Since U.S. President Joe Biden took office in January 2021, the semiconductor issue has taken a prominent place in the new government’s agenda. It not only relates to job creation and the “Build Back Better” plan, but it concerns the renaissance of U.S. semiconductor manufacturing leadership and, most of all, the future of U.S. leadership in the age of Asia. After Biden’s first two in-person summits – with Japanese Prime Minister Suga Yoshihide on April 16 and South Korean President Moon Jae-in on May 21 – the U.S. president reiterated the importance of semiconductor supply chain security and resilience. After the summits, the U.S. reiterated its intention to cooperate on semiconductor issues with Japan and South Korea. The long-term security and resilience of semiconductor supply chain is at risk, however, if that cooperation does not expand to other like-minded countries.
With the ongoing progress in China’s semiconductor market and support from the Chinese government, the United States’ current method of containing China’s technological progress in semiconductors is insufficient. In order to prevent China’s eventual dominance in the semiconductor industry, the U.S., Japan, Taiwan, and South Korea must form a “Semi Quad” to coordinate their policies and efforts.
There are two major obstacles to establishing a resilient local semiconductor supply chain that would be invulnerable to geopolitics: The manufacturing base is over-clustered in Asia, especially in Taiwan, and there is a manufacturing decline in the United States. These dangerous conditions impact the U.S. semiconductor supply and pose a national security concern. Therefore, the Biden government expects to spend $50 billion in subsidies to revitalize U.S. semiconductor leadership. Intel, which had pledged to invest $20 billion in two fabs, is going to play an important role in the U.S. semiconductor renaissance. It might be the right move in the short term, as subsidies will increase the incentive for building fabs, but, arguably, the long-term goal of Biden’s plan might not be attained for the following reasons.
First of all, the shift of semiconductor manufacturing facilities from the U.S. to Asia – first to Japan and then to Taiwan and South Korea – over the past decades was driven by competitive advantages, just like other industries.
The production of chips is extremely capital and skilled-labor intensive. The capital expenditure for a 3 nm fab is $19 billion, which will see depreciation in just five years. This requires a high profit margin as well as a huge volume of sales to compensate for the high depreciation costs incurred from the new investment for advanced process technology. However, Intel failed to grab a share of the fast-growing mobile device market over the past 15 years, while, in the meantime, the PC market has been shrinking. In addition, the failure of Intel in 7 nm chips put the company at a relative disadvantage in competing with TSMC and Samsung. Therefore, the cost of staying at the leading edge in advanced semiconductor manufacturing is affordable only to TSMC and Samsung, which have already planned to build such fabs.
Other advantages of Asian manufacturers include satellite suppliers clustering, better customer service, and human resources. These produce a more cost-efficient edge over their competitors in the United States. The unit cost of chip production, not subsidies, is the fundamental factor behind competitiveness, which leads to success in the long run. In addition, a pure-play foundry – one that produces chips for other companies, rather than design its own products – is not a high-profit business model. A recent report by OECD estimates that the average gross margin of the pure-play foundry in 2018 is around 20 percent. This is why, over the past decades, Intel has gradually outsourced production to Asian fabs in order to concentrate on higher margin products.
At present, the division of labor in the semiconductor industry is quite defined. The United States dominates the upstream market with a lot of IP and machines required to produce integrated circuits (ICs), while Japan leads in raw materials, especially special chemical materials. Taiwan and South Korea are more cost-efficient in production.
However, these countries are all facing China’s catch-up in technology. History has shown us that if the Chinese know how to manufacture a product or service, sooner or later, China will take over the market due to cheap labor, lack of regulatory compliance, government subsidies, low taxes and duties, and compulsory technology transfer or theft. Today, Chinese-dominated industries range from materials, such as steel and chemicals, to consumer products like clothes and smartphones. Even China’s online and social media companies, including Alibaba, Tencent, WeChat, and ByteDance (the owner of TikTok), can compete with the West.
China consumed 143.4 billion worth of chips in 2020, a 9 percent yearly increase, and only $22.7 billion worth (15.9 percent) were produced in China. China has set an ambitious target for domestically-made chips to account of 70 percent of total consumption by the year 2025 and poured a new round of investment into semiconductors. The Chinese government has also set aside $155 billion in its latest five-year plan in support of the IC industry. We can expect another wave of transfers, legal or not, in intellectual property and talent to China over the next few years.
While China’s current problem lies in its inability to manufacture high-end chips, its semiconductor companies shorten the learning curve very quickly through technology transfer, including, among other means, hiring experts from abroad and using compulsory technology transfer – or even theft. Chinese companies are poaching from all over the world all available technology and talent along the full spectrum of IC design and production, including related machines. For example, three recently founded Chinese companies, X-Epic, Shanghai Hejian Industrial Software, and Amedac, have hired veterans from Synopsys and Cadence Design Systems to develop Chinese-made electronic design automation (EDA) tools. Other semiconductor companies in Taiwan, Japan, and South Korea have been targeted by China’s hacking and talent poaching.
Therefore, it is clear that the U.S. government alone cannot stop China from advancing its semiconductor industry. Given the small portion of semiconductors made in the United States, the current U.S. policy neither limits China from the acquisition of high-sensitive chips and technology, nor outweighs Xi Jinping’s call for technology autonomy in semiconductor production. The governments of the U.S., Japan, Taiwan, and South Korea must coordinate efforts to respond to aggressive tactics from China.
Given that different semiconductor technologies are scattered among the U.S., Japan, Taiwan, and Korea, and the degree of technology protection among manufacturers may differ, it is difficult to fully prevent advanced semiconductor technology from entering China. If the Biden administration wants to fully contain the high-speed progress of semiconductor technology in China, further containment policies and multi-country coordination are needed. Unless the U.S. can form a “semi-Quad” alliance, a mechanism like Quad or Five Eyes, and coordinate policies with Japan, South Korea, and Taiwan, fair competition and resilient trade in semiconductors will be difficult to realize.