The Zhitong Finance App learned that TSM.US (TSM.US) announced financial results for the third quarter of 2026 on Thursday. According to the data, TSMC's Q3 consolidated revenue was NT$1.4942 trillion (US$46.72 billion), up 50.9% year on year and 17.6% month on month, exceeding the company's own financial range of 44.6 billion to 45.8 billion US dollars; net profit to the mother was NT$763.06 billion, up 68.7% year on year and 8.0% month on month, breaking the single-quarter record for the sixth consecutive quarter, and the 11th consecutive quarter of profit growth; diluted earnings of NT$29.43 per share, higher than the market estimate of approximately NT$28.3 billion. The Q3 gross profit margin was 66.4%, down 1.3% month-on-month, mainly diluted by the 2nm process slope climbing.
By process, the most advanced 2 nm shipments accounted for 8% of wafer sales in the third quarter (3% in the previous quarter), 3 nm accounted for 33%, 5 nm accounted for 28%, 7 nm accounted for 10%, and advanced manufacturing processes (7 nm and below) accounted for 79% in total — the sharp rise in volume and price of advanced nodes is the core of revenue exceeding expectations.
TSMC expects fourth-quarter revenue to be between $52.3 billion and $54.3 billion, up about 14% month-on-month, and gross margin between 65% and 67%; US dollar revenue for the full year 2026 is expected to increase by about 44%. In terms of capital expenditure, the company reiterated its 2026 capital budget of $60 billion to $64 billion.
TSMC management emphasized many times during the conference call that “demand is higher than supply” and “very, very high,” and that the gap between supply and demand is still quite large. Regarding AI demand, management said that AI-related demand is supported by strong basic demand from AI accelerators. At the same time, the emergence of intelligent AI CPUs and the continued growth in demand for network chips are further driving demand for chips.
Wei Zhejia, CEO of TSMC, said during the conference call that the compound growth rate of AI accelerators was previously expected to reach the upper middle level in the 50% range from 2024 to 2029, and the current demand growth rate is likely to exceed this expectation; the company “has no intention of leaving opportunities to others” and hopes to seize every segment of the business as much as possible.
TSMC expects capital expenditure from 2026 to 2028 to be significantly higher than the previous three years to support multi-year customer demand driven by AI. Although TSMC did not provide capital expenditure guidance for 2027 (and this is exactly what the market wants to know), the company has strong confidence in the AI supertrend.
Here are the details of TSMC's Q3 earnings conference call.
I. Executive speeches
1. Wendell Huang (Wendell Huang) - Chief Financial Officer
Good afternoon, and thank you all for joining our conference call today. My introduction will begin with financial highlights for the third quarter of 2026, followed by performance guidance for the fourth quarter of 2026.
In the third quarter, in NTD, our revenue increased 17.6% month-on-month, mainly supported by strong demand for advanced process technology. In US dollars, revenue increased 16.2% month-on-month to $46.7 billion, above the upper limit of our third-quarter results guidance. The gross margin fell 1.3 percentage points month-on-month to 66.4%, mainly due to the dilution of profit margins brought about by the 2nm process climbing. Total operating expenses accounted for 7.8% of net revenue, and operating margin was 58.7%. Overall, our third-quarter earnings per share reached NT$29.43, and our return on net assets (ROE) was 45.5%.
Now let's look at revenue by technology node. 2nm process technology contributed 8% of wafer revenue in the third quarter, with 3nm, 5nm, and 7nm accounting for 33%, 28%, and 10%, respectively. Advanced processes — defined as processes at 7nm and below — account for 79% of wafer revenue.
Next, let's take a look at each platform's contribution to revenue. High-performance computing (HPC) revenue increased 22% month-on-month, accounting for 68% of third-quarter revenue. Smartphone business revenue increased 10%, accounting for 21%. The Internet of Things (IoT) revenue business grew 5%, accounting for 4%. Revenue from the automotive business increased 23%, accounting for 4%. The consumer electronics (DCE) revenue business declined by 12%, accounting for 1%.
