5-7 Nm Possibly The Long-Term Technological Ceiling For China’s Semiconductor Industry – Analysis

By Zhou Chao

Since the beginning of Donald Trump’s first term, the U.S. government has continuously ramped up pressure on China’s advanced technology sector, with the chip industry becoming a major target. Recently, before leaving office, President Joe Biden intensified sanctions and technological blockades against China’s semiconductor sector. The U.S. government has explicitly prohibited TSMC and Samsung from exporting chips below 7nm to China and has added 140 Chinese chip companies to its blacklist. From mid to late December 2023, the focus shifted to China’s mature-process chip industry. In response to the persistent pressure from the U.S., China’s chip industry has actively worked to counter external challenges, and eventually achieved a series of breakthroughs.

In terms of advanced processes, Huawei’s Mate60, launched in 2023, was confirmed to utilize 7 nm process technology, marking a significant breakthrough in Chinese-produced high-end chips. Regarding mature processes, data from international consulting firm Knometa Research showed that by 2022, there were 167 12-inch wafer fabs worldwide, with 13 new fabs coming online in 2023, bringing the total to 180. In terms of quantity, nearly half of the world’s 12-inch wafer fabs are located in China. Due to its complete industrial chain, robust production capacity, and cost advantages, China holds a clear comparative advantage in the production of mature-process chips.

In late November 2023, European chip giant STMicroelectronics announced during its Investor Day event that it will cooperate with China’s second-largest wafer foundry, Hua Hong Semiconductor, to produce 40 nm microcontroller units (MCUs) in China by the end of 2025. Recently, U.S. Secretary of Commerce Gina Raimondo publicly stated that a series of measures aimed at restricting China’s access to advanced semiconductor technologies have been a “fool’s errand”. It seems that, whether in advanced or standard processes, China’s semiconductor industry has made significant achievements and progress. However, ANBOUND’s founder Kung Chan recently pointed out that while China’s chip industry is already quite powerful, in terms of advanced processes, 5-7 nm may represent the long-term technological ceiling, with breakthroughs being extremely difficult.

Firstly, the technological bottleneck for 7 nm and below has already emerged. Huawei’s latest smartphone was once rumored to use 5 nm chip technology. However, a recent teardown analysis by the authoritative research firm Tech Insights (TI) of the Kirin chip in the Mate 70 series revealed that the previously rumored Chinese-made 5 nm chip does not exist. The Mate 70 series is equipped with the Kirin 9020, which is an upgraded version of the previous-generation Kirin 9010. The chip’s manufacturing process is the same as its predecessor, still at the 7 nm level. Compared to the Kirin 9010, the main improvement in the Kirin 9020 is a slight increase in the frequency of the large cores, optimizations in the middle cores, and a reconfiguration of the small cores. Compared to previous generations, the Kirin 9020 can only be considered an incremental upgrade, not a major design overhaul.

Since the release of the Mate 60, there have been numerous rumors about Chinese chip technology reaching the 7 nm level and making progress toward 5 nm. However, in the case of 7 nm technology, existing analyses have pointed out that due to the lack of extreme ultraviolet (EUV) lithography machines, production can only be achieved through multi-patterning techniques using the n+2 or n+3 process to achieve higher transistor density. Industry analysts have noted that if 5 nm chips are produced using deep ultraviolet (DUV) lithography and multi-patterning techniques, the overall manufacturing cost will be 50% higher than that of EUV-based processes, and the yield rate remains an uncontrollable factor. Information from the Financial Times shows that China’s 7 nm and 5 nm process technologies are 40-50% more expensive than TSMC’s, but the yield rate is less than one-third of TSMC’s. This level of technological cost is considered a relatively risky approach in the wafer manufacturing industry. All in all, China has encountered a clear technological bottleneck in chip technologies at 7 nm and below.

Secondly, can Chinese companies rely on their own capabilities for R&D and product commercialization to effectively break through this bottleneck? The answer is unlikely. Semiconductors are among the most internationalized and hierarchical manufactured products. They are fundamentally based on technology, cost, and efficiency, with manufacturing realized through the global division of labor. This global division can be seen through the three key stages of semiconductor fabrication. In fact, the world has established a complete supply chain for semiconductor production. The first step involves the U.S. providing EDA design tools and Japan supplying semiconductor materials. The second step is the design phase, where the U.S. provides the primary design, while other countries also contribute designs as needed. The third step is chip packaging in the U.S., mainland China, South Korea, and Taiwan. The fourth step involves testing, done in South Korea, mainland China, and Taiwan. The fifth step is shipping the chips. A global industrial chain, along with the flow of talent and knowledge, is an essential prerequisite for the smooth operation and development of any country’s semiconductor industry.

Over the past two decades, China has seen a large number of students studying advanced semiconductor technologies in the West every year, gaining substantial technical input that has greatly contributed to the growth of China’s semiconductor sector. Data shows that in 2019 alone, approximately 3,000 Taiwanese semiconductor engineers were working in mainland China, accounting for nearly 10% of the 40,000 core engineers in the industry that year. A few high-level executives also joined mainland Chinese companies, with the most notable example being Semiconductor Manufacturing International Corporation (SMIC). After former TSMC executive Mong-Song Liang joined SMIC, he led a team of engineers to successfully develop SMIC’s 7 nm chip technology. However, in recent years, with the increasing Western sanctions led by the U.S., the influx of international talent and technology into China has become much more difficult, and the situation has changed significantly compared to the past. In fact, according to research from multiple media outlets, since 2021, many Taiwanese technicians have started to distance themselves from mainland Chinese companies.

