Semiconductor 2024: The emergence of domestic production of automotive chips, the listing boom of AI chips, and the huge market of memory chips | Year-end Review of Workplace Dividends ⑥
"Workplace Bonus" (ID: ZhiChangHongLi)
Whether it is the domestic substitution in the stock market or the new industries such as artificial intelligence in the incremental market, it means unlimited development opportunities for the semiconductor industry.
To build China's own semiconductor industrial system, not only chip R & D talents are needed, but also a large number of chip production process talents, sales and marketing talents, etc.
Data shows that at the chip R & D end, the most in-demand position is the R & D engineer, followed by the R & D manager and the R & D director. The former is the main force at the chip R & D end, and the latter two belong to the R & D management layer. Specifically, the Most in-demand R & D engineers include analog design engineers, RF design engineers, CPU architects, and signal integrity engineers - and they are precisely the positions with the fiercest competition, that is, both highly competitive and in short supply.
Although the industry is recovering, it cannot be directly linked to the recovery of the recruitment market. In terms of salary, in large domestic first-tier cities such as Beijing, Shanghai, and Guangzhou, the average annual salary of a chip R & D director (with more than 15 years of work experience) is about 1 million yuan, the average annual salary of an R & D manager (with more than 10 years of work experience) is about 700,000 yuan, and the average annual salary of an ordinary R & D engineer (with more than 3 years of work experience) is about 380,000 yuan.
Automotive Chip: The Localization Rate Is Expected to Increase to 15% ╱ 01
AI Chip: Domestic Large Factories Start to Purchase Domestic AI Chips ╱ 02
Memory Chip: It Is in an Important Period of Transformation ╱ 03
Automotive Chip: The Localization Rate Is Expected to Increase to 15%
According to CCTV News, on the morning of November 14, the annual output of new energy vehicles in China exceeded 10 million for the first time - China has become the first country in the world to reach an annual output of 10 million new energy vehicles.
Compared with traditional fuel vehicles, new energy vehicles are more dependent on electronic systems and require more chips to realize functions such as battery management, power control, and on-board charging.
New energy vehicles are of great significance in China, and their popularity has driven the vigorous development of automotive chips such as vehicle-grade SoC. According to data from Beroe Consulting, the market size of automotive chips in the global and Chinese markets in 2023 was 279.364 billion yuan and 65.818 billion yuan respectively, with China accounting for 23.56%.
"Workplace Bonus" (ID: ZhiChangHongLi)
Since the "shortage tide" of automotive chips, the domestic automotive industry chain and capital market have been strengthening the independent and controllable layout of chips, and a number of excellent domestic enterprises have emerged. In the MCU field, JoulWatt, NavInfo (AutoChips), ChinaCore Technology, GigaDevice, Hangshun Chip, BYD Semiconductor, ChipON, Chipsea Technology and other companies have emerged. Recently, the first fully domestic, independently controllable high-performance vehicle-grade MCU - Dongfeng Automobile DF30 - has officially made its debut. Some analyses suggest that the current localization rate of vehicle-grade MCUs has approached 10%.
In recent years, GalaxyCore, SmartSens, BYD Semiconductor and other domestic enterprises have also been continuously promoting the development of vehicle-grade CIS products, and their market competitiveness is gradually improving. It is worth noting that OmniVision under Will Semiconductor is one of the leading players in the global automotive field.
In addition to supply chain enterprises, automotive enterprises are also increasingly involved in the chip sector. BYD, Geely, Changan, Great Wall, Xiaomi, SAIC, Dongfeng, FAW, NIO, Ideal, XPeng, etc., either directly participate in chip manufacturing or participate in the ecological chain to assist in chip manufacturing, all are building their own chip supply security bottom line.
China sells more than 30 million vehicles annually, accounting for about one-third of global vehicle sales, but the local automotive chip supply only accounts for about 10%. With the joint efforts of the industrial chain ecology, more and more automotive chip achievements are accelerating to be implemented. The industry predicts that the localization rate of domestic automotive chips is expected to increase to 15% in 2024.