Next, let's look at the balance sheet. By the end of the third quarter, our cash and marketable securities holdings reached NT$4 trillion, or $126 billion. In terms of debt, current liabilities increased by NT$110 billion month-on-month, mainly due to: the current portion of bonds payable and bank loans increased by NT$46 billion; dividends payable increased by NT$26 billion. Judging from financial indicators, the number of receivables turnover days decreased by 1 to 28 days. Inventory turnaround days decreased from 5 to 82 days, mainly due to increased wafer shipments.
Regarding cash flow and capital expenses, in the third quarter, our operating activities generated approximately NT$920 billion in cash, capital expenditure was NT$443 billion, and NT$156 billion was paid for cash dividends for the fourth quarter of 2025. Overall, at the end of the quarter, our cash balance increased by NT$488 billion to NT$3.6 trillion. In US dollars, total capital expenditure for the third quarter was $13.8 billion.
This concludes my financial summary. Now let's take a look at the current quarter's performance guidelines.
Based on the current business outlook, we expect fourth-quarter revenue to be between $52.3 billion and $54.3 billion. In median terms, this represents a 14% month-on-month increase in revenue and a 58% year-over-year increase. Based on an exchange rate assumption of US$1 to NT$31.8, gross margin is expected to be between 65% and 67%, and operating margin between 56.5% and 58.5%.
Now let me introduce our core message. I'll start by talking about profitability for the third and fourth quarters of 2026.
Compared with the second quarter, our gross margin for the third quarter fell 130 basis points month-on-month to 66.4%, mainly due to the dilution of profit margins due to the initial rise in 2nm technology. The fourth quarter gross margin guideline we just gave was 65% to 67%, with a median value of 66%, down 40 basis points from month to month, mainly due to the dilution of profit margins brought about by 2nm technology that continued to climb rapidly, and the dilution of profit margins brought about by overseas fabs. These effects are expected to be partially offset by cost improvement measures, including increased productivity and capacity optimization across process nodes.
Although the cost of overseas fabs is still high, thanks to the company's overall scale expansion and cost improvement measures, we currently expect the gross margin dilution due to the downhill slope of overseas fabs to be close to 2 percentage points. To support the strong structural needs of our customers, we are speeding up the construction of more fabs, particularly in the US. As a result, we expect the gross margin dilution from overseas fabs to expand to 3% to 4% in the next few years. We will take advantage of the growing size of our Arizona fab and continue to improve our operations to optimize our cost structure. We will also continue to work closely with our customers and suppliers to manage the impact.
Overall, with the two fundamental competitive advantages of leading manufacturing technology and large-scale production sites, we are confident that TSMC will continue to be the most cost-effective manufacturer in every region where we do business.
Next, let's talk about the 2026 capital budget. At TSMC, higher levels of capital expenditure are always associated with higher growth opportunities in the next few years. Given the continued strong multi-year customer demand, including the emerging smart AI market, we are reaffirming a capital budget of $60 billion to $64 billion for the full year 2026.
We continue to expect capital expenditure investments from 2026 to 2028 to be significantly higher than in the past three years as we continue to invest heavily to support customer growth. Furthermore, despite the tight supply of equipment, we always work closely with equipment suppliers in advance to prepare for production capacity, just as our customers work with us to plan production capacity in advance.
Even if capital expenditure investments are made at this level over the next few years to support future growth, we remain committed to achieving profitable growth for our shareholders. We also remain committed to maintaining a sustainable and steadily growing cash dividend per share at both the annual and quarterly levels.
2. C.C. Wei (C.C. Wei) - Chairman and CEO
Good afternoon everyone. First, let me start with the immediate outlook for demand.
In the third quarter, our revenue reached 46.7 billion US dollars, higher than previous guidance in dollar terms, mainly driven by strong demand for advanced process technology. Entering the fourth quarter of 2026, we expect the business to continue to be supported by strong demand for advanced process technology, including the rapid climbing of 2nm technology.
Looking ahead, we continue to observe that consumer and price-sensitive terminal markets are facing challenges due to rising component prices and the impact of macroeconomic uncertainty. As a result, as we enter 2027, we remain cautious in our business planning while focusing on business fundamentals to further strengthen our competitive position.
Having said that, the demand for our advanced process technology remains extremely strong due to strong technical differentiation and support from a broader customer base. As a result, we now expect revenue in US dollars to increase by approximately 44% year over year for the full year 2026.
Next, let's talk about the outlook for AI-related demand.
