The thoughts and new things of A-share MCU chip manufacturers: selling goods at a loss under the dilemma of low-end internal involvement, overseas giants cutting into 32 or setting off the industry reshuffle, and combining automotive-grade and AI products into a breakthrough

Cai Lian News Agency, April 9 (Editor Yu Qi) This year's artificial intelligence set off a new wave, driven by computing power semiconductor chip prosperity significantly improved, MCU after last year's sharp decline in demand after the bottom reversal of the call is getting stronger and stronger. In terms of A-share listed companies, GigaDevice and Espressif's share price increased by more than 40% during the year.

It is reported that MCU, or micro-control unit, is to appropriately reduce the frequency and specifications of the central processing unit (CPU), and integrate peripheral interfaces such as memory, USB, and even LCD driver circuits on a single chip to form a chip-level computer. Market analysis said that with the further improvement of MCU computing power, the main frequency of high-frequency MCU has been raised to the GHz level, which can already meet the needs of low-computing power artificial intelligence at the edge end. At the same time, integrating artificial intelligence on MCUs and using only one chip to achieve end-side deployment is becoming a new trend.

Zheshang Securities recently said that the "hero behind ChatGPT" is supported by computing power such as GPU or CPU + FPGA, which will bring an increase in demand for high-end chips and drive the average price of chips. Cinda Securities also said that AIGC promotes the switch of AI industrialization from software to hardware, and the semiconductor + AI ecosystem is gradually clear. According to IC Insights data and forecasts, the global MCU market will reach 27.2 billion US dollars in 2026, with shipments of about 35.8 billion.

At present, the MCU market is relatively concentrated in the industry, and overseas leaders still occupy an absolute leading position. According to Omdia statistics, the top 5 MCU manufacturers in the world in 2021 are NXP, Renesas Electronics, STMicroelectronics, Infineon, and Microchip Technologies. Domestic leading manufacturers include GigaDevice Innovation, Zhongying Electronics, Espressif Technology, Fudan Microelectronics, National Technology, etc.

32-bit MCUs have become mainstream and continue to cannibalize the 8-bit market Intense "involution" in the mind: low-priced strategic products and even losing money The next few years may be the last opportunity for domestic MCUs to go to the next level

According to the bus or data processing bits, MCU can be divided into 4 bits, 8 bits, 16 bits, 32 bits or even 64 bits, that is, the number of bits of binary digits processed by the CPU each time, the more bits there are, the more significant data numbers, the higher the accuracy, and the stronger the computing power.

Since the beginning of this year, the two MCU giants STMicroelectronics and Texas Instruments have successively brought new 32-bit new products. According to STMicroelectronics, the digital trend will lead to a 3-4 times or even larger growth in the 32-bit MCU market. Huachuang Securities Research Report said that as electronic products move towards intelligence and networking, driven by high-end demand such as edge intelligence, 32-bit MCUs have become mainstream and continue to encroach on the 8-bit market.

The data shows that in the mainland general-purpose MCU market size, 32-bit and 8-bit occupy the mainstream of the market. According to the Northeast Securities Research Report, listed companies such as Nasda, GigaDevice Innovation, Fudan Microelectric, Zhongying Electronics, Espressif Technology, National Technology, Xinhai Technology, and AMEC Semiconductor have 32 MCUs. However, on the whole, the domestic 8th market still accounts for nearly half.

On the other hand, Guojin Securities Research Report pointed out that most domestic manufacturers are still actually concentrated in low-end fields such as small household appliances and consumer electronics. According to industry insiders, domestic MCU manufacturers have adopted a "low-price strategy" to seize the market under the fierce "involution", a listed manufacturer's M0+ level MCU pricing does not make money at all, or even loses money, while other MCU manufacturers can only follow up in order to produce, which makes the entire industry move towards a "vicious circle" and it is difficult to get rid of the dilemma of low-end competition.

Under the current MCU competition pattern in China, some insiders believe that the next few years may be the last opportunity for domestic MCUs to go to a higher level. At the same time, the two MCU manufacturers entered the 32nd place, or will cause a big reshuffle in the domestic industry pattern.

New things brought about by the transformation of new energy vehicles and AI industry: GigaDevice and other domestic manufacturers launch automotive MCU products to gradually break through National Technology and Nasda actively explore the combination of MCU and artificial intelligence

However, some industry analysts said that under the background of vehicle electrification, intelligence and networking, the value chain of automotive products will be reshaped, and automotive-grade MCUs may become the key areas of domestic manufacturers in the future.

At present, China has the fastest growing and largest new energy vehicle market in the world, and the production and sales of new energy vehicles have remained the first in the world for eight consecutive years. According to the research report of Huachuang Securities, automotive MCUs are mainly 32-bit chips. Among them, the increase in the application of battery management systems and vehicle controllers for new energy vehicles will drive the growth of MCU usage. For example, BYD's fuel vehicle F3 is equipped with 12 MCUs, while the number of MCUs in its electric model Tang has increased to 55, and the usage has increased significantly. At the same time, industry analysts said that whether from the short-term core shortage or long-term localization, the acceptance of local MCU companies by local car manufacturers is becoming high, and the upstream and downstream of the industrial chain are increasing support for domestic MCUs.

According to the Prospective Industry Research Institute, the global automotive MCU market will be worth $10.2 billion in 2025. The MCU market size of the automotive electronics industry in mainland China is expected to be 5.4 billion yuan in 2025.

At present, although international manufacturers occupy a leading position in the field of automotive-grade semiconductors, domestic manufacturers such as GigaDevice have gradually launched automotive-grade MCU products, deeply deployed in automotive applications, and begun to gradually break through with points, strips, and surfaces, and the era of MCU localization 2.0 is still kicking off. According to the research report of Huachuang Securities, GigaDevice Innovation, Fudan Microelectric, Xinhai Technology, Zhongying Electronics, Guoxin Technology, Jifa Technology, Xinwang Micro, BYD Semiconductor and other manufacturers are all working on automotive-grade MCU products and have successively passed AEC-Q100 certification, leading manufacturers have mass-produced or will soon launch automotive-grade MCU products, domestic manufacturers are expected to gradually make breakthroughs.

In addition, some market participants believe that in the new round of artificial intelligence transformation, MCU chips are still long-lasting, including Renesas and many MCU manufacturers are actively exploring the combination of MCU and artificial intelligence, and MCU chips will also make a big difference in the field of AI in the future. National Technology recently revealed on the interactive platform that the Cortex-M7 core MCU can support AI applications in machine learning with high computing power, and will be reasonably laid out according to market demand, R&D plans and technical conditions in the future. Nasda replied on the interactive platform that it currently has a dedicated computing self-developed platform for deep learning applications, which is responsible for lightweight deep learning model inference acceleration for MCU chips.