12 September 2023

Battle for Semiconductors: Will There Be Winners?

Leonid Kovachich

On the eve of U.S. Secretary of Commerce Gina Raimondo’s visit to China on August 27-29, the White House took some conciliatory steps. The U.S. removed 27 Chinese companies from the so-called Unverified list compiled by the Bureau of Industry and Security at the U.S. Department of Commerce. This list includes companies for which the agency cannot verify information on their transactions and whose exports from the U.S. are restricted in some way. While the Chinese Foreign Ministry certainly welcomed that move, the basis of U.S. technology policy toward China remains unchanged. The regional fragmentation of the semiconductor industry will only increase over time. However, the degree of supply chain interdependence and global division of labor in this area is so great that the creation of technological regional blocs under the influence of geopolitical considerations will inevitably lead to supply chain disruptions, multiplied costs, and possibly a slowdown in the growth of technological capabilities for all parties.

Key players

Semiconductors form the backbone of all modern electronics. They are used not only in computers and smartphones, but also in household appliances, cars, children’s toys, military equipment, etc. In other words, semiconductor circuits are indispensable in the production of almost any modern good containing electronic components. Historically, the semiconductor industry evolved in the United States in the mid-1950s. It was there that the first operable integrated circuit was invented and produced. Up until the mid-1980s, Silicon Valley in the U.S. state of California had retained its undisputed global leadership: the U.S. share in global semiconductor production exceeded 50%. However, gradually, under the sway of globalization and international division of labor, as well as with the growing technological complexity of the semiconductor circuitry, the production chain was lengthened to become dispersed across different countries. In the mid-1980s, Japan took over some of the key production processes in this area. Later, Taiwan secured a strong position in the mass final production of semiconductors. Finally, in the 2000s, the Netherlands became an absolute leader—and later a monopolist—in the production of advanced equipment required for lithography of semiconductor circuitry on a silicon wafer. As a result, the current production chain may involve thousands of suppliers scattered around the world, many of them being absolute monopolists on the market. For example, U.S.-based companies such as Cadence Design and Synopsis control 90% of the market for electronic design automation tools (EDA Tools), essential at the initial stage of microchip design. The Netherlands-based ASML is the world’s only supplier of equipment for ultra-deep ultraviolet (EUV) lithography on silicon slabs or wafers. Japan’s Tokyo Electron supplies state-of-the-art equipment for plasma etching, a necessary process for removing layers of material from the wafer surface after lithography. Finally, Taiwanese companies account for more than 50% of the entire global semiconductor end-market, as well as more than 90% of the market for advanced chips made in the 10nm process and below. Meanwhile, South Korean manufacturers control up to 64% of global production of dynamic random-access memory (DRAM) chips.

It is important to realize that China, too, plays a key role in the global semiconductor industry. First, the country is the world’s largest consumer of chips, as it dominates the global production of electronic products. Approximately one-third of all consumer electronics in the world are manufactured in China. In 2022, semiconductors worth $573.5 billion were produced globally, with China accounting for 53.7% of all sales of these products. It is natural that global chip makers have sought to localize production closer to their markets. Thus, both the largest Taiwanese contract manufacturer TSMC and South Korean SK Hynix and Samsung have their own production facilities in China. For example, Chinese plants produce up to 40% of the total volume of NAND chips (non-volatile memory chips) manufactured by Samsung and 40% to 50% of DRAM chips put out by SK Hynix. In addition, it was profitable for global manufacturers to locate less technologically advanced but more labor-intensive production stages in China. For example, China still accounts for more than a quarter of the global chip testing and packaging market. Intel, Texas Instruments, and many others have located their respective facilities in China. Finally, China is the largest producer and supplier of rare metals (gallium, germanium, etc.) required for semiconductor production.