China still faces significant import dependency on key semiconductor raw materials. For example, AlSiC alloys, which feature high thermal conductivity, a thermal expansion coefficient matching that of chips, low density, lightweight, high hardness, and strong bending resistance, are mainly used in communication components. However, no Chinese companies have achieved large-scale industrialization of this product. Another example is interface thermal interface materials, which are commonly used in IC packaging and electronic heat dissipation. In China, the production companies for interface thermal materials are small in scale and lack high-end products. For both of these product types, U.S. and Japanese companies are the primary sources of imports. Existing analyses have pointed out that in the three main stages of the semiconductor industry, i.e., design, manufacturing, and packaging/testing, Chinese companies’ main comparative advantage lies in the packaging and testing area. However, even within this domain, Chinese companies are highly dependent on imported raw materials. If Western sanctions continue to intensify, the challenges in this area are likely to increase.

In the mature process technology field, Chinese companies are likely to find it difficult to provide effective support for the development of advanced process technologies.

First, the achievements of Chinese companies in the mature process domain are still far from satisfactory. Data from research organizations such as Knometa Research, Yole Group, and IC Insights show that although the overall production capacity share of mature process chips in mainland China has rapidly increased in recent years, Taiwan still holds the largest share of capacity. Looking at the leading mature process chip manufacturers, European companies dominate, followed by the U.S. and Japan.

Take Texas Instruments (TI), a leading company in the automotive chip sector, as an example. As a dominant player in analog chips with annual revenues close to USD 20 billion, TI has substantial cash reserves to handle price wars. Despite the intense competition in the Chinese market, it still enjoys solid profits in the U.S. and European markets. In 2023, even with a decline in profits, TI’s net income still reached USD 6.5 billion. At the same time, it has strong cost-control capabilities to launch price wars. Although the process technology for analog chips no longer evolves toward smaller line widths, TI, operating under the integrated device manufacturer (IDM) model, which combines chip design, manufacturing, and packaging/testing, has continuously upgraded its chip production processes. To date, TI’s analog chip processes have evolved to the industry-leading LBC10. With each generation of advancement, TI achieves improvements in yield rates, reductions in chip area, and savings in raw material consumption, ultimately leading to lower costs. Moreover, TI has adopted a new strategy in China, employing the strategy of “universal price cuts to strike all competitors, and targeted pricing to contain major Chinese manufacturers”. For some products, TI’s pricing in the Chinese market is 50% lower than in international markets. Although over the past two years, due to chip shortages, many Chinese manufacturers have turned to domestic chip suppliers for procurement, the increasing competitive pressure from international giants suggests that the ability of Chinese companies to resist in the future may not look bright.

Second, within the industry, it is generally believed that profits from mature process technologies should support the R&D of advanced process technologies. However, on a global scale, in recent years, as consumer trends for general industrial and consumer goods have gradually weakened, the demand for mature process chips has effectively been declining. The main sources of profit for chip giants are increasingly concentrated in advanced process products and equipment. Recently, warning statements from companies like Micron, Broadcom, and Marvell have highlighted this trend. As for the domestic Chinese market, although a series of consumption stimulus policies, such as those boosting electric vehicle sales, have objectively driven up chip demand, the sustainability of these policy-induced effects remains in question. Additionally, due to intense internal competition, as of December 5, Wind Information’s statistics show that 14,648 chip-related companies in China have gone bankrupt. In terms of industry sectors, bankruptcies have been relatively high among companies in consumer electronics and trade, and the business outlook for companies in the analog chip sector is also not optimistic. Although this trend may enhance the market dominance of leading companies, the profitability and financing capabilities of domestic top-tier companies still struggle to effectively support the development of advanced process chips, especially under increasing isolation. The international macro environment also suggests that even if Chinese production advantages in mature process chips are further enhanced and international markets are opened, these markets may not necessarily provide truly ideal returns.

Third, it should also be noted that, in terms of advanced process chips, there is no objective and strong market demand domestically in China. In recent years, the construction of smart computing centers in the country has been booming. By the first half of 2024, its domestic computing power scale reached 246 EFLOPS, with the growth rate of smart computing power exceeding 65% year-on-year. More than 13,000 computing power application projects have been implemented in various sectors such as industry, education, healthcare, and energy. However, industry research has found that the average occupancy rate of these smart computing centers remains around 20%-30%, with some enterprise-level centers seeing occupancy rates as low as 10%. Due to the severe oversupply, the rental price for an NVIDIA H100 server (8 GPUs) has dropped from RMB 120,000 – RMB 180,000 per month at the beginning of the year to RMB 75,000 per month, a reduction of around 50%. A smart computing center with a thousand-card cluster based on NVIDIA H100 GPUs would have an initial investment of around RMB 350 million, with annual maintenance costs of about RMB 50 million. Yet, with an average occupancy rate of 20%, the revenue from renting out the center cannot even cover half of the normal operating costs. In other words, there is no strong demand for advanced process chips in the Chinese market, and the demand needed to support their effective operation is quite limited. Under such a situation, even if advanced process chips with independent intellectual property rights are developed, the market demand to support them is unlikely to be substantial. It is also important to note that the construction of intelligent computing centers in recent years has been supported by a series of local policies, such as financial subsidies, tax incentives, and land supply. With local government finances increasingly strained, it is unlikely that such strong policy measures will continue to be available to actively create demand.

Final analysis conclusion:

Despite years of sustained pressure from the West, China’s semiconductor industry has still made significant progress and gradually accumulated strong capabilities, especially in the mature process field. However, with changes in both the internal and external environment, as well as the evolving trends in the industry, such technological capabilities are increasingly limited when it comes to breakthroughs in advanced processes. It also faces difficulties in providing continuous financial and demand support. For the foreseeable future, 5-7 nm will essentially be the technological ceiling for China’s semiconductor industry.

• Zhou Chao is a Research Fellow for Geopolitical Strategy programme at ANBOUND, an independent think tank.