20 kHz. The first version said SiC MOSFETs; whether a given Taycan model year uses SiC or silicon IGBTs should be checked.
The digital cockpit
BMW's iDrive 8 runs on the BMW Curved Display, which combines a 12.3-inch information display and a 14.9-inch control display. Correction: the first version described dual 14.9-inch screens. It also named the Qualcomm Snapdragon 8155, which is a cockpit SoC rather than an MCU, with a Hexagon DSP that can handle voice processing; whether iDrive 8 uses that chip is as reported and not confirmed here. 5G connectivity was offered on some iDrive 8 cars.
Over-the-air updates show the value of updatable firmware. Correction: in Tesla's terminology "MCU" means media control unit, the infotainment computer, which in recent cars uses an AMD Ryzen processor. Autopilot neural networks run on the separate FSD computer, not on the Ryzen. Lane detection improvements credited to Tesla's 2023.26.1 update are as reported.
Balancing power in hybrids
Mercedes-Benz's EQ Boost 48 V system is reported to use STMicroelectronics SPC58 MCUs to manage its starter-generator, recovering up to 12 kW of energy during braking (EPA, 2021, as reported). Mercedes uses both belt-driven and integrated starter-generators under the EQ Boost name, so the figure depends on the engine.
Lightyear 0's roof-mounted solar panels are reported to use NXP LPC5500 MCUs for maximum power point tracking (MPPT). Correction: Lightyear claimed up to 70 km of solar range per day in good conditions, not per week. Lightyear stopped production of the Lightyear 0 in early 2023.
Adapting to the road
GM's MagneRide 4.0, used on the Cadillac CT5-V, changes the viscosity of magnetorheological fluid in its dampers with a magnetic field. It is reported to use Microchip dsPIC33 controllers and to adjust the dampers within about 1 ms, improving handling on uneven roads.
Audi's quattro and torque-vectoring systems are reported to use Renesas RH850/E2x MCUs to distribute torque between wheels, improving cornering grip by 30% (Audi AG, 2023, as reported).
Comfort meets efficiency
Tesla's Cabin Overheat Protection keeps the cabin below about 40 °C (105 °F) after the car is parked by running the fans or air conditioning. The first version said it uses Texas Instruments MSP430 MCUs and draws power from the 12 V battery; both details are as reported and not confirmed here.
Audi's Digital Matrix LED headlights use a digital micromirror device with about 1.3 million micromirrors to shape the beam and project light markings, such as lane guidance, onto the road. Correction: the first version called Infineon's TLF35584 an MCU. It is a safety power-supply IC that powers and supervises MCUs such as AURIX. The micromirror chip itself comes from Texas Instruments' DLP family. Which Audi models carry Digital Matrix LED (the first version named the e-tron GT) should be checked per market.
Guarding the connected car
GM's Ultifi software platform is reported to use Renesas RH850/U2A MCUs with hardware security modules to authenticate firmware using AES-256, blocking unauthorised OTA updates. In practice, firmware authentication normally relies on digital signatures, with AES used for encryption and message authentication codes.
Tesla's Sentry Mode watches the car while it is parked and alerts the owner through the Tesla app if a threat is detected. Correction: the first version said it monitors CAN bus traffic for anomalies. Sentry Mode uses the car's external cameras and records events; it is not a CAN intrusion detection feature.
Industry-leading microcontrollers: the outlook
Market trends
The global automotive MCU market was projected to grow from $7.4 billion in 2023 to $12.1 billion by 2028 (IC Insights, 2023, as cited in the first version; IC Insights is now part of TechInsights, and the figures are not re-verified). Electric vehicles drive demand: high-performance MCUs for 800 V battery systems were reported to be growing at 14% CAGR, faster than MCUs for combustion engines (Statista, 2023, as reported). Market sizes differ a lot between analysts, depending on whether they count only MCUs or include processors and SoCs, so compare like with like.
A simple sizing example (illustrative)
To see why safety MCUs cost more than body MCUs, consider an electric power steering ECU running a 20 kHz current loop. Each loop has 50 µs. If the control code takes 15 µs at 300 MHz and the safety software (plausibility checks, monitoring) another 10 µs, the core is about 50% loaded before diagnostics, communication and margin. A lockstep core does not add throughput, it doubles the silicon to detect faults, so the designer pays for safety in area and power, not speed. These numbers are illustrative only.
