The Zhitong Finance App learned that according to the TSM.US (TSM.US) 24Q3 law meeting minutes, the company's business in the third quarter was supported by strong smartphone and AI-related demand for 3 and 5 nm. Looking at the fourth quarter, it is expected that the business will continue to be supported by strong demand for advanced processes, driving an increase in the operating rate of 3 and 5 Nm. It is expected that server AI processors will grow more than three times and reach mid-teen revenue growth. TSMC said that strong support and customer base will allow the company to continue to seize industry growth, and it is expected that annual revenue growth will be close to 30% according to the USD scale.
The company's management said that 24Q3 revenue increased by 12.8% in NTD, supported by powerful smartphones and AI-related demand at 3,5nm. GM grew 4.6 pct qoq to 57.8%, mainly due to higher operating rates and cost optimization. Operating expenses were 10.4%, and opm increased by 5 pct qoq to 47.5%. Q3 EPS is 12.54 TWD and ROE is 33.4%.
The company's dividend policy is sustainable and growing, and is reaping previous investments. The increase in free cash flow is the reason for reaping past investments. Looking at the balance sheet, cash has reached organic growth to maximize shareholder benefits. The company will pay free cash flow back to shareholders. As to whether the next five years will be as good as 21-24, TSMC said that the previous 23 years had an increase of 20 to 30%, and 25 years were also a very healthy year. There has been no long-term CAGR update yet, but the company thinks it is very healthy.
The company is based on customer needs. Some customers require flexibility, as well as government subsidy support. Arizona has stable customer needs, and with the support of the US federal government and local government, plans to build 3 FABs, double size of the usual ones. Now we are building 4nm. This is a great year for us and our customers, highlighting our strong manufacturing capabilities. We expect the first Fab to be mass-produced in early '25, and we expect the quality of the Arizona factory to be the same as that of the Taiwanese factory. The second and third plants are based on advanced customer processes and are expected to be mass-produced 28 or 30 years ago. The Japanese factory also has government support, and is now progressing very well. The specialty process factory has passed verification. Mass production will begin this quarter, and the quality is the same as that of the Taiwanese factory. Construction of the second fab began in the fourth quarter of the second year. The second fab was used by strategic customers for consumer, automotive, industrial, and HPC. The European factory is supported by local governments in the EU and Germany, and held a ceremony with strategic customers in August for automotive and industrial applications. Mass production is expected to begin in 2027.
Q&A
[Q] Regarding investment in AI, how do you view the sustainability of AI and how do you plan the development of CAPEX? What do you think of the AI cycle?
We judge that the demand for AI is real. We are constantly in contact with customers, including hyperscaler customers. Every AI innovator cooperates with us, and we know the deepest and broadest. Based on our experience, we are using AI and ML in FAB and R&D to improve our efficiency, speed, and quality. A 1% increase in efficiency is $1 billion for us. We're not the only company benefiting from AI. Other companies using AI will also improve efficiency. We believe many companies are using AI to improve their efficiency. So we think it's real.
【Q】How do you judge the semiconductor cycle? Has it peaked?
We think the need for AI is just the beginning. One of our key customers said that the demand now is Insane's; it's just the beginning, and the future will shift from sci-fi to reality. Other than AI, semiconductors are now stable and starting to recover.
【Q】What do you think of CAPEX in the next few years?
We don't have a specific number of CAPEX we can share in 25 years; we determine CAPEX through a very disciplined process. Higher CAPEX is always based on higher growth opportunities in the next few years. If growth is good, we'll invest. Looking forward to next year being a year of healthy growth, we will update the numbers in January.
[Q] Regarding GM's guidelines, overseas factories dilute the gross profit margin by 2 or 3 points. What do you think of it in '25?
It's still too early for the details of 25 years. Indeed, we expect the dilution of 3nm climbing to gradually decrease next year, we will sell the value, and it will be a year of healthy growth, so it is positive. However, when production capacity rises in overseas factories, gross margins will be diluted, and there will also be N5 to N3 movements. N2 will also have ramps in 26 years, and there will also be expenses during the ramp period. With every advanced process, preparation costs are increasing. Electricity costs have also increased, rising 14% in October, and have doubled over the past few years. We think the electricity bill in Taiwan is the highest. This affects at least one point gross profit margin. We haven't released a controlled exchange rate. If the exchange rate changes by 1%, there will be a 40 basis point change in gross margin.
[Q] In terms of overseas production capacity, do the US and Japanese factories have very low gross margins?
We'll have an update on depreciation in January next year. The profitability of overseas factories is indeed worse. The main thing is that they are smaller. Next year, they are just starting to climb, and the costs are higher. But we'll gradually improve. At Arizona and Komamoto, we have more than one factory, FAB1 is starting to improve profitability, and FAB2. Arizona also has FAB3, so we think the gross margin will be diluted by 2 or 3 points every year for the next three to five years.
[Q] There are currently no price adjustments, and the gross margin is 57% or 58%, which is higher than that of some customers. Regarding the future bargaining power of customers, what impact will it have on long-term gross margin? What about mature process pricing strategies? Will the company have antitrust?