From globalization to technological sovereignty

Therefore, the semiconductor industry has become one of the most globally dispersed production sectors. At present, no country can ensure the production of a microchip from start to finish by solely relying on its own resources and production base. Until a certain time, such a global mode of division of labor had suited everyone. Moreover, Washington had long winked at China’s development of its military-industrial complex due to interpenetration of civilian and military capacity. The leakage of U.S. technologies to China occasionally worried the United States, only in the context of these technologies being transferred to Iran, which at that time had been under sanctions for decades. That negligence had lasted until the mid-2010s, when China first published its Made in China 2025 import substitution program for key technologies, followed by Next Generation Artificial Intelligence Development Plan, which recognized the leading role of emerging technologies, including artificial intelligence (AI), in achieving global dominance and developing the military potential. It was then that the issue of China’s technological development and threats to U.S. national interests came to the fore in America’s political and expert community.

Strict export control measures against Chinese technology companies were first adopted in 2018. At that time, the U.S. accused telecommunications company ZTE of supplying Iran with products containing U.S. semiconductor technology in circumvention of the U.S. sanctions. The United States banned ZTE from purchasing chips created with American technology. This brought the company to the brink of bankruptcy, as there were no other alternatives for ZTE: as was discussed above, U.S. technology is used in the production of any modern chip, one way or another. The ZTE case was settled rather quickly after personal talks between Chinese President Xi Jinping and U.S. President Donald Trump. The company was ordered to pay a $1.3 billion fine, replace its top management, and introduce U.S. Compliance officers into the team. It is important to realize that the U.S. sanctions against ZTE had been imposed even before a full-scale trade war between the U.S. and China broke out. Yet, this was a turning point for both the U.S. and China.

The Americans realized that they had a powerful leverage of technological pressure in their hands. China, in turn, realized its own vulnerability. At that same time, in 2018, Keji Zhibao, a newspaper affiliated with China’s Ministry of Science and Technology, began publishing a series of articles reviewing Beijing’s vulnerabilities in key fundamental technologies. Chinese officials also began to speak more frequently about the need to ensure technological sovereignty.


The U.S. has become more active in using the technological leverage for putting more pressure. In 2019, the Trump administration put Huawei on the U.S. Department of Commerce’s blacklist; among other things, the sale of U.S. chips and other components was banned, as well as the use of Android OS for Huawei. However, this measure did not have a serious impact on Huawei’s business.

First, the Trump administration immediately introduced temporary export permits for Huawei so as not to create economic shocks for American companies, which in 2018 alone supplied Huawei with products $11 billion worth.

Second, nothing could deter Huawei from procuring critical components in third countries. Just a few months later, the company announced its own Harmony OS as an alternative to Android. By the end of the year, the company reported revenue growth of 18%. The following year, the Trump administration extended the sanctions imposed on Huawei. The company fell under the so-called Foreign Direct Product Rule, which prohibits the supply of equipment and components, including from third countries, if they contain American technology. At that time, Huawei was cut off from receiving advanced chips in any way, because contractors from third countries refused to cooperate with this company, fearing secondary U.S. sanctions. As a result, Huawei was forced to sell Honor, its smartphone division.

In the meantime, the sanctions imposed on China by the Trump administration were fragmented. A grace period was introduced for most export restrictions, which, in fact, deferred the enforcement of sanctions for a long period. In addition, Trump’s technological crackdown on China was a pinpoint strike. Huawei, whose name was already on the rumor mill, including among U.S. political figures, was hit hard. However, other Chinese technology companies continued to grow relatively unhindered. Sales of Chinese chip makers and developers rose 18% to $150 billion in 2021. China’s largest contractor SMIC reported sales growth. Although SMIC did not escape being blacklisted by the U.S. Department of Commerce, this did not prevent the company from mastering the production of chips in the 14-nm process. In addition, according to some media reports, SMIC was able to master the 7nm process via reverse engineering of a chip from TSMC. Chinese memory chip maker YMTC has caught up with its American and Korean competitors. The company has developed its own fourth-generation 3D NAND chip, consisting of 232 layers. Apple was even going to make YMTC the exclusive supplier of memory chips for the iPhone.

War of technologies

All these factors forced the U.S. to take a new look at the technology standoff, given that China’s progress in semiconductors has since been tied directly to U.S. national interests. In its technology policy, Washington began to focus on two fronts simultaneously: first, restricting China’s access to advanced technologies as much as possible, and second, stepping up government support for its own innovations and encouraging the repatriation of production capacity to the American soil. Washington understands that semiconductors are the basis both for the development of civilian technologies (and hence economic growth) and for the development of modern weapons systems, i.e., ensuring the interests of national defense.

