According to the information provided by TSMC’s Corporate Presentation Committee, only the China facility is generating significant profits, while operations in Japan, the United States, and Germany are currently incurring losses.
The Arizona factory in the U.S. began construction in 2021 and entered mass production by late 2024, accumulating losses of NT$39.4 billion (approximately $1.2 billion) over four years. The Kumamoto factory in Japan started construction in 2022 and began mass production by late 2024, with losses of NT$7.933 billion (about $248 million) over three years. The Dresden factory in Germany broke ground in 2024, with mass production expected in 2027, and has so far incurred losses of NT$500 million (around $15.6 million).
The Nanjing factory in China began construction in 2016, started mass production in 2018, and became profitable from 2020, generating a total profit of NT$68.2 billion (approximately $2.131 billion) over the past three years. Loss figures prior to 2020 are not publicly disclosed.
Losses during the construction and early production phases of chip factories are normal due to significant capital investments in building facilities and purchasing equipment. Profitability typically requires achieving economies of scale post-mass production. The Nanjing plant took 2 years to build and 2 years to realize mass production before it started to make a profit.
However, TSMC’s other overseas factories have not met the “profitability timeline” set by the Nanjing plant. The U.S. factory took three years to reach mass production, one year longer than the Japan factory, indicating high construction costs in the U.S. and less developed infrastructure in Arizona compared to Kumamoto. In contrast, TSMC’s factories in Taiwan can achieve mass production within just one year.
For chip manufacturers, the most crucial factor is order expectations. Before building a factory, they must ensure that the order volume exceeds the mass production scale; otherwise, they are unlikely to invest in construction. Another Taiwanese chipmaker partnering with India’s Tata Group is only providing technology licensing—neither building the factory nor securing orders. This indicates that India’s chip demand is unstable, and chip manufacturers are concerned about long-term losses.
The U.S., China, Europe, and Japan offer relatively stable order sources, which is the primary reason TSMC is willing to establish factories in these regions. Government subsidies help reduce initial capital expenditures and expedite mass production. The Nanjing factory’s success is largely due to China being the world’s largest chip consumption market, with no language or labor culture barriers and relatively straightforward regulations, enabling significant cost reductions.
The U.S. factory faces the largest losses due to high costs across the board, including inadequate infrastructure, complex government regulations, low administrative efficiency, a shortage of construction workers, raw material scarcity, and significant labor culture differences compared to East Asia. Consequently, building the U.S. factory costs four times more than a Taiwan factory. Even with government subsidies, the U.S. has failed to deliver fully—out of the $6.6 billion promised during the Biden administration, TSMC received only $1.1 billion, and the Trump administration has no intention of settling the remaining $5.5 billion.
Japan’s construction challenges fall between those of China and the U.S., with communication being the primary issue due to language barriers. Chip manufacturing requires extensive technical communication, both internally and with clients, making talent critical. This is why TSMC sends Taiwanese employees to overseas sites during the initial stages. Although the labor cost of the chip factory accounts for less than 10%, the labor cost can not only calculate the salary; the training of local talents and the labor culture of grinding are the main costs.
The Japan factory’s experience can be applied to the Germany factory. It is reasonable to expect mass production in 2027 without issues, with operations likely smoother than in Japan. Germany, particularly Saxony, already has a semiconductor industry foundation and a more complete supply chain than Japan, facilitating efficient operational planning and execution.
To avoid the negative experience of the U.S. factory, Europe must simplify administrative processes and regulations. If TSMC’s Germany operations run smoothly, advanced process factories could be established in Europe more quickly.
From a broader perspective, the U.S. government has difficulty overcoming the cost problem, and rebuilding manufacturing comes at a steep price. Forcing TSMC to set up factories while imposing tariff barriers only increases chip procurement costs for U.S. companies and restricts their market access. Moreover, TSMC is currently overly reliant on the booming AI industry, which remains on the brink of a bubble collapse.
TSMC’s CEO estimates that 30% of future advanced processes will come from U.S. factories, but I believe this is overly optimistic. Given the Trump administration’s ambitions, at least 50% of TSMC’s advanced process chips will likely be produced in Arizona, with Taiwan’s supply chain relocated to the U.S. Only then can the U.S. ensure leadership in chips over China and Europe.
In the AI industry, the key players are the U.S., China, and Europe. Washington does not want TSMC to have the capacity to supply its competitors. China, meanwhile, is aggressively bypassing the current logic chip track, preparing alternatives from design to manufacturing, such as advancements in photonic chips.
It is reasonable to predict that within five years, China will achieve mass production and profitability in cutting-edge logic chips, with photonic chip technology reshaping the chip supply chain. The U.S., during this period, will struggle with high costs and shrinking markets.
The chip industry may split into “two systems,” presenting an opportunity for Europe. By maintaining a foothold in the current chip race and collaborating with Chinese technology to share markets, Europe can secure autonomy.
Given the high energy consumption of the chip industry and downstream sectors, Europe-China technical cooperation should also include new energy. Chip factories and AI data center operators must invest in power generation infrastructure, so supporting the chip industry and applied technologies can also promote energy independence. Over the next five years, U.S. new energy development will likely be constrained, creating opportunities for China-Europe collaboration in this field.
Due to expanding losses, TSMC’s U.S. factory is reportedly set to raise 4nm foundry prices by over 30%. Meanwhile, Huawei’s Ascend 910C chip, amid U.S. export restrictions on NVIDIA, is becoming the only option for Chinese AI data centers. In other words, TSMC’s price hikes exacerbate challenges for U.S. companies already strained by tariff wars while providing Chinese competitors with more breathing room and development opportunities.
The profitability of TSMC’s Nanjing factory demonstrates China’s ability to create a high-quality industrial development environment, offering a benchmark for Europe.
Positioned between China and the U.S., Europe can learn from many experiences and avoid detours in cutting-edge technology. The foundation of technological innovation lies in corporate profitability, and the future performance of TSMC’s Germany factory will indicate Europe’s prospects in new technology development.