This conversation focused on TSMC's financial performance for the second quarter of 2025. It started with a summary of the quarter's operations by CFO Huang Wend. He also provided performance guidance for the third quarter. Then, Huang Wend and the Chairman and CEO shared key information about the company. The meeting also included a Q&A session, during which forward-looking statements and their potential risks and uncertainties were discussed.

In the second quarter, the company's revenue increased by 17.8% year-on-year, reaching $30.1 billion, benefiting from strong demand for 3 nanometer and 5 nanometer technologies despite adverse foreign exchange impacts. The gross margin was 58.6%, while the operating profit margin was 49.6%. Advanced process technologies (7 nanometers and below) accounted for 74% of wafer revenue. HPC platform revenue grew by 14%, representing 60% of total revenue. Revenue for the third quarter is expected to be between $31.8 billion and $33 billion, a year-on-year increase of 38%, with a gross margin forecasted to be between 55.5% and 57.5%, and an operating profit margin between 45.5% and 47.5%. Annual capital spending is expected to be maintained between $38 billion and $42 billion.

In the second quarter of 2025, TSMC's gross profit margin was 58.6%, slightly lower than the previous quarter, mainly due to adverse foreign exchange rates and margin dilution from overseas factories. The gross profit margin for the third quarter is expected to further decrease to 56.5%, with the main factors still being exchange rate issues and production costs of overseas factories (such as in Kumamoto, Japan and Arizona, USA). Despite these challenges, TSMC expects the long-term gross profit margin to reach 53% or higher and anticipates a 30% increase in annual revenue, benefiting from strong demand for 3D, 3nm, and 5nm technologies, especially in the AI and high-performance computing sectors. The company remains cautious about the potential impact of tariff policies and will continue to invest in technology leadership, manufacturing excellence, and customer trust to strengthen its market competitiveness.

TSMC, based on customer demand, value, flexibility, and government support, has decided on a global manufacturing layout, with a special emphasis on investment plans in the United States. The total investment plan is $16.5 billion for advanced semiconductor manufacturing in the US, including the establishment of six manufacturing plants, a packaging plant, and a major research and development center in Arizona. The first plant has already entered mass production, using N4 process technology equivalent to the Taiwan factory. The construction of the second plant, using 3-nanometer technology, has been completed and is accelerating production to meet customer demand. At the same time, construction has begun on the third plant using 2-nanometer and 16-nanometer technology, with plans to increase production to meet AI-related demand. TSMC expects to establish an advanced semiconductor manufacturing cluster in Arizona to meet the demand for smartphones, AI, and high-performance computing applications, with 30% of 2-nanometer and more advanced capacities located there. Additionally, plans are in place to establish two advanced packaging facilities and research centers to enhance the AI supply chain. TSMC will continue to play a key role in the US semiconductor industry, helping customers succeed.

TSMC is expanding its manufacturing capacity globally, including multiple projects in Japan, Europe, and Taiwan, to meet the global IC industry's demand for advanced process technology. In Japan, mass production for the latter part of 2024 has already begun, with plans to build more factories in Europe and Taiwan. At the same time, TSMC focused on the development of its advanced technologies such as N2, A16, and A14, which will provide significant improvements for high-performance and energy-efficient computing. Mass production is expected in the coming years, strengthening its leading position in the global semiconductor industry.

At the recent meeting, the discussion focused on the strong momentum of the demand for data centers and AI that TSMC is facing, as well as how the company is attempting to narrow the supply-demand gap. In addition, observations on the development of device AI were mentioned, as well as the reasons for the expected decline in revenue in the fourth quarter, which is mainly based on customer feedback, especially in the consumer market.

The discussion focused on the progress of Edge AI in customer product design, pointing out that although the number of devices is growing moderately, the project scale is continuously increasing. It is expected that there will be a significant growth in the next six months to a year. At the same time, regarding the business guidance for the fourth quarter, the company has adopted a more conservative approach, taking into account the impact of tariffs and other uncertainties. However, it emphasized that if there is an opportunity, they will leverage their technological leadership and excellent manufacturing capabilities.

In the discussion, the focus was on the significant impact of foreign exchange rate changes on gross profit margins, as well as the company's confidence in maintaining or increasing future gross profit margins. In addition, there was also a discussion on the shipment of AI accelerator chips to the Chinese market and whether this may prompt the company to adjust its long-term growth forecast.