In October 2022, the Biden administration imposed unprecedented export restrictions on China. Under the new rules, U.S. companies are prohibited from supplying China with high-performance chips and computer goods containing such chips (e.g., GPUs used to develop AI systems). In addition, exports of components that are used in the manufacture of supercomputers or for the development of semiconductor manufacturing have been banned. Supplies of certain equipment for chip production are prohibited. The Foreign Direct Product Rule applies to 28 Chinese companies (this list includes China’s all leading technology companies as well as specialized research institutes). Finally, third-country companies operating in China require special licenses from the U.S. Department of Commerce to supply logic chips with FinFET (fin-shaped field-effect transistors) architecture – 14nm and below; with DRAM – 18nm and below; NAND FLASH – with 128 layers and more – if such products are manufactured using U.S.-developed technology. In addition, professionals with U.S. citizenship and green cards are prohibited to perform certain work that directly or indirectly supports the development and production of semiconductors at certain facilities in China.

In August 2023, the U.S. released a draft of new measures aimed at restricting the flow of U.S. capital into China’s technology sector. If these measures take effect, U.S. private and venture capital investors will be prohibited from investing in Chinese companies that are involved in quantum computing, AI and advanced semiconductors. This being said, a complete ban on investments in the AI industry, as follows from the draft decree of the U.S. President, will apply only to Chinese companies that supply products to enterprises of the military-industrial complex. In other cases, U.S. investors will only need to notify the relevant U.S. regulatory authorities of their intention to invest in respective Chinese companies.


In parallel with prohibitive measures against China, the U.S. authorities are introducing incentives to develop America’s own competencies in the semiconductor industry and to build up the national production base. In 2022, the CHPIS and Science Act was passed, which envisages the allocation of $52 billion in government subsidies for the development of production within the United States. These subsidies will be available to all companies, including the ones of foreign jurisdiction, that decide to develop semiconductor production in the United States. An important condition for receiving support: potential recipients must commit not to invest more than $100,000 in China over a 10-year period if these investments result in the expansion of existing production capacity in China by 5% or more. It is also now prohibited to introduce new product lines or expand existing production with mature technologies by more than 10%. Companies that fail to meet these terms will have to repay the subsidies provided to them within 10 years.

Living under sanctions

U.S. technology restrictions would have had a very limited impact upon China unless key semiconductor technology suppliers from other countries had joined them. Therefore, considerable efforts of U.S. diplomacy were aimed at convincing its partners, mainly the Netherlands, South Korea and Japan, to join the technology restrictions. To a certain extent, the U.S. succeeded in doing so. On July 23, 2023, Japan announced export restrictions on 23 types of equipment needed for semiconductor manufacturing. Moreover, unlike the U.S. sanctions, the Japanese barriers apply to the equipment required for the production of chips using more mature technologies starting with the 45-nm process. Following Japan, the Netherlands also joined the export control measures: as early as 2019 it did not only ban supplies of the equipment for ultra-deep ultraviolet (EUV) lithography, but, starting in June 2023, also some machines for deep ultraviolet (DUV) lithography. Taken together, these restrictions should deny China the opportunity to rapidly develop its own semiconductor industry.

Export bans imposed by the U.S. and its allies are seriously hampering China’s technological development. China is being deprived of the necessary equipment to produce chips. Chinese companies have managed to establish mass production of chips in the 28nm process and are actively mastering the 14nm process. Of course, China cannot produce the most advanced chips, which are used, for example, in the latest generation of smartphones. Nonetheless, the bulk of consumer demand for semiconductors falls on the chips of previous generations. It is important, however, that China still produces these chips using foreign equipment. For example, China bought lithography equipment from ASML even in the 28nm process. The development of such equipment is surely underway in China, but a domestic lithography machine for the 28-nm process can only be expected by the end of this year at best. Moreover, Chinese companies do not have sufficient competencies to create automation design tools for the latest generation of electronics. Huawei this year said it has developed its own EDA tools to create chips in the 14nm process. However, the experimental software and hardware still need to be scaled up for mass production, as well as to ensure interoperability and compatibility in process setup.