The AI market is still very strong. Recent developments are very positive for the long-term demand for AI. As we move from chatbots to AI assistants, we have observed a continuous increase in the adoption rate of intelligent AI, which is driving the demand for AI to grow. Furthermore, we have also observed a high adoption rate of open source AI models. This will improve efficiency, lower the threshold for future use of AI, and further drive demand for advanced chips.
As a result, AI-related demand is being supported by strong basic demand from AI accelerators. In addition, the advent of smart AI CPUs and the growing demand for network chips are further driving demand for chips. Therefore, we believe semiconductors will continue to be an integral part of the AI supertrend.
In our view, the value of semiconductors in the global supply chain will continue to increase. This will create more opportunities for our customers to obtain value, and will also create more value opportunities for TSMC. Whichever solution is used, the most advanced semiconductor process technology is required. With our technological leadership, superior manufacturing capabilities, global capacity expansion, and customer trust, we are fully prepared to seize all the opportunities brought by years of AI-related structural requirements. Therefore, all AI companies are cooperating with TSMC.
Next, let's talk about TSMC's production capacity expansion strategy.
As a foundry, our primary responsibility is to provide our customers with the best technology and effective capacity support to unleash their ability to innovate. In order to respond to the structural growth in long-term demand in the semiconductor market, we work closely with our customers and their customers to jointly plan production capacity. Furthermore, we have now taken another step forward and sent our own team to check the US data center construction progress and power supply. The feedback so far has been excellent, and the cloud service provider (CSP) is well prepared.
Internally, TSMC uses a strict capacity planning system to assess market demand from both top-down and bottom-up perspectives. It's an ongoing process. Given the complexity of advanced process technology itself and the long design introduction and delivery cycle, we have a very clear understanding of the customer's product roadmap and production plans for many years to come. As a result, we still have strong confidence in the AI supertrend.
We are expanding our production capacity at full speed to meet our assessed needs and increase capital expenditure investments to support our customers' future growth. TSMC is expanding at an unprecedented rate, and construction of about 20 fabs around the world is currently underway. In the next few years, we will be building 13 advanced process and advanced packaging fabs in Taiwan, and another 5 to 6 fabs overseas.
Our new process node N2 has entered large-scale mass production in the fourth quarter of 2025, and the yield performance is good. N2 is successfully climbing in multiple stages at two locations in Hsinchu and Kaohsiung. We anticipate that by 2027, the annual production capacity of N2 family technology will more than double the current level, which will be supported by strong demand for smartphones and HPC/AI applications. We are also using our excellent manufacturing capabilities to increase N2 wafer output and further enhance the production capacity support we provide to our customers.
Additionally, we have begun preparing production capacity for A14 technology. Risk trial production in 2027 is progressing smoothly, and mass production will begin in 2028. Therefore, TSMC's semiconductor technology and manufacturing capabilities will continue to play a key role in supporting the global semiconductor industry and unlocking customer innovation.
Now let's talk about the latest developments in TSMC's global N3 production capacity expansion.
As we said before, historically, once a process node reaches its target capacity, we usually don't add additional production capacity. However, according to our assessment, we are successfully implementing our global capacity plan and adding three additional fabs to support 3nm technology's strong multi-year demand pipeline.
In Taiwan, China, our new 3nm fab in Tainan Science Park is progressing well, and mass production is scheduled to begin in the first half of 2027. Furthermore, driven by strong customer demand, we decided to add another 3nm fab in Tainan. The fab is scheduled to begin mass production in the first half of 2028.
In Arizona, we have completed the construction of our second fab and are currently moving in and installing the equipment. The fab is scheduled to begin mass production in the third quarter of 2027, slightly ahead of schedule. In Japan, we are advancing the adoption of 3nm technology at our second fab as planned, and mass production is scheduled to begin in 2028. In addition to all these new fabs, we are continuing to convert 5nm equipment to support 3nm production capacity in Taiwan.
We are also using our superior manufacturing capabilities to improve the production efficiency of fabs in all regions. In order to increase wafer output, we are also focusing on optimizing production capacity between different process nodes, including providing flexible capacity support between N7, N5, and N3 processes. Therefore, we are using various means to do our best, wherever we can, expand production capacity as quickly as possible, and support our customers to the greatest extent possible. We don't intend to leave any business opportunities to anyone else, and we want to take every business as much as possible.
This concludes my core message, thank you all.
II. Q & A session
1. Citibank analyst Laura Chen
Good afternoon. Thanks also to C.C., for sharing TSMC's production capacity expansion strategy. I'd like to ask some more questions. Given the increasingly aggressive demand for AI semiconductors, how does TSMC view AI revenue CAGR growth now?
Furthermore, how does TSMC balance customer capacity needs with possible concerns about data center power supply? Recently, we have also seen some financing issues, as well as issues with other key components such as HBM, APS, and advanced packaging. In terms of network equipment, we are also seeing a tightening supply of lasers. If these trends cause a mismatch between semiconductor supply and actual AI development, how will TSMC adjust production capacity to minimize the risk of excessive construction in the supply chain?
This is my first question.
Wei Zhe Family
This is a very, very long question. Laura, let me share it with you. We have a very advanced capacity planning process, and we also use a very strict evaluation method. We will consider all possibilities, including CSP, capital expenditure, data center construction, financial status, electricity supply and other parts supply, etc. We make our own judgments.
Based on our assessment of needs, we still believe that demand is higher than supply. So we are working very hard to close this gap.
Laura Chen
So far, based on our thorough research on the supply chain, we are still quite optimistic about long-term demand. In that case could you elaborate a bit more? What are the latest changes in AI CAGR growth? I remember we were talking about about 50% more before.
Wei Zhejia
Laura, I can only tell you right now: demand is extremely strong. We're not seeing any signs of slowing down. As to whether it's more than 50% CAGR, it doesn't matter. It could be higher.
Laura Chen
My second question also relates to very strong demand and capital expenditure cycles. We see that this capital expenditure cycle will clearly be much higher than before. TSMC also previously announced that it will adopt high-NA EUV around 2030. Should we think this will further improve production efficiency and thus accelerate revenue growth more significantly in the longer term? Furthermore, as the intensity of capital expenditure increases, how should we understand the company's pricing power, as well as long-term gross margin and capital efficiency returns?
Wei Zhejia
Let me answer that question. Yes, in terms of long-term production efficiency, we will continue to improve. The High-NA is definitely a very good device. Currently, I think our plan is to use it in 2030. It is emphasized once again that our capital expenditure will be much higher than today.
You're asking if gross margins or profitability can be increased. Let me put it this way: there is no upper limit to TSMC's profitability, but a sharp increase in profit margins will not happen overnight. Our philosophy or strategy has always been to work with our customers. We want to help our customers succeed, and then we also gain value for ourselves.
2. J.P. Morgan analyst Gokul Hariharan
My first question is about capacity expansion. There appears to be a shortage of supply in the market. C.C., I'd like to hear your thoughts. I think we haven't seen TSMC increase production capacity by more than 10% each year for a long time. If you look at the company's annual report, I don't think there has been a 10% increase in production capacity for about 10 years. Given the obvious acceleration of current production capacity construction, can we begin to see wafer production capacity growth of close to 10% or even more than 10% in the next few years? Can you comment on the planned increase in wafer production capacity over the next few years?
This is my first question.
Wei Zhe Family
As I mentioned earlier, demand is extremely strong and growing. Of course, we have our own judgement and will look at it from every angle. As I said, we communicate with the customer, communicate with the customer's customer, and then we will further check the data center construction situation ourselves.
We include all results in our judgments and evaluations. The result: demand is very, very high. The gap between demand and supply is still quite large. That's why we're mobilizing all our forces to increase production capacity, as I just said.
As to whether the gap will widen or narrow, we are working very hard to resolve this. I can't answer that right now because our customers are still growing in demand.
Huang Renzhao
OK, Gokul. I think it's too early to discuss future capacity growth percentages. I just want to remind everyone that although the annual growth figure may not be large, the growth of advanced process nodes is actually the main driving force behind the increase in production capacity.
Gokul Hariharan
Understandable. Thank you. Second, I think you mentioned more aggressive overseas expansion plans, particularly in the US. Recently, there has also been a lot of discussion in the market, such as the possibility of building a new fab cluster in Texas. I think one of your customers also mentioned discussions with TSMC. C.C., can you talk about overseas plans, such as more accelerated capacity building? Also, given that one of your customers mentioned some collaboration with TSMC last week, is there any contact with SpaceX or Tesla? What might be the general framework for these collaborations? Is it similar to how you work with some of your customers in Japan and ESMC?
Wei Zhe Family
OK, Gokul. We are working very hard to expand production capacity in the US. This is necessary because most of our large customers are in the US, and they are asking TSMC to increase production capacity. The US government is also very supportive. This is where we want to expand.
Recently, we have further expanded the scope of our capacity assessment to see if there are other expansion opportunities outside of Arizona. Texas is one of them. We've sent a team to Texas to conduct all relevant assessments. We are also communicating with a customer trying to build a fab, more specifically Terafab. We're in discussions with them to see if we can also utilize the actual production capacity of the project. It's still under discussion. If we come to any clear conclusions, we'll share them with you.
Gokul Hariharan
OK. But all of these fabs will eventually be operated by TSMC, right? I don't think you guys are going to be part of the Terafab project or anything like that. Can you help us understand what principles you're using here? Are the principles similar to those used by JASM or ESMC?
Wei Zhejia
We're still discussing it. Let me put it this way: it's possible either way. But the fab must be operated by TSMC.
3. Krish Sankar, analyst at TD Cowen
I have two questions. First question, C.C., some customers talked about the so-called “secondary wafer market,” where one customer transfers their wafer quota to another. Is this just a very small part of the business? Is this a customer-specific situation, or do you think it's becoming a new trend where less demanding customers transfer part of their quota to more demanding customers?
Wei Zhe Family
If this is a new strategy, a new approach, the answer is no. TSMC's principle is that we do not allow customers to sublease or auction production capacity. However, in some cases, customers can cooperate through a mask licensing agreement. However, this isn't our common practice. The scale is very, very small.
Krish Sankar
Then ask a quick question. You did mention a high-NA agreement with Asmack starting in 2030. I'm curious. Historically, TSMC has been at the forefront of EUV adoption. However, it appears that by 2030 or 2031, when TSMC uses High-NA EUV, some of its competitors may already be using this technology. Do you think this would create a competitive disadvantage? Or aren't you worried about your competition leading the way in terms of high-NA EUV?
Wei Zhe Family
We're always evaluating new devices. If we want to use it for mass production, it must have: the best technology; a very reasonable cost; and most importantly, the highest degree of maturity. We said we'll start using it in 2030 because we think High-NA will mature by then. We are also working closely with Asmack to plan to establish a 12-inch mask capability by 2031. At that point, we think the cost of High-NA will be more reasonable and suitable for TSMC to put into large-scale production.
Let me emphasize this: this is a very good device. At a smaller line width, it does have some advantages. As to whether we use it later than the competition would put TSMC at a disadvantage? It won't. Because we are confident: we are still the best company to provide our customers with the most advanced technology.
4. Morgan Stanley analyst Charlie Chan
I'm happy to hear that in the face of recent stock market adjustments, demand for AI semiconductors is still very strong. My first question really relates to the company's unprecedented capacity expansion strategy. As far as I understand, one of your IDM customers is actively expanding production capacity. There are some concerns in the market that this customer may increase in-house production rather than outsource to TSMC.
First, I'm not sure how you guys are going to handle this kind of business opportunity. Second, I remember when C.C. mentioned last time that A14 might become a larger node than N2. I'm wondering if this assumes that this big IDM company will outsource more CPUs on A14 nodes, such as server CPUs?
Wei Zhe Family
Charlie, we treat every customer equally. Whether it's an IDM customer or a pure Fabless customer, we treat them the same. Of course, we also consider the customer's own in-house production capacity. We calculate the support we can provide and the scale of capacity construction based on our own judgment.
In fact, as far as A14 is concerned, we have seen more and more customers participating in A14 events so far, and the scale is already higher than 2nm. We expect: A14 will be a larger node than current N2 technology.
Charlie Chan
I get it. In other words, there is no need to assume that the IDM company will also increase outsourcing; we can still see A14 becoming a larger node than N2. Am I understanding it right?
Wei Zhe Family
Charlie, as I said, frankly, I'm not going to comment on specific customers. But let me assure you: we'll take everything into account.
Charlie Chan
The second question is a follow-up question about advanced packaging. Judging from what we've observed, the reticle size is getting bigger. C.C., you mentioned TSMC's response plan last time or about two quarters ago, including SoW, 3.5D, etc., so many projects are currently underway.
My question is: Do you think these SoW 3.5D and other solutions can really satisfy customers who are rapidly developing larger reticle-sized chips, such as 14x reticle chips? Second, if a customer initially chooses to use Intel's EMIB-T but is unable to obtain sufficient supply for some reason, will TSMC also provide CoOS solutions or production capacity to these customers?
Wei Zhe Family
OK, Charlie. In fact, I still think that TSMC's advanced assembly, or so-called advanced packaging technology, is very competitive. But most importantly, we use these technologies to help our customers' products enter the market smoothly. This is our number one priority. We must provide the best technology and sufficient production capacity.
Currently, I can show you a reality: we have achieved 5 times the reticle size in actual production. I don't think any other company other than TSMC can do this right now, and our yield is excellent. As a result, our customers are very satisfied with the results. We are advancing 9x reticle size and continuing to move towards 14x reticle size. Of course, we are also promoting CoWoS to ensure reasonable cost reductions in the future.
But looking at it now: CoWoS is still the number one choice for customers. Over time, yes, our competitors have offered very competitive solutions. But I still want to emphasize: Currently, we are the only company that can ensure mass production of 5x reticle size advanced packages, and we are still very competitive. I want to emphasize that.
5. UBS analyst Sunny Lin
My first question is, I want to know more about how we should look at capital spending beyond 2026. Remember a few quarters ago, you mentioned that if you do a good job, you could make capital expenditure grow slightly lower than revenue growth. But now it's clear that overall customer activity is likely to be further strengthened, and you are also willing to support customers to expand on a larger scale.
So I'm wondering: Looking at it now, is it possible for capital expenditure to be in line with sales growth, or even slightly faster than sales growth, just like 2026? Meanwhile, due to very tight supply, equipment prices are also rising. Therefore, entering 2027, will the rise in equipment prices also become an important factor driving up capital expenditure?
Huang Renzhao
OK, Sunny. It is too early to talk about capital expenditure figures beyond 2026. As we said in our preparation statement, we have very strong confidence in the AI supertrend. As a result, we expect capital expenditure from 2026 to 2028 to be significantly higher than in the past three years. This is a higher level of investment to support customer growth. We have no intention of leaving any business opportunities to anyone else.
Equipment price inflation is real. Equipment prices, materials, labor costs, etc. are all rising. However, the majority of capital expenditure has always been related to long-term demand. The composition of future capital expenditure will also be the same. As far as capital expenditure intensity is concerned, as we have always said: if we get the job done right, revenue growth will exceed capital expenditure growth. Based on what we've seen so far, this still holds true.
Sunny Lin
If so, I'd like to quickly ask about capital expenses. Earlier, Wendell mentioned that due to the acceleration of overseas expansion, the gross margin dilution of overseas business will reach 3% to 4% starting in 2027. As far as the US is concerned, your current focus is still on Arizona. Does that mean Arizona's second phase of expansion will be very rapid? Or are you actually speeding up the expansion of the third phase or even other stages? As a result, will the margin of gross margin dilution expand further next year?
Huang Renzhao
As we said, we are doing everything possible to expand production capacity wherever we can to meet strong customer demand. We're also speeding up in various regions, including the US. We have anticipated that in the second half of the five-year cycle, the gross margin dilution brought by overseas fabs will reach 3 to 4 percentage points. This still holds true. As a result, in the second half of the five-year cycle, dilution from overseas fabs will reach 3 to 4 percentage points.
6. Bank of America analyst Haas Liu
My first question is about your thoughts on potential changes in the structure of AI requirements. You mentioned in your preparation statement that AI accelerator chip requirements are expanding to smart AI requirements, including CPUs and network switching chips. We understand that overall AI investment is still very strong, and trends in various fields are even strengthening.
Can we learn more: Are you seeing customer demand shifting to more efficient computing solutions? Especially as commercialization becomes more important. How should we understand the incremental opportunities this brings to the manufacturing and packaging business?