In addition, foreign media reported recently that China is urging domestic automakers to purchase up to 25% of chips locally by 2025, aiming to reduce reliance on imported chips, which will also help improve the competitiveness of domestic chips.
Almost all types of automotive chips have local enterprises promoting domestic solutions, and the localization trend in some fields is increasingly prominent. For example, in the field of intelligent driving chips, HiSilicon, Horizon Robotics, Black Sesame Technologies have become representatives of domestic solutions. Especially, the mass production of Horizon Robotics' Journey 6 chip is accelerating the squeezing of the market share of some international giants. However, some fields still need to be further cultivated, and the cockpit chip is considered one of the most difficult fields.
Vehicle-grade chips, especially high-end vehicle-grade chips, still restrict the development of China's automotive industry. Therefore, recruitment is essential. Taking the chip design company Hygon Information Technology as an example, the average salary of R & D personnel in 2023 can reach 854,400 yuan, and this figure can reach 917,500 yuan in Cambricon. In the 2024 campus recruitment competition, Hygon offers an extremely attractive salary, with the salary for fresh graduates with a master's degree in Digital IC Design directly set at 450,000 yuan.
AI Chip: Domestic Large Factories Start to Purchase Domestic AI Chips
This year, the stock price of "the first AI chip stock" Cambricon has been rising steadily, attracting much attention. In recent years, the call for the localization of AI chips has remained high, and start-up companies such as Enflame Technology, Moore Threads, Biren Technology have already become star targets in the capital market.
After the explosion of ChatGPT, the outside world realizes that whether it is processing large-scale data in the cloud or intelligent applications of edge devices, a powerful AI chip is needed to provide computing power support, such as in the fields of intelligent security, autonomous driving, and cloud computing, which all require high-performance AI chips.
The situation of China in the chip field is not optimistic. On December 2, the US Department of Commerce announced new export control measures on semiconductors to China, listing 140 Chinese enterprises in the Entity List.
Opportunities and challenges coexist. The next day, the China Internet Association, the China Association of Automobile Manufacturers, the China Semiconductor Industry Association, and the China Communications Enterprises Association jointly issued a statement - the relevant US chip products are no longer safe and reliable, and it is recommended that domestic enterprises carefully purchase US chips. This means that the opportunity for domestic chips has come. It is reported that while domestic supply is available, enterprises such as China Mobile and Baidu have begun to purchase domestic AI chips.
At present, the domestic enterprises focusing on AI chips mainly include Hanbo Semiconductor, Skylark Technology, Moore Threads, Maximojo Technology, AXERA Technology, Muxi Technology, HifiveTech, MulticoreWare, Skylark Technology, Starblaze Technology, etc. Among them, the most powerful model of domestic AI acceleration chips is Huawei Ascend 910B. This chip adopts a 7nm process and has 49.6 billion transistors. Its FP16 computing performance is about 320 Tflops, and its overall performance reaches the level of NVIDIA A100. In addition, the 910B AI chip launched by Cambricon, its FP16 dense computing power reaches 128 TFLOPS, exceeding NVIDIA A100; the Higon 2000 AI chip launched by Hygon Information Technology is the first domestic cloud inference chip based on a 5nm process.
In addition, from the perspective of cloud demand and enterprise actions, the research and development of the next-generation large model in 2024 will still be the core of competition among major technology companies, and the trend of the increasing proportion of multimodal models has driven greater computing power demand. Each chip manufacturer is fiercely competing in AI PC, AI Phone, and AI Car, comparable to the "Spring and Autumn and Warring States Periods".
"Workplace Bonus" (ID: ZhiChangHongLi)
The vehicle end may be the most confusing AI chip battlefield in 2024.
Workplace Bonus noticed on the recruitment website: A company is recruiting a Senior Expert in AI Chip Cooperation Planning, requiring the candidate to be based in Shanghai/Nanjing, with a salary range of 800,000 - 1.5 million yuan. Its main responsibility is to be responsible for the cooperation and docking of the externally purchased AI chips for the intelligent computing products; the construction of the indicator system, the selection and model testing, and the formulation of the relevant verification plans. In addition to having knowledge of computer architecture, this type of position also requires familiarity with the hardware architecture of the CPU/GPU or AI accelerator.