The discussion focuses on the revenue contribution of TSMC's N2 process technology, especially compared to the N3 process, as well as revenue growth expectations for the next few years. The revenue contribution of the N2 process is expected to be higher than the N3 process, not only because of its more competitive pricing, but also because capacity improvements will drive faster revenue growth.

The discussion focused on the supply and demand situation of TSMC's N5 and N3 technology nodes in the next two years. Due to the large number of AI products still at the 4-nanometer technology node and transitioning to 3 nanometers, the N5 production capacity is extremely tight. TSMC is using its technological advantages to flexibly adjust the capacity between 7 nanometers, 5 nanometers, and 3 nanometers, even using N7 capacity to support N5, while also transitioning N5 to N3 to meet demand. The overall production capacity of its leading technology nodes is in a highly tight state, and TSMC is working hard to narrow the gap between demand and capacity.

The conversation revolves around how TSMC is adjusting wafer prices and utilizing AI technology to address the structural cost increase caused by exchange rates and high costs of overseas factories. TSMC points out that despite the unfavorable structural factors, these can be effectively offset through optimization of other factors, such as price adjustments and the application of AI technology. Specifically, the application of AI technology in operations and research and development has significantly increased productivity for the company, equivalent to billions of dollars in value.

In recent conversations, it was mentioned that TSMC plans to build multiple semiconductor fabs in Taiwan and overseas to meet the strong demand for 5-nanometer, 3-nanometer, and future 2-nanometer chips in global data centers. Observers pointed out that this scale of capacity expansion is unprecedented in TSMC's history, demonstrating exceptionally strong market demand. TSMC stated that despite the high demand, they are working hard to close the supply-demand gap and ensure that there is enough capacity to support the increasingly growing data center demand.

In the discussion, the focus is on the comparison of N2 and N3 technologies, especially in terms of their return on investment ratio and future development. N2 technology is highly regarded for its higher profitability, despite its high capital expenditure, its production performance is excellent. It was mentioned in the discussion that the development of N2 technology is progressing as planned, with an expected increase in production in the second half of this year, and revenue is expected to be achieved in the first half of next year. At the same time, it was also emphasized that N2 technology is superior to N3 technology in terms of profitability, and its development progress is in line with expectations.

Although the company has increased its sales guidance in 2025, it has kept its CapEx (capital expenditure) guidance unchanged, a decision that takes into consideration macroeconomic uncertainties. For the coming years, especially 2026 and 2027, the company expects CapEx to potentially rise to address ongoing business opportunities. Additionally, the company emphasizes the importance of growth in cloud AI and AI accelerators, actively constructing new facilities to increase the capacity for AI and CoWoS (Co-package Wafer-on-Substrate) to meet strong market demand.

The discussion focuses on the development of TSMC's AI chip technology, especially the issues of increasing chip size and energy consumption, and how the company is addressing these challenges through advanced packaging technology and node development strategies. In addition, it explores TSMC's strategies in mature node technology, including how to meet customer demand by developing specialty technologies such as CMOS sensors and high voltage technology, and how to avoid excess capacity issues in the industry.

During the discussion, TSMC assessed the market potential of humanoid robots in the semiconductor industry, especially their contributions and growth momentum in the AI hardware field. Humanoid robots are seen as the next frontier in AI hardware, and their application in the semiconductor market will involve significant computing and sensor requirements. The discussion also touched on the issue of capital intensity, particularly when investing in new nodes or structural mega trends, TSMC's capital intensity has reached or exceeded 40%. For the upcoming 2-nanometer technology, TSMC expects capital intensity to rise again to meet demand. Additionally, the discussion also involved the possibility of customer demand pull-in ahead of schedule, as well as TSMC's capacity and plans for the fourth quarter.

The conversation mainly revolves around the relationship between capital expenditure and future growth opportunities, as well as the application of AI technology in the semiconductor industry, particularly in TSMC's adoption and leading-edge technology nodes such as 16 nanometers. The discussion emphasizes the strong growth of AI demand and the importance of improving power efficiency for data center applications, while also mentioning the progress of TSMC technology in meeting these needs.

TSMC is accelerating the construction progress of its second factory in the United States in response to customer demand, a decision positively influenced by the US government's policy of increasing tax credits for investment. Additionally, TSMC's expansion plans in the US will not impact its investment plans in Japan and Germany, as investments in these regions are focused on different technologies and industries. Furthermore, the detailed allocation of TSMC's overseas capital expenditures and domestic capital expenditures has become a point of discussion.