Wei Zhe Family
All I can say is: AI demand is extremely strong. As to whether the demand comes from the accelerator or more from the CPU side, this is not something I can comment on. However, we do see continued growth in demand for intelligent AI CPUs. Yes, this is a trend right now. Other than that, everything is pretty strong.
I would like to emphasize once again: according to our judgment, we have considered all parameters, and the current demand is actually still much higher than what we can currently provide. We are working very hard to close this gap. That's all I can say.
Haas Liu
My second question is about the yield learning curve under the future dual technology route strategy. Historically, I think you've been improving the yield of new process nodes by producing chips with smaller chip sizes and simpler designs, thereby improving the learning curve for new processes. However, for future process node families that support both traditional power supply solutions and Super Power Rail technology: traditional power supply is mainly used in mobile devices and consumer products; Super Power Rail is used in HPC and AI chips. If you run two process architectures at the same time, will it make the learning curve more difficult or increase manufacturing complexity? How should we understand the impact of this on gross margin, including yield and cost control aspects?
Wei Zhe Family
In fact, the current situation is different from past years. Yes, we are now facing very strong demand from both smartphones and AI/HPC applications. Therefore, the small chip size itself occurs at the same time as the accelerator's large chip size.
Let me say: this is good for TSMC. Because we're good at that. We're prepared for this situation, and our customers are happy to work with us. So of course there is a learning curve. But we did a pretty good job. So the customer is currently very satisfied.
7. Macquarie analyst Arthur Lai
My question has to do with electricity as a constraint. I'd like to know: Have grid restrictions caused you to lower customer forecasts in the past? Additionally, as grid restrictions become more severe, is this likely to prompt customers to accelerate adoption of Super Power Rail products for better power performance?
Wei Zhe Family
Undoubtedly, the electricity supply situation is a consideration we are watching. This is why we are further checking the US electricity supply. We've done the relevant research. So far: Very good condition.
As to whether this situation will change customer behavior to move more towards back rail technology? The answer is: no. Whether to use backside power supply and this type of metallization scheme depends on the product's own needs. Some accelerators might be used. However, some products, such as mobile phones, won't be used. So it depends on the customer's product.
Arthur Lai
Back to the previous discussion about AI CPU requirements for smart devices. I want to know: according to our previous demand forecasting framework, does more than 50% of the growth include CPU? If we see customers drastically increase CPU demand forecasts, when will AI demand growth data be updated next time?
Wei Zhe Family
The CPU is not included in our AI requirements calculation. The reason is very simple: we can't determine whether a CPU user is a traditional server CPU or an AI server CPU. Furthermore, this also depends on the customer's application and the customer's own production capacity allocation. So, not right now. As of today, we're still sticking to the initial, very narrow definition of AI requirements.
8. Stifel Financial Analyst
Entering 2027, N3 began contributing incremental profits, and this stage is now over. N2 is still in the upward phase of profit dilution, and profit margin dilution for overseas businesses is expanding. In terms of offsetting factors, there are recent reports that prices for both your mature and advanced processes have risen sharply. Can you tell me more about the situation? If the price does change, does it usually apply to an existing contracted order, or does it mainly apply to additional production capacity beyond what the customer has already promised?
Wei Zhe Family
As we've discussed before: Our pricing strategy is strategic. The price also depends on: market segmentation; geographical location; technology; other factors. Once we provide the flexibility of a physical production location, such as manufacturing in the US, then the price will be different from what we manufacture in Taiwan. Regardless, we strive to get our customers to recognize our value and our shareholders to maximize the value created by our technology.
Sahej Singh
My second question concerns a quote you mentioned in your preparation statement. I was very impressed by this statement: “We don't leave any business opportunities on the table.” About a week or two ago, GlobalFoundries announced that it would provide TSMC with a silicon intermediate layer. I'd like to know: How do you consider the relationship between procurement from a specialized foundry and internal supply in these strategic decisions?
Wei Zhe Family
We designed this partnership agreement with GlobalFoundries as part of our overall back-end supply strategy. In fact, our top priority is to help our customers succeed in the marketplace. We expect our customers to receive full support across all aspects of supply, including front-end and back-end.
Currently, the back-end does have some shortage of supply. Therefore, we want to work with our so-called partner GlobalFoundries. In this way, both TSMC and GlobalFoundries can support TSMC customers' products and ensure that products have sufficient production capacity support when entering the market.