Sell our value is an ongoing strategy. We are partners with our customers, and the process is progressing very well now. We are our suppliers and partners; we don't use bargaining power on them, we help them develop. Our suppliers and customers all work together, and our strategy is to make our customers successful as TSMC customers. If our customers are good, we are fine. It is wrong that TSMC's gross margin is now higher than that of its customers; we will never reach the gross margin of some AI chip suppliers. As a capital-intensive industry, we need high gross margins to survive and achieve healthy growth. Regarding anti-monopoly or unfair competition, we previously proposed a new FAB model, including wafer manufacturing, packaging, testing, and mask manufacturing. These are all becoming more and more important. These are all part of our revenue. Our competitors, such as IDM manufacturers, also have their own packages, masks, and tests. We think this new FAB model is more suitable for us. Our share in this market is only about 30%, and we have not yet reached dominance, and not enough to fight against monopoly.
[Q] Intel now wants to sell its manufacturing division. Does the company expect more demand from Intel? Will the company buy some of Intel's factories? What do you think of Samsung's IDM and will it be outsourced?
It won't be possible to buy some factories. From a business perspective, we think IDM in California is a good customer, and we continue to get sizable orders from them. The question of whether it will increase is too specific. Let's take a look at it in a few quarters.
[Q] Regarding long-term growth opportunities, the company guided a CAGR of 15 to 20% in '21. Will AI grow this opportunity? What about after 26 years? How about looking at growth opportunities in the next five years?
Will the next five years be as good as 21-24? We had 20-30% growth before 23 years, and 25 years were also a very healthy year for us. We haven't had a long-term CAGR update, but we think it's also very healthy.
[Q] Will AI demand make the company grow faster?
We hope so, but we don't have a long-term CAGR figure to share today.
[Q] Regarding 2nm and A16, HPC demand is strong. How will 2nm production capacity be built? Will chiplet reduce 2nm demand? What will the migration to the A16 process be like?
Indeed, Chiplet is indeed a strategy for many HPC customers, but many of our customers are interested in 2nm. They have all come into contact with us, and we have seen a lot of demand, which is higher than our previous expectations, and higher than 3nm. The A16 is very attractive for AI server chips. The demand is also very good. We are working hard to prepare 2nm and A16 production capacity.
[Q] Regarding long-term AI planning, the company will take several years to build a factory. How can the company prepare production capacity for demand after 25 years? Does the customer guarantee the company? How does Hyperscaler's Capex change Derisk?
We communicate with many customers, and almost every AI company works with us, including these hyperscalers. Looking at long-term market needs, we have some pictures. We have collaborated with them on a rolling basis. We are disciplined in building production capacity to prepare the right scale of production capacity to support the needs of our customers and maximize shareholder interests.
[Q] There are some energy challenges in Taiwan. How should the company respond when building 2nm production capacity in Hsinchu and Kaohsiung? Now that Hyperscaler is building a nuclear power plant, will the company use nuclear power?
We do build a lot of factories in Taiwan, and we need electricity, water, and land. We continue to cooperate with relevant departments to inform us about our needs and plans, and get assurances that they will support us. We still can't share how they prepare electricity, such as nuclear power and green electricity. But we can definitely get enough electricity, water, and land.
[Q] Regarding non-AI requirements, what do you think PC and mobile phones will be in demand in 25 years?
The growth of PCs and mobile phones is still low single digit, but judging from the content, there are more chips with AI functions. Demand for silicon wafers is growing faster than terminals because of their larger size. We expect both businesses to gradually grow because AI applications will be healthier.
【Q】Regarding advanced packaging, what is the revenue outlook for the next few years? Is it possible to achieve the company's average gross profit margin?
Advanced packaging will grow at a higher rate than the company in the next five years, reaching a high single-digit ratio. Gross profit is close to average gross profit margin, but not yet.
[Q] Regarding free cash flow, will the company consider increasing cash margin? How does the company balance ongoing investment with shareholder interests?
Our dividend policy is sustainable and growing; we invest before we harvest. The increase in free cash flow is why we have reaped our past investments. Looking at the balance sheet, we use cash to achieve organic growth and maximize shareholder benefits. We pay free cash flow back to shareholders.
[Q] Compared to the traditional Foundry 2.0 model, what are the growth opportunities for each part of the Foundry 2.0 model?
In the foundry 2.0 model, advanced manufacturing processes and advanced packaging have higher growth, and mature processes and traditional packaging are not in such a good situation.
[Q] Can you update the outlook for semiconductor growth for the whole year? How about next year?
Very similar to what I said last time; our growth rate is a bit better than last time. But the industry as a whole is about the same as previously estimated. We think it's still too early for 25 years of growth; we'll talk about that next quarter.
【Q】What are the construction plans for the latest Cowos production capacity this year and next year?
We have invested a lot of effort to expand cowos production. Roughly speaking, our customer demand exceeds our supply. Even if production capacity doubles this year and continues to double next year, it's not enough. We are still continuously working to meet customer needs.