How Far Domestic MCU Substitution Has Actually Got: Share, Category and Verification Cost
Trends · Published 2026-09-17 · Updated 2026-09-17
Share alone can't describe substitution progress; look at the easiest categories, the genuinely hard automotive ones, and the hidden wall of verification hours.
Nobody can answer how far domestic MCU substitution has got with a single percentage. The ICNAV database stores share figures for five international vendors (Omdia 2025) and none for Chinese makers, so a sentence such as "domestic MCUs have reached X%" cannot be written from our own data. What can be written is more useful: which categories bend today, which one does not and why the obstacle there is liability rather than silicon, and the verification hours still paid after a pin-compatible part arrives. Three angles give a map of the boundary, not a progress bar.
Share: what the stored numbers actually describe
Our MCU share records are Infineon 23.2%, NXP 18.7%, STMicroelectronics 16.8%, Renesas 14.3% and Microchip 10.2% — the sum of those five, long tail excluded. What holds the leaders up matters more: three of the five carry automotive notes, Infineon as the automotive semiconductor number one for six consecutive years, Renesas as main RH850 supplier, NXP as the automotive processor leader. The largest share sits exactly where substitution is hardest.
The other side of the market is fragmented by design. Microchip is catalogued as strong in the 8/16-bit long tail, and families such as AVR, PIC16 / PIC18 and RL78 sit comfortably in it, so general-purpose and niche volume was never owned by five vendors. For Chinese makers the only wording we can quote is GigaDevice as the number one domestic MCU shipper by volume — an ordering, not a share. Any percentage seen elsewhere needs its scope stated; the usual trick is to count white-goods and consumer shipments and call that the whole MCU market. The MCU category lists 10 companies, leaders and challengers on one sheet.
Category: supply-driven volume moves, liability-driven volume does not
The categories that bend share a profile: qualification in months, failures absorbed inside the OEM, short expected life, switching decided on price and delivery. Consumer electronics, home-appliance control, simple control loops, plus general-purpose 8/16-bit and mid-range Cortex-M3 and M4 parts live there. Domestic lines are climbing out of that band: GD32E5 is catalogued as the domestic mainstay for motor control and digital power, GD32H7 as the flagship of domestic high-performance MCUs, GD32F4 as pin-compatible with STM32F4. Those records prove availability, not victory.
The band that does not bend is written down concretely. Powertrain and chassis is RH850's position in the database, body and domain control is S32K3, the automotive main controller is AURIX TC4x — catalogued as the world's number one in that role — and cockpit and cluster display is TRAVEO T2G. The wall is not the core. It is qualification cycles, ownership of the defect rate when a batch misbehaves in the field, since stop-line and recall liability is a contract clause rather than a datasheet parameter, and a decade-long supply commitment with predictable change notices. One level higher sit security parts such as EdgeLock SE05x, OPTIGA TPM and the LPC5500 family, where a switch replaces a key-injection and production-line procedure instead of a piece of silicon.
Substitution therefore advances along responsibility boundaries, not clock speeds. Infineon will not lose its automotive position soon; what moves is the volume already competing on lead time and price.
Verification cost: pin compatibility is the first kilometre
Pin compatibility removes a board redraw; it removes no retest. The hours hide in five places.
- Registers and peripherals: trigger relationships, DMA mapping and interrupt vectors rarely match name for name, and code that compiles is not code whose timing is right.
- Clocks and security packages: clock trees, boot configuration and read protection, OTP and the secure-boot chain, plus key injection that now has to be rebuilt.
- Compiler and library versions: HAL and RTOS pinning, and timing drift when optimisation levels change.
- EMC and whole-machine retest: identical pins do not mean identical drive strength or slew rate, so emissions and immunity are re-measured on the product.
- Ramp and lot consistency: several small lots and refreshed failure data before anyone calls the substitution finished.
The sequence falls out of that list. Start with safety-free general 8/16-bit and simple control parts, then pin-compatible mid-range parts of the GD32F4-versus-STM32F4 kind where only peripheral drivers move, then Ethernet or USB designs, then motor control and digital power where the loop must be re-identified — GD32E5 against STM32G4 and XMC7000 — and leave automotive and security parts to last, one product line at a time.
| Step of the switch | What must be redone | Where it usually stops |
|---|---|---|
| General 8/16-bit, simple control | Peripheral drivers and factory firmware | Programmer and debug interface mismatch |
| Mid-range M3 and M4 | Register diffing, clocks, security package | It builds, but the timing is wrong |
| Ethernet or USB parts | Stack porting, whole-machine EMC | Emissions fail, one board spin |
| Motor control and digital power | Loop re-identification, ageing data | Prototype passes, production lots vary |
| Automotive and security parts | Qualification, defect liability, supply clauses | Terms cannot be agreed |
How to run the substitution
- Build the list by category, not by vendor; sort active parts into the five steps above and start at the bottom.
- Ask three questions per candidate: is there a real pin-compatibility claim, must key injection be built in-house, can the supply horizon be contracted.
- Budget verification hours explicitly; EMC and machine retests are counted per event, not per part.
- Move one step at a time — never port drivers and re-identify a control loop in the same change.
- Do not treat "domestic" as a performance conclusion; the database proves ordering and category position only.
- Do not swap automotive parts for a label; qualify a second source on body and lighting loads instead.
Data caveats
Share figures are the Omdia 2025 values stored for five international vendors, a top-five sum that excludes the long tail. Our database holds no share, shipment or price record for Chinese MCU vendors, so this article quotes none. Product-line positions and vendor wording come from curated ICNAV records and are not live pricing. Editorial judgement by the ICNAV team, not procurement or investment advice.
Written by ICNAV 编辑部 · ICNAV editorial team. References: the ICNAV company directory and public industry sources.