The Costs of U.S.-China Semiconductor Decoupling

Justin Feng

Source Link

As Washington and Beijing move toward “partial” technological decoupling, semiconductor chips have emerged as a key national security priority. In recent years, the United States has increased export controls against Chinese technology firms, and both sides have pursued industrial policies to promote domestic semiconductor manufacturing. A long-term concern for investors and multinational firms will be whether ongoing developments eventually snowball into a full-fledged U.S.-China semiconductor decoupling. Based on their deep integration in- and complementary positions along the global semiconductor value chain, a complete chip decoupling would carry immense economic and innovative costs.

The Global Semiconductor Value Chain

Beginning in the late 1980s, integrated device manufacturers (IDMs) that independently design and manufacture chips (e.g., Intel) found it increasingly difficult to simultaneously deliver the capital intensity of manufacturing and R&D expenditures for design. This gave rise to the foundry model consisting of pure play foundries (e.g., TSMC) that focus solely on manufacturing and fabless chip companies (e.g., Nvidia) that specialize in design. Today, all types of chipmakers—IDMs, foundries, and fabless firms—rely on a highly specialized global value chain and open trade to move around various chip components.

The U.S. leads the world in most knowledge-intensive activities along the semiconductor value chain, such as chip design, electronic design automation (EDA), semiconductor manufacturing equipment (SME), and core IP (see Table 1). In 2019, half of the top 20 chip design companies and 4 of the top 5 EDA and core IP companies by revenue were headquartered in the United States. Meanwhile, 75 percent of the world’s semiconductor manufacturing is concentrated in Northeast Asia. With decades of industrial policy support, robust infrastructures, and highly skilled workforces, Taiwan and South Korea are particularly strong in advanced manufacturing and possess a combined 100 percent of the global fabrication capacity in 7- and 5- nanometer processing nodes. China and the United States, the world’s two largest chip consumer markets, only rank fourth and fifth, respectively, in wafer fabrication. In the relatively less skill- and capital-intensive back-end manufacturing activities (i.e., assembly, packaging and testing), China ranks first in global market share, followed by Taiwan and Malaysia.

Integrated supply chains, specialization, and cross-border collaboration have continuously spurred cost-efficiency and performance improvements, which in turn enables the development of information technology, digital services, and emerging technologies (e.g., electric vehicles, artificial intelligence, and quantum computing). As major participants with complementary specializations, Washington and Beijing have both benefited from global semiconductor value chain optimization. However, geopolitical tensions, Covid-19 demand shocks, and extreme weather events in Taiwan have exposed the fragility of global semiconductor value chains during the past two years.

Chip shortages stemming from supply chain disruptions severely stalled auto and computer production for much of 2021. Such disruptions combined with rising U.S.-China geopolitical tensions have led both sides to adopt initiatives—such as the CHIPS for America Act and Chinese 14th Five-Year Plan—to bolster domestic semiconductor manufacturing. Furthermore, the Trump and Biden administrations have tightened export restrictions, added China’s leading chip manufacturer SMIC and other technology firms to the Commerce Department’s Entity List, and blocked Chinese bids to acquire U.S.-listed strategic technology companies. Such actions further convinced Beijing of the need to pursue semiconductor self-sufficiency. To the extent that China’s indigenization efforts succeed, global value chains may see deeper fragmentation. From 2016 to 2020, technology-related foreign direct investment (FDI) between the U.S. and China declined by 96 percent while R&D spending on home-grown technological innovation surged.

Table 1. Main Segments of the Global Semiconductor Value Chain.

Design Wafer Fabrication Assembly, Testing and Packaging

Intensive Input(s) Knowledge Capital and Labor Labor

Value Added to End Product 29.8% 38.4% 9.6%

Primary Activities Schematic model and physical layout of chip designs are created in specialized labs. Nanometer scale integrated circuits from chip designs are printed onto silicon wafers. Silicon wafers are sliced into individual chips, packaged, tested, and sent to end use producers.

Foundry/Fabless Model Fabless Designers Foundries, or “Fabs” OSATs

Foundry/Fabless Model Firms Qualcomm (USA), Nvidia (USA), MediaTek (Taiwan), HiSilicon (China) TSMC (Taiwan), UMC (Taiwan), SMIC (China), Hua Hong (China), Global Foundries (USA) JCET (China), ASE (Taiwan), SPIL (Taiwan), Amkor (USA)

Integrated Device Manufacturers (IDMs) Samsung (South Korea), SK Hynix (South Korea), Intel (USA), Micron (USA), Texas Instruments (USA), NXP (Netherlands), Tsinghua Unigroup (China)

Market Share Leader USA Taiwan China

*Other segments not displayed include Research and Development (R&D), Electronic Design Automation (EDA), Core IP, Semiconductor Manufacturing Equipment (SME), and Raw Materials.