Memory Chip: It Is in an Important Period of Transformation
With the emergence of various large models, "AI Large Model +" is gradually being applied to diversified scenarios such as office, medical, industrial, government affairs, e-commerce, games, and streaming media. The continuous increase in data volume has given rise to a strong demand for computing power. At the same time, how to efficiently process and store large-scale data is currently two major challenges faced by computing power facilities. For this reason, the development of memory chips [1] is crucial for improving the performance and storage capacity of electronic devices.
The memory chip market has a huge space, currently accounting for 27% of the entire semiconductor market, with an overall market size of over 150 billion US dollars, and it is expected to continue to grow to 50% of the entire semiconductor market in the future.
In the field of memory chips, Chinese manufacturers have also risen rapidly in recent years and have occupied a place in the global competitive landscape. Yangtze Memory Technologies Co., Ltd. (YMTC), Hefei ChangXin Memory Technologies Co., Ltd. (CXMT) and other enterprises are catching up. Through huge investments and technological innovations, they are gradually narrowing the gap with international giants. RimeData Research Institute predicts that in 2024, the global market share of Yangtze Memory Technologies Co., Ltd. will be about 8%, and the global market share of Hefei ChangXin Memory Technologies Co., Ltd. will be about 10%, still having a gap with overseas leaders. Domestic enterprises are increasing their investment in technological research and development, especially in high-density storage technology, low-power consumption technology, and the integration of storage and computing, and have made important progress.
Since the end of last year, the prices of major memory chip products have stopped falling, and the market has begun to show signs of recovery. This is mainly due to the explosion of generative AI and the recovery of the consumer electronics market. In recent years, many memory chip companies have listed on the STAR Market of the A-share market, such as Puya Semiconductor (Shanghai) Co., Ltd., Dongxin Semiconductor Co., Ltd., Hengshuo Semiconductor Co., Ltd. and Innovative Memory Technology Co., Ltd..
Looking globally, the logic of the shortage of semiconductor talents is similar. Especially in China, the shortage of semiconductor talents is all-round, from Semiconductor materials, equipment, EDA tools, and IP in the upstream of the industrial chain to Chip design and manufacturing in the midstream and downstream, there are talent gaps.
However, if you want outstanding talents to join, the corresponding salary cannot be low. According to the latest data statistics and analysis of Zhiyouji, in the salary range of Yangtze Memory Technologies Co., Ltd., the proportion of 30 - 50K is the highest, reaching 36.1%. In 2023, the average salary of Yangtze Memory Technologies Co., Ltd. is 24.8K, which has decreased by 16 percentage points compared with 2022. From the perspective of academic qualifications, the average salary of undergraduate employees is about 33.6K, while the average salary of master's employees is about 35.2K.
In terms of regional differences, the average salary in Wuhan is 26.3K, while the average salary in Shanghai is as high as 39.4K. According to the latest information released on its official website, Yangtze Memory Technologies Co., Ltd. is currently conducting large-scale recruitment in Wuhan, Shanghai, Beijing, Shenzhen, Sichuan and other places, providing more than 70 positions, covering multiple positions such as Analog Circuit Engineer, Advanced Packaging Technology Development Engineer, Electronic Component Engineer, Industrial Engineer, etc.
"Workplace Bonus" (ID: ZhiChangHongLi)
According to the observation of Workplace Bonus, a typical recruitment of a Senior Expert in Chip Architecture in the Storage Field can usually offer a salary of 800,000 - 1.6 million yuan, requiring the ability to lead the specification definition, storage competitiveness analysis, and chip architecture design in the chip storage field according to the needs of cloud products. However, for senior expert-level positions, at least 10 years of relevant work experience for a bachelor's degree or 8 years of relevant work experience for a master's degree is required.
[1] Memory chips, also known as memories, are a common type of integrated