According to the information obtained from the Wisdom Financial APP, Taiwan Semiconductor (TSM.US) disclosed in the minutes of the 24Q3 conference that the company's business in the third quarter was supported by strong demand for 3nm and 5nm from powerful smartphones and AI-related sectors. Looking ahead to the fourth quarter, business is expected to continue to be supported by strong demand for advanced processes, driving an increase in the utilization rate of 3nm and 5nm, with expectations for server AI processor revenue to grow by more than three times, achieving mid-teen revenue growth. Taiwan Semiconductor stated that strong support and customer base will enable the company to continue to capture industry growth, with full-year revenue growth expected to be close to 30% in USD terms.
The company's management stated that revenue in 24Q3 increased by 12.8% QOQ in New Taiwan Dollars, supported by strong demand for 3nm and 5nm from powerful smartphones and AI-related sectors, with a GM growth of 4.6% QOQ reaching 57.8%, mainly due to higher utilization rates and cost optimization. Operating expenses were at 10.4%, with OPM increasing by 5% QOQ to reach 47.5%. Q3 EPS was 12.54 TWD, and ROE was 33.4%.
The company's dividend policy is sustainable and continuously growing, reaping the benefits of previous investments. The growth in free cash flow is a result of reaping past investments. Looking at the balance sheet, cash has achieved organic growth, maximizing shareholder value. The company will return free cash flow to shareholders. As for whether the next five years will be as good as 2021-2024, Taiwan Semiconductor stated that growth will be between 20-30% until 2023, with 2025 also expected to be a healthy year. There is currently no long-term CAGR update, but the company believes it is healthy.
Based on customer demand, some customers require flexibility and there is government subsidy support. Arizona has stable customer demand, with support from the US federal government and local government, planning to build 3 FABs, double the size of a typical fab. Construction is now underway for 4nm, which is a good year for us and our customers, highlighting our strong manufacturing capabilities. We expect the first fab to start mass production in early 2025, with the quality of the Arizona plant expected to be the same as that of the Taiwan plant. The second and third fabs will be based on customer needs for advanced processes, with mass production expected to start in 2028 and 2030. The Japanese factory also has government support, making good progress, with the specialty process factory already validated and mass production starting this quarter, with quality equal to that of the Taiwan plant. Construction of the second fab will begin in the fourth quarter of the second year, with the second fab being used by strategic customers for consumer, automotive, industrial, and HPC applications. The European factory is supported by the EU and the local government in Germany, with a ceremony held with strategic customers in August for automotive and industrial applications, with mass production expected to begin in 2027.
<h2>Q&amp;A</h2>
[Q] Regarding investment in AI, how do you view the sustainability of AI and how do you plan for CAPEX development? How do you view the cycle of AI?
We believe that the demand for AI is real. We continuously engage with customers, including hyperscaler customers, and collaborate with every AI innovator. We have the deepest and broadest knowledge in this area. Based on our experience, we use AI and ML in FAB and R&amp;D to enhance our efficiency, speed, and quality
A 1% efficiency improvement is worth $1 billion to us. We are not the only company benefiting from AI. Other companies using AI will also improve efficiency. We believe that many companies are using AI to enhance their efficiency, so we think it's real.
[Q] How do you judge the semiconductor cycle? Has it peaked?
We believe that the demand for AI is just beginning. One of our important customers said that the current demand is insane, just the beginning, and the future will shift from science fiction to reality. Apart from AI, the semiconductor industry is now stable and starting to recover.
[Q] How do you view CAPEX in the coming years?
We do not have specific CAPEX numbers to share for the next 25 years. We decide on CAPEX through a very disciplined process, with higher CAPEX always based on higher growth opportunities in the coming years. If the growth is good, we will invest. Next year looks to be a year of healthy growth, and we will update the numbers in January.
[Q] Regarding GM's guidance, overseas factory dilutes gross margin by 2 to 3 points, how do you see 2025?
It's still too early to discuss the details for 2025. Indeed, we expect the dilution from the 3nm ramp to gradually decrease next year. We will sell the value, and it will be a year of healthy growth, so we are positive. However, the overseas factory capacity ramp will dilute the gross margin, and there will also be the transition from N5 to N3, with N2 ramping in 2026, incurring costs during the ramp. The preparation costs for each advanced process are increasing. Electricity costs have also risen, up 14% in October, doubling over the past few years. We believe Taiwan has the highest electricity costs. This will impact at least one percentage point of gross margin. We have not hedged the exchange rate, and a 1% exchange rate change will result in a 40 basis point change in gross margin.
[Q] Looking at overseas capacity, are the margins very low for US and Japan factories?
We will have an update on depreciation in January next year. The profitability of overseas factories is indeed lower, mainly due to smaller scale and higher costs as they are just starting to ramp up next year. But we will gradually improve. In Arizona and Komamoto, we have more than one factory, with fab1 improving profitability and fab2 starting. Arizona also has fab3, so we believe there will be a 2 to 3 percentage point dilution in gross margin every year for the next three to five years.
[Q] With current gross margins at 57% to 58% without price adjustments, higher than some customers, how will the ability to negotiate with customers affect long-term gross margins? What about mature process price wars? Will the company engage in anti-monopoly practices?
&#34;Selling our value&#34; is an ongoing strategy, and our progress with customers is going well. We also view our suppliers as partners and help them develop instead of using negotiation power against them. Our strategy is to make customers successful as TSMC's customers. If customers do well, we do well too. It is incorrect to say that TSMC's gross margin is higher than customers; some AI chip suppliers will never reach our gross margin. As a capital-intensive industry, we need a high gross margin to survive and achieve healthy growth