Integration challenges
Cybersecurity risks
- Example: researchers Charlie Miller and Chris Valasek showed in 2015 that they could control a Jeep Cherokee remotely through its Uconnect infotainment unit, and FCA recalled about 1.4 million vehicles (NHTSA recall 15V-461). Correction: the first version gave the year as 2022.
- Solution: MCUs and secure elements designed for ISO/SAE 21434 cybersecurity processes, with hardware firewalls, secure boot and HSMs (for example Microchip's CryptoAutomotive™ security ICs), plus network segmentation through gateways.
Power efficiency
- Thermal management: Ford's Mustang Mach-E is reported to use TI TMS570 MCUs with dynamic voltage scaling to reduce heat in its 88 kWh (usable, extended range) battery pack.
Software complexity
- AUTOSAR adoption: Volkswagen's MEB platform is reported to have standardised on AUTOSAR Classic for 70% of its ECUs, cutting development time by 40% (VW Group, 2022, as reported).
Future trends
AI at the edge
The first version stated that NXP's S32G3 integrates a 4-TOPS NPU for real-time traffic prediction, reducing cloud dependency. NXP describes the S32G family as vehicle network processors for gateways and domain control; the NPU claim could not be confirmed and should be checked against NXP's data sheet. AI acceleration is, however, appearing in automotive MCUs and processors more generally.
RISC-V disruption
Western Digital is reported to use SiFive E6-A RISC-V cores in automotive storage controllers, cutting licensing costs by 60% (The Linley Group, 2023, as reported). More broadly, several automotive chip makers have announced RISC-V cores for future MCUs, and the open instruction set lets them add custom safety and security extensions.
Zone architectures
Instead of one ECU per function, zonal designs group ECUs by location in the car and connect them to a few central computers over Ethernet. Tesla's Model Y is reported to consolidate around 70 ECUs into 4 controllers with a simplified wiring harness. Correction: the AMD Ryzen chip in recent Teslas powers infotainment, not the body or zonal controllers.
Sustainability and process
The first version named an STMicroelectronics "STM32Auto" built on 40 nm FD-SOI. ST's automotive MCU line on FD-SOI that we could confirm is Stellar, built on 28 nm FD-SOI with embedded phase-change memory, so the product name and node should be checked. A 50% reduction in leakage compared with 65 nm parts was claimed in the first version and is not re-verified; FD-SOI does give lower leakage and supports body biasing.
Common misconceptions
- "The Autopilot chip is an MCU." Perception and cockpit computers are SoCs with external DRAM. MCUs do the real-time control around them.
- "ISO 26262 certifies a chip." It certifies a process and a safety case. A chip is "developed according to" ISO 26262 and comes with a safety manual the system designer must follow.
- "More MHz is better." For control, deterministic interrupt latency, memory with no wait states and good timers matter more than peak clock rate.
- "Security is the gateway's job." Every ECU that accepts updates needs secure boot and key storage.
Driving the future with smarter silicon
From life-saving ABS to supporting AI-driven automation, microcontrollers are the backbone of automotive innovation. As vehicles become connected, electrified platforms, MCUs will face greater demands for speed, security and efficiency. For engineers, staying ahead means mastering RISC-V, AI acceleration and cybersecurity, while automakers must balance performance with supply chain resilience.
FAQ
How many microcontrollers are in a car?
It varies widely. A basic car may have a few dozen ECUs; premium cars are usually quoted at 70 to 150 ECUs, and some contain several MCUs each.
What is the difference between an automotive MCU and a consumer MCU?
Automotive parts are qualified to AEC-Q100, built and documented for ISO 26262 safety work, supported for 10 to 15 years or more, and include safety hardware such as lockstep cores, ECC and self-test.
Who are the main automotive MCU suppliers?
Infineon, NXP, Renesas, STMicroelectronics, Texas Instruments and Microchip are the most frequently named, with rankings that change from year to year.
Why is RISC-V interesting for automotive?
It removes per-core licensing fees and lets chip makers add their own extensions, but it needs a mature safety-qualified tool chain and software ecosystem to compete with Arm, TriCore and RH850.
References
As cited in the first version of this post; not re-verified unless noted.
- European Environment Agency (EEA). (2020). Impact of ABS on Road Safety.
- SAE International. (2022). LiDAR Resolution Improvements via MEMS.
- Audi AG. (2023). Torque Vectoring in the Audi e-tron GT.
- IC Insights. (2023). Automotive MCU Market Forecast.
- NHTSA. Jeep Cherokee cybersecurity recall, recall 15V-461 (2015; year corrected from 2022).
Your comments will be moderated before it appears here.