Consequently, to produce its own chips, even using mature technologies of previous generations, China needs to build the entire supply chain of raw materials, hardware and software support. No other country at the current stage of technology development has been able to accomplish this incredibly complex and costly exploit. China is certainly ready to invest huge amounts of money in semiconductor technology development, but this does not guarantee success. China’s State Semiconductor Development Fund, or the so-called Big Fund, has accumulated more than $30 billion, but it has not been able to grow a single technology startup into a competitive semiconductor company. For example, Wuhan Hongxin Semiconductor Manufacturing Co, which received almost $20 billion, including from the fund, had gone bankrupt before it could launch any production.

Restrictions on chip imports to China also affect the development of related technologies and related industries. In the first half of 2023, China’s semiconductor imports fell by 22%, while imports of chip-making equipment fell by 23%. Inspur, a leading Chinese manufacturer of server hardware used for AI development, has already warned investors about difficulties with chip supply. The company forecasts a 30% drop in revenue as a result of U.S. semiconductor restrictions. Leading U.S. chip makers have responded to the export restrictions by developing chips specifically for China that are not subject to the export ban. NVIDIA, for example, released the A800 and H800 GPUs for China instead of the banned A100 and H100. Chinese companies have purchased $4 billion worth of these processors to be delivered in 2024. However, the development of new AI products, including generative AI, requires more processing power. According to various estimates, a complex model with as many parameters as ChatGPT requires about 30,000 of the most powerful A100 GPUs. No Chinese company currently boasts such computing power. While American tech giants such as Microsoft, Google and Amazon are freely investing billions in artificial intelligence platforms, Chinese companies are bound by both technological and investment constraints.

Nevertheless, containing China does not guarantee the successful evolution of the U.S. semiconductor industry. First, $52 billion in subsidies for all companies in the semiconductor sector is a very insignificant amount. For example, the construction of only the first phase of the TSMC plant in Arizona is estimated at $12 billion, while the entire project is expected to exceed $40 billion. In the meantime, the economic feasibility of building semiconductor plants in the US is questionable. The plant in Arizona, according to the project, will be able to produce up to 600 thousand chips per year by 2026. TSMC put out more than 14 million chips last year. And by the time the Arizona plant is expected to set up the 3nm process in 2026, such chips will have already been produced in Taiwan for two years. It is not known whether massive government subsidies will ensure U.S. technological leadership and independence from Asian partners. In addition, China as a key supplier of raw materials for the semiconductor industry also has serious leverage. For example, China has introduced export licenses for gallium and germanium. With China accounting for about 80% of the world’s total gallium exports and 60% of the world’s germanium exports, restrictions on the export of these metals could already lead to a significant increase in the costs of chip production and will subsequently reduce the growth potential for the entire industry.

Conclusion


The semiconductor industry is one of the most dispersed global industries. No single country currently possesses the full range of manufacturing chains required to manufacture finished semiconductor products. China, as the largest market for semiconductors and the source of raw materials essential for their production, plays an important role in global supply chains. The U.S. and China standoff, mounting export restrictions, and providing incentives for artificial relocation of production facilities will inevitably lead to the transformation of global production chains. Both the pace of development of Chinese capabilities in this area and the economic well-being of U.S. partners depend on the intensity of new export restrictions introduced by the United States. Given that, according to various estimates [1], semiconductor companies around the world are losing from 15% to 40% of their revenue from the existing export restrictions, an increase in U.S. sanctions pressure may lead to the degradation of innovation potential, including among the world’s industry leaders due to a sharp decline in their income levels. On the other hand, dependence on the Chinese market creates strong incentives for companies to seek ways to circumvent existing sanctions, so their fragmentation may limit the effectiveness of U.S. technology policy toward the PRC. In the long run, China will increase investment in basic research and development to ensure technological independence. The U.S. faces the challenge of balancing its technology policy to keep a hold on the existing gap with China in semiconductors for generations to come but, on the other hand, not to destroy key drivers of growing technological competencies for itself and its allies. Yet, as Chinese technological capabilities further evolve, it will become increasingly difficult for the U.S. to keep the right

No comments: