Is the United States’ investment in chip manufacturing effective? 13 images dive deep into the global semiconductor supply chain.

The initial commercial results of developing a homegrown advanced packaging ecosystem in the United States are emerging.

Author | ZER0

Advertisement

Edit | desert shadow

The U.S. Semiconductor Industry Association and the Boston Consulting Group jointly released a 38-page global chip industry report “The Resilience of the Semiconductor Supply Chain is Emerging.”

▲ Government incentives by major regions (arranged by GOP size from left to right)

According to the report, the semiconductor industry has become susceptible to geographical concentration, with at least 50 points throughout the supply chain where one region accounts for more than 65% of the global market share.

Advertisement

The report estimates that private sector investment in wafer manufacturing will be approximately US$2.3 trillion from 2024 to 2032, compared with US$720 billion in the 10 years before the US Chip Act was enacted (2013-2022). The United States is expected to account for 28% of these capital expenditures, compared with a pre-CHIP Act investment pace that would account for only 9% of global capital expenditures.

Without the CHIP Act, the U.S. share is expected to drop to 8% of global capacity by 2032.

The report also predicts that by 2032, top wafer manufacturing capabilities will expand from Taiwan and South Korea to the United States, Europe and Japan; between 2022 and 2032, the United States’ wafer fab capacity will increase by 203%, becoming the country with the largest growth rate in the world; By 2032, the United States will reverse a decades-long downward trend and increase its share of total global fab capacity to 14% from the current 10%.

This report provides an updated view of the impact of current policies on future investment in the global semiconductor supply chain and its impact on resilience. It broadly defines resilience as an improvement in the geographical diversification of the supply chain and analyzes and predicts the impact of related industry trends on the next ten years. The likely impact in 2020, forecasting changes in wafer manufacturing distribution and ATP capacity, also takes into account geographic diversification in other parts of the supply chain, including chip design, core IP, EDA, equipment and materials.

The global semiconductor supply chain is highly specialized, with different regions having advantages in different fields.

For example, companies headquartered in the United States are leading in chip design, core IP, and EDA; American, European, and Japanese companies are leading in the equipment field; mainland China, Japan, Taiwan, and South Korea are leading in semiconductor materials; South Korean and Taiwanese companies are leading in the field of equipment. Leading in advanced chip manufacturing; assembly, test and packaging (ATP) mainly concentrated in China and Taiwan.

The globally integrated nature of supply chains enables regional specialization, giving each specialized firm access to global markets. However, geographical concentration also creates vulnerability, and it is expected that there will be significant geographical diversification in the future, mainly starting from the two fields of wafer manufacturing and packaging and testing. The report believes that due to cost pressures, packaging, testing and assembly companies are unlikely to be headquartered in the United States unless it is in some advanced packaging facilities near new wafer fabs.

Taiwanese companies have announced plans to build seven new wafer fabs on the island. Taiwan's chip manufacturing leader TSMC has also cooperated with Sony, Denso, and Toyota to improve the manufacturing capacity of Japan's Kumamoto factory. Mainland China is making new fab investments in Shenzhen, Tianjin and Shanghai.

Japanese chip manufacturing startup Rapidus has established an advanced 2nm chip production line at its new factory in Hokkaido. South Korea announced a $47.1 billion plan to build 16 new fabs in a large chip cluster in Gyeonggi Province.

From 2020 to the end of 2023, 80 new semiconductor manufacturing projects were announced in the United States alone, expected to create 50,000 direct new jobs.

Europe has made significant investments in new capacity, with seven major fab investments announced since 2020. Much of the capacity is being built in eastern Germany, including Intel's investment in Magdeburg and TSMC's investment in a new factory in Dresden with leading European semiconductor manufacturers.

In southern France, GlobalFoundries has partnered with STMicroelectronics to invest $3.1 billion in a wafer fab in Crolay. Poland is also preparing to open a new Intel advanced packaging factory.

The report predicts significant investment flows between regions between now and 2032.

Investing heavily in cutting-edge technology allows a region to compete at the forefront of innovation, but will not be fully reflected in monthly wafer starts; investing in mature processes, on the other hand, allows a region to realize more money in the short term. and employment values, the risk is the creation of excess capacity in a segment where demand may be fixed or weakened.

Advanced Logic's investment model has become more dispersed globally, with Taiwanese and Korean companies investing significantly more in the United States, Europe and Japan.

Advanced logic production will be almost 100% distributed in South Korea and Taiwan by 2022, and more than 40% will be distributed outside these regions by 2032.

In 2022, the United States is not producing any advanced logic chips. By 2032, the United States will produce nearly 30% of logic chips using processes smaller than 10nm.

Europe and Japan will also produce about 12% of chips above 10nm when planned fabs come online.

In logic processes in the 10 to 22nm range, Japan will develop a 5% market share from scratch, while mainland China's share will increase from 6% to 19%. Logic greater than or equal to 28nm will remain well distributed, with small share changes in most areas.

Mainland China's share will increase the most, from 33% in 2022 to 37% in 2032.

Among other process technologies, dynamic random access memory (DRAM) will remain highly concentrated in South Korea, but the U.S. share will increase from 3% to 9%, a 3-fold increase.

The geographic concentration of NAND memory will increase. South Korea's market share is expected to rise from 30% to 42% by 2032, with Japan and South Korea combined accounting for approximately 75% of capacity.

Discrete, analog and optoelectronic chips (DAO) will maintain good distribution, with participation shares of 5% or higher in all major regions.

The net result of these specific, strategic, and targeted moves by companies is to increase the industry's resilience to a more “even” share of global capacity by region. The United States will increase its share of global production capacity from 10% to 14%.

The report predicts that production capacity in each major region will increase by more than 80% in the next ten years. Capacity growth in the United States, at 203%, will be faster than elsewhere and much faster than in the previous decade. In terms of thousands of wafers per month (300 mm equivalent), this means an increase from 1121 kWh in 2022 (thousands of wafers per month) to 3393 kWh in 2032 (a 203% increase).

Currently, there are more than 3,000 fabless design companies in mainland China, with annual revenue growing at double-digit rates. Mainland China's local chip design mainly focuses on consumer electronics, industrial control systems and smart device chips, but is less competitive in advanced CPUs, GPUs and FPGAs as well as corresponding high-end servers and computer power management.

U.S. leadership in EDA should not be taken for granted. Between 2018 and 2023, BGI Jiutian, a leading EDA software provider in mainland China, achieved a 6-fold revenue growth.

The $110 billion semiconductor equipment market covers more than 50 types of specialized equipment, but is highly concentrated in some areas. The three segments of lithography, deposition, material removal and cleaning account for 70% of the market share, and each segment is dominated by a few major suppliers.

One European company holds 87% of the photolithography market. In deposition, material removal, and cleaning, three companies (two in the U.S. and one in Japan) hold 70%-80% of the market.

The top 15 equipment suppliers have production facilities in a total of 17 countries. These measures also include establishing new training centers to increase the talent pool outside their regions.

Mainland China currently accounts for 20% of global equipment spending and 18% of global equipment imports. Export controls in the United States, Japan and the Netherlands have increased the urgency to develop domestic alternatives. According to reports, at least five Chinese manufacturers are engaged in mass production; small and medium-sized enterprises have created lithography demonstration equipment; Northern Huachuang and China Microelectronics have entered the etching market at larger nodes.

The $64 billion semiconductor materials market includes chemicals and materials used in both the front-end ($40 billion) and back-end (ATP, $24 billion) of the supply chain.

Silicon wafers and photoresists account for about half of the total front-end materials market ($19.5 billion), but other subcategories such as gases, wet chemicals, CMP slurries and sputter targets are also critical to various steps in the manufacturing process . Likewise, substrates and leadframes account for about half of the back-end market ($12.8 billion).

Most of the leading semiconductor materials companies are headquartered in Japan, the United States, and the European Union. Japan has at least three leading suppliers in multiple segments of the front-end and back-end materials markets.

A survey released in December 2023 by the U.S. Department of Commerce’s Bureau of Industry and Security found that industry respondents “expressed serious concerns about domestic sources of three materials: bare wafers, gases, and wet chemicals.”

The post-pandemic chip shortage has also highlighted procurement challenges related to packaging substrates, which connect chips to circuit boards. Furthermore, certain raw materials, including gallium, rare earths and many other critical minerals, still primarily come from a single region.

Some people recycle old equipment to obtain these critical materials, and the vast majority of mining and refining of these materials is done in mainland China.

Currently, the ATP market worth US$95 billion is mainly concentrated in Northeast Asia. South Korea has significant back-end capacity similar to existing fabs. China and Taiwan together account for nearly 60% of the world's ATP production capacity. Of the 36 ATP facilities announced since 2020, 25 are expected to be in China and Taiwan.

In the long term, with continued policy support and overseas investment, the report predicts that ATP production capacity will be transferred to other regions, including Latin America, Europe and less developed regions in Southeast Asia. Southeast Asia is already the center of ATP activity, accounting for about 20% of the global total.

In advanced packaging, which accounts for less than half of the ATP market, technological breakthroughs could open the door for high-cost regions such as the United States to play a greater role in ATP. A key innovation is the heterogeneous integration of chiplets.

The initial commercial results of the US Chip Act's development of the domestic advanced packaging ecosystem in the United States are emerging: Amkor announced that it will build a $2 billion factory in Peoria, Arizona, to package chips produced by TSMC for Apple; Intel will The factory in Rio Rancho, New Mexico, has invested US$3.5 billion in advanced packaging; SK Hynix plans to invest approximately US$4 billion to build an advanced packaging factory in Indiana, USA; Samsung also plans to build an advanced packaging factory in Texas .

One potential area of ​​concern is ≥28nm logic. The current fab construction trajectory has production capacity well in excess of future demand, with much of that capacity located in large fabs in mainland China. If the trajectory does not change, high utilization-driven fab economics may result in significant pressure to lower wafer prices, which may cause fabless companies to reconsider process technology selection decisions.

Over the next decade, semiconductors will play a key role in the global economy, from everyday products to the forefront of defense and artificial intelligence.

Few industries have supply chains and ecosystems that are so complex and globally intertwined. However, many factors, from geopolitical tensions and a more complex regulatory environment to labor shortages and rising costs, are underlining the need to diversify supply chains and invest in improving resilience.

Likewise, governments recognize the strategic importance of semiconductors and seek to reduce strategic dependence by attracting and incentivizing new local or adjacent investments. But flexibility is not the same as self-sufficiency, and the costs of self-sufficiency will be staggering.

Advertisement