Read The Diplomat, Know The Asia-Pacific
SMIC’s 7nm process advancement – despite heavy U.S. sanctions – will have major implications for East Asia.
Semiconductor Manufacturing International Corporation (SMIC), the largest chipmaker in China, has reportedly achieved a major breakthrough. TechInsight, a Canadian tech media outlet, revealed that SMIC had advanced its technology to a quasi-7-nanometer (nm) process, which might be a stepping stone for a true 7nm process. According to TechInsight, SMIC products made from the quasi-7nm process had been shipped for a year. Some media argued that the SMIC’s advancement showed that the U.S. blockade was too little, too late, and out of date.
SMIC’s most advanced chip process node successfully made in the past was 14nm, although it has always made strong attempts to move toward an advanced process node (below 10nm). However, due to SMIC’s inclusion on the Entity List by the U.S. Bureau of Industry and Security in December 2020, which was designed to limit SMIC’s ability to reach advanced technology nodes of 10 nanometers or below, it has been blocked from obtaining the necessary Extreme Ultraviolet Lithography (EUV) machines from ASML of the Netherlands.
The use of an EUV machine is not necessary, in theory, to make the advanced process nodes. Taiwan Semiconductor Manufacturing Company (TSMC), the global leader in semiconductor manufacturing, used Deep Ultraviolet Lithography (DUV) machines in the early stage of its 7nm volume production. But using DUV machines requires more layers of masks, which means more times of exposure and more complexity. This will lead to a lower yield rate and a higher cost for each chip, making such a process commercially inviable nowadays.
But the semiconductor industry is of a strategic importance for China. Having the capacity to make advanced chips is more important than the prices of these chips. It appears that SMIC is indeed moving ahead to use this older technology to achieve technological breakthroughs. In October 2020, it was reported that SMIC had successfully developed “quasi-7nm” chips with the FinFET N+1 process using DUV machines.
Get briefed on the story of the week, and developing stories to watch across the Asia-Pacific.
TSMC’s chairman, Dr. Mark Liu, said that the 7nm process was a full node stride and a watershed in semiconductor manufacturing. The biggest difference between the 7nm and 14nm processes is that the number of transistors per unit area of the 7nm process increases greatly, and its energy consumption is reduced substantially. These makes 7nm chips far more powerful than 14nm one, yet also more economical. For example, in 2020, the cost of a 7nm chip was $233, which was not only lower than the $331 cost of a 16nm chip, but also lower than the $238 cost of a 5nm chip. In addition, the performance of NVIDIA’s A100 Tensor Core data-center processor, which uses TSMC’s 7nm process, increased by 20 times, so that the data center, which originally required 25 racks, can be reduced to a single rack.
Enjoying this article? Click here to subscribe for full access. Just $5 a month.
In other words, 7nm chips not only lower the cost of ownership, but also deliver high computing performance, which makes AI, cloud computing, and 5G economically viable both in business and military applications.
China’s Semiconductor Industry Sprints to Improve Self-Sufficiency
There is a big gap between chip consumption and chip manufacturing in China, meaning its chip self-sufficiency rate is low. In 2021, the size of China’s semiconductor market was about $186.5 billion, of which only $31.2 billion worth of chips were manufactured in China, both by foreign and domestic companies – a self-sufficiency rate of 16.7 percent. Furthermore, only $12.3 billion worth of chips were manufactured by China-headquartered companies, accounting for merely 6.6 percent of domestic consumption.
To reach the goal outlined by the “Made in China 2025” initiative, a self-sufficiency rate of 75 percent needs to be achieved by 2030. Under such pressure, it is not difficult to understand why China has subsidized semiconductor companies to build factories through various policy incentives. While there are notorious “unfinished fabs” cases in the development of the semiconductor industry, the failure has not caused China to retreat from its policy to fully support semiconductor factories.
The experience of developing the electronics industry in the past has made Chinese policy planners understand that, even though China’s semiconductor industry lags behind foreign manufacturers in terms of its production scale and technology, there are two effects that will urge China to support a large number of semiconductor companies through policies. First, a large number of Chinese manufacturers can “eat up” the market and compress the space for second- and third-tier wafer foundries. According to one report, by the end of 2024, China will be leading the world by building new 31 chip factories, surpassing 19 in Taiwan and 12 in the United States. Since most of the 31 new factories in China will be making mature processes nodes, there is little impact on the leading manufacturers, such as TSMC, Intel, and Samsung, all of which use advanced processes. However, China’s “fab sea” tactic may exert huge pressure on other mature process manufacturers.
Given that excess inventory has emerged in some areas of the electronics industry, and the market expects that there will be excess production capacity in chip manufacturing after 2023, price competition is inevitable. Foundries using mature processes will not be able to compete with Chinese semiconductor factories that enjoy major policy subsidies. Some second- and third-tier foundries may have to withdraw from the market, which will allow Chinese foundries to dominate the mature process market.
Second, if one or two Chinese companies can stand out among the large number of policy-supported foundries, there is a hope that this “national champion” can compete or even dominate the advanced process market. Lenovo in the PC/laptop sector and Huawei and ZTE in communications were all developed using such a model. And SMIC may be the leading Chinese company that can compete in the international advanced semiconductor arena and break technology strangleholds set by the United States. The 7nm advancement discovered by TechInsight is the best proof.
In the past, Chinese chip design companies were sanctioned by the United States and could not use TSMC’s advanced process to launch new products. If SMIC can extend the 7nm process to be used on other Chinese manufacturers’ products, it will allow China to accelerate its advancement in AI, high-speed computing, and 5G etc. The acceleration will enable China to achieve its goal of moving from a “manufacturing giant” to a “world manufacturing power.”
The Effectiveness of Entity List Needs to be Reassessed
We currently know very little about SMIC’s 7nm shipments, yield rates, and prices; it is not even clear whether there are other applications. However, the advancement of the 7nm process is expected to allow China to make breakthroughs in artificial intelligence and high-speed computing. In turn, that will also increase China’s economic and military threats to not only Taiwan, but all of East Asia.
Enjoying this article? Click here to subscribe for full access. Just $5 a month.
China has set goals to achieve “complete modernization” based on “informatization,” “intelligence,” and “mechanization” by the People’s Liberation Army’s 100th anniversary in 2027. Breakthroughs in AI, quantum computing, and hypersonics all require the assistance of advanced chips. Only blocking China from acquiring EUV machines will not prevent China from advancing in advanced process chips, which will eventually help its military buildup. More efforts are needed.
The United States is now trying to exert diplomatic pressure on Japan and the Netherlands to extend the current EUV embargo to include DUV machines. Some may argue that isolating China will only accelerate its march to self-sufficiency. From the past history of China’s industrialization, China’s ambition will not stop until the country dominates the entire market. Therefore, only limiting China from obtaining EUV machines will not suit the original purpose of keeping China from making advanced technology nodes of 10nm or below.
Che-Jen Wang is an assistant research fellow at the Institute for National Defense and Security Research (INDSR), Taiwan.