For antitrust or unfair competition, we previously proposed a new fab model, including wafer manufacturing, packaging, testing, and mask manufacturing, which are becoming increasingly important and are part of our revenue. Our competitors, such as IDM manufacturers, also have their own packaging, mask, and testing capabilities. We believe that this new fab model is more suitable for us. Our market share in this market is only about 30%, not yet dominant enough to be subject to antitrust scrutiny.
[Q] Intel is now considering selling its manufacturing department. Does the company expect more demand from Intel? Will the company purchase some of Intel's factories? How do you view Samsung's IDM, will they outsource?
The company will not purchase any part of the factories. From a business perspective, we believe that the IDM companies in California are very good customers for us, and we continue to receive sizable orders from them. Whether there will be an increase is too specific, we will have to see in a few quarters.
[Q] Regarding long-term growth opportunities, the company's guidance for 2021 is a 15-20% CAGR until 2026. Will AI drive this opportunity? What about after 2026? How do you see the growth opportunities for the next five years?
Whether the next five years will be as good as 2021-2024, we have been growing at 20-30% until 2023, and 2025 was also a very healthy year for us. We have not updated our long-term CAGR yet, but we believe it is also very healthy.
[Q] Will the demand for AI accelerate the company's growth?
We hope so, but we do not have any long-term CAGR numbers to share today.
[Q] Regarding 2nm and A16, there is strong demand for HPC. How will the construction of 2nm capacity be? Will chiplets reduce the demand for 2nm? What will the transition to the A16 process be like?
Indeed, chiplets are a strategic choice for many HPC customers, but we have many customers interested in 2nm. They have been in contact with us, and we see a lot of demand, higher than we expected, compared to 3nm. A16 is very attractive for AI server chips. The demand is also very good. We are working hard to prepare the capacity for 2nm and A16.
[Q] Regarding long-term AI planning, how many years does the company need to build factories? How is the company preparing for demand after 2025? Do customers have guarantees for the company? How does the change in hyperscaler CAPEX derisk?
We have been in communication with many customers, almost everyone in the AI field collaborates with us, including these hyperscalers. Looking at the long-term market demand, we have some insights, and we work with them on a rolling basis. Our capacity construction is disciplined to prepare the appropriate scale to support our customers' needs and maximize shareholder value.
[Q] There are some energy challenges in Taiwan. How will the company address the construction of 2nm capacity in Hsinchu and Kaohsiung? Currently, hyperscalers are building nuclear power plants. Will the company use nuclear power?
We have indeed built many factories in Taiwan, requiring electricity, water, and land. We continue to cooperate with relevant departments, informing them of our needs and plans, and have obtained their assurance of support. How they prepare for electricity, such as nuclear power or green energy methods, we cannot share at this time. But we are certain that we can obtain sufficient electricity, water, and land.
[Q] Regarding non-AI demand, what do you think about the demand for PCs and smartphones in the next 25 years?
The growth of PCs and smartphones is still in the low single digits, but in terms of content, there are more chips with AI functions. The growth rate of silicon chip demand is higher than that of end devices because of their larger size. We expect both businesses to gradually grow as the application of AI becomes more prevalent.
[Q] In terms of advanced packaging, what is the revenue outlook for the next few years? Can it reach the company's average gross margin?
Advanced packaging will grow at a rate higher than the company's overall growth for the next five years, reaching a high single-digit percentage. The gross margin is approaching the average gross margin, but not yet.
[Q] Regarding free cash flow, will the company consider increasing cash gross profit? How does the company balance ongoing investments and shareholder interests?
Our dividend policy is sustainable and continuously growing, we are reaping the benefits of our investments before. The growth of free cash flow is the result of reaping the benefits of past investments. Looking at the balance sheet, we use cash to achieve organic growth, maximizing shareholder value. We will pay back free cash flow to shareholders.
[Q] In the Foundry 2.0 model compared to the traditional foundry model, what are the growth opportunities for each part?
In the Foundry 2.0 model, advanced processes and advanced packaging have higher growth potential, while mature processes and traditional packaging do not have such favorable conditions.
[Q] Can you provide an update on the annual semiconductor growth forecast? How about next year?
It is quite similar to what was mentioned last time, our growth rate is slightly better than before. But overall, it is similar to previous estimates for the entire industry. We believe it is still too early to predict the growth for the next 25 years, we will discuss it in the next quarter.
[Q] What are the latest capacity plans for Cowos production for this year and next year?
We have put in a lot of effort to expand Cowos production. Roughly speaking, our customer demand exceeds our supply, even if the production capacity doubles this year and continues to double next year, it will not be enough. We are still working hard to meet customer demand

Taiwan Semiconductor stated at the 24Q3 earnings conference that it expects annual revenue growth to be close to 30%. Revenue in the third quarter increased by 12.8%, mainly driven by demand from smartphones and AI, with the 3nm and 5nm processes operating at full capacity. The management pointed out that a strong customer base and continuous free cash flow will support maximizing shareholder value. The outlook for the next five years is optimistic, with plans to build three new FABs in Arizona and achieve mass production by early 2025

- TSMC's Q3 2024 earnings report highlights strong demand for 3nm and 5nm processes driven by smartphones and AI, with a projected revenue growth of nearly 30% for the year.  
- The company anticipates a significant increase in AI server processor demand, with Q3 revenue rising 12.8% quarter-over-quarter and gross margin improving to 57.8%.  
- TSMC plans to expand its manufacturing capacity in Arizona and Japan, supported by government initiatives, while maintaining a sustainable dividend policy and focusing on maximizing shareholder value.

Taiwan Semiconductor 24Q3 Legal Affairs Meeting Minutes: Expected annual revenue growth close to 30%