Overview of MCU process changes: development paths under the tide of AI application and cost reduction

1. Three major process echelons: from mature processes to advanced nodes

1. 40nm and above: absolutely mainstream in the current market

• It is still the most important process node for MCU in the world, adapting to traditional scenarios such as home appliances, basic industrial control, and general consumption.
• Core constraints:eFlash embedded flash memoryEconomic process nodes are only above 40nm
• Representative products:
  • ST：STM32F1/F4（65/90nm）、STM32H5/H7（40nm）
  • Renesas: RA6M4, RX66T, etc. (40nm)

2. 28nm: The main node of high-end MCU, the first choice for car companies and AIoT

• Driving factors: Automotive regulations, Internet of Things, and industrial automation require stronger performance and lower power consumption
• Break through bottlenecks:MRAM/RRAM/PCMand other new types of storage to replace eFlash and adapt to advanced technology
• Foundry focuses on expanding production, taking both cost and performance into consideration
• Representative products:
  • NXP：i.MX RT1170、MCX N、S32K
  • ST: Stellar P6 automotive MCU
  • Renesas: RH850/E2x, U2B, U2A series

3. 18nm / 16nm: cutting-edge technology for AI and automotive domain control

• Applicable scenarios: client-side AI inference, image and audio processing, automotive massive data stream processing
• Technical support: Embedded storage shrink technology is mature and supports further evolution of the process
• Representative products:
  • NXP: S32Z, S32E (TSMC 16nm FinFET)
  • Renesas: RL78/F, RL78/G23
  • ST：18nm FD-SOINew product line, scheduled for production in 2025, equipped with ePCM to significantly improve energy efficiency, density and RF performance

2. Changes in the industrial structure brought about by process upgrading

• Shipments are still at40/65/90nmMainly based on mature processes, complete ecology, short development cycle, and controllable costs
• 28nm/16nm/18nm products are growing rapidly, focusing onSmart cars, edge AI, high-end industry
• MCU original in-house production lines generally only reach 40nm, and advanced nodes are highly dependent on foundries such as TSMC, UMC, and SMIC.
• ST’s 18nm FD-SOI process is manufactured byCooperate with Samsungdevelop

3. Localization and cost reduction are key trends

• satisfySupply chain security and localizationpolicy requirements
• Close to China’s new energy vehicle market,Shorten response time and accelerate delivery
• In response to the rise of local MCU manufacturers,Reduce costs and improve competitiveness