Source: Author’s compilation based on multiple sources.

The Costs of U.S.-China Semiconductor Decoupling

As semiconductor chips become increasingly central to national security, Washington and Beijing will have to consider policy implications for the private sector. Boston Consulting Group (BCG) estimates that U.S. companies would lose 18 percent of their global market share and 37 percent of their revenues—leading to the loss of 15,000 to 40,000 high skilled domestic jobs—if Washington pursued a hard technological decoupling and completely banned domestic semiconductor companies from selling to Chinese clients. While this scenario is unlikely to occur in the immediate term, many in Washington are calling for broader outbound investment restrictions, tougher semiconductor export controls, secondary sanctions, and research blockades. In contrast, the U.S. would only lose approximately 8 percent of global market share and 16 percent of revenues if Washington does not expand the existing Entity List export restrictions.

Another risk with attempting technological decoupling is that many U.S. allies are reluctant to cut off trade with China, as Treasury Secretary Janet Yellen has noted. For instance, Seoul reportedly found it not “fully acceptable” when Washington proposed a semiconductor industry alliance between the U.S., South Korea, Japan, and Taiwan in March 2022. If “friendshoring” the entire semiconductor value chain was difficult, “reshoring” would be economically unfeasible. A joint report by the Semiconductor Industry Association (SIA) and BCG estimates that fully self-sufficient local supply chains would require at least $1 trillion in upfront investment, incur $45 to $125 billion in incremental recurrent annual operational costs for the entire industry, and result in a 35 to 65 percent overall increase in chip prices.

The costs of decoupling would be even higher for China, whose producers are completely reliant on imports of electronic design automation (EDA) tools and semiconductor manufacturing equipment (SME) designed by foreign firms such as ASML and Applied Materials. In the wafer fabrication sector, China’s newest foundries typically produce at the 28-nanometer level and its national champion SMIC is currently making progress in 14 nanometer technology. Chinese domestic chipmakers have made tremendous advancements but remain around two generations behind global industry leaders (e.g., TSMC, Samsung, and Intel) who can all run high volume production at the 5-nanometer processing node. Furthermore, Beijing’s efforts to end its dependency on foreign advanced chips have been hampered by U.S. Commerce’s regular denial of advanced EDA and SME export licenses for Chinese firms on the Entity List. As a result of this considerable technology gap, Chinese automakers who developed in-house 10 nanometer automotive chips have been unable to find domestic foundries capable of manufacturing their advanced designs, forcing them to outsource production to foreign firms despite the government’s self-sufficiency goals.

Aside from economic costs, decoupling global semiconductor value chains would also lead to future innovation loss for the entire industry. Broad market access barriers and restrictions on technology transfers could threaten the beneficial feedback loops produced by 30 years of value chain optimization. From the industry perspective, the worst outcome would be a full semiconductor decoupling that leads Chinese chip firms to “design out” U.S. technology standards, creating two separate ecosystems that results in higher costs and innovation loss for all semiconductor end-users globally.

Conclusion

As the essential hardware that powers almost all modern electronics and emerging technologies, semiconductor chips have emerged as an important geopolitical resource in the U.S.-China strategic rivalry. Moving forward, U.S. national security will keep prioritizing domestic technological innovation, supply chain resilience, and slowing China’s semiconductor industry through export controls. From China’s perspective, relying on the United States and its allies for advanced chips and inputs poses a major national security risk. Thus, Beijing will continue to emphasize semiconductor self-sufficiency and indigenous production.

Following the global chip storage in 2021, semiconductors clearly become a staple national security issue for both countries. However, Washington and Beijing should remain mindful of economic costs as they craft new policies. U.S. and Chinese semiconductor firms are deeply integrated within a complex and highly interdependent global value chain which has enabled decades of industry advancements. Given the enormous economic and innovative costs of localization, a full decoupling of global semiconductor value chains would be highly impractical. Balancing national security interests with economic reality will be a key challenge for both sides. As a general rule, policymakers should implement specific and targeted policies rather than broad restrictions to preserve the benefits of global semiconductor value chains while safeguarding national interests.