Good afternoon, everyone, and welcome to TSMC's second quarter 2024 earnings conference and conference call. This is Jeff Su, TSMC's director of investor relations and your host for today. Today's event is being webcast through TSMC's website at www.tsmc.com, where you can also download the earnings release materials. If you are joining us through the conference call, your dial-in lines are in listen-only mode. The format for today's event will be as follows: first, TSMC senior vice president and CFO, Mr. Wendell Huang, will summarize our operations in the second quarter 2024, followed by our guidance for the third quarter 2024. Afterwards, Mr. Huang and TSMC's chairman and CEO, Dr. C.C. Wei, will jointly provide the company's key messages. Then we will open both the floor and the line for the Q&A session. As usual, I'd like to remind everybody that today's discussions may contain forward-looking statements that are subject to significant risks and uncertainties, which could cause actual results to differ materially from those contained in the forward-looking statements. Please refer to the safe harbor notice that appears in our press release. And now, I would like to turn the microphone over to TSMC's CFO, Mr. Wendell Huang, for the summary of operations and the current quarter guidance.

Thank you, Jeff. Good afternoon, everyone. Thank you for joining us today. My presentation will start with the financial highlights for the second quarter of 2024. After that, I will provide the guidance for the third quarter 2024. Second quarter revenue increased 13.6% sequentially in NT or 10.3% in U.S. dollars, as our business was supported by strong demand for our industry-leading three- and five-nanometer technologies, partially offset by the continued smartphone seasonality. Gross margin increased 10 basis points sequentially to 53.2%, mainly reflecting cost improvement and a more favorable foreign exchange rate, partially offset by the margin impact from M3 RAM. Due to the operating leverage, total operating expense accounted for 10.5% of net revenue as compared to 11.1% in the first quarter. Thus, operating margin increased 0.5 percentage points sequentially to 42.5%. Overall, our second quarter EPS was TWD 9.56 and ROE 26.7%. Now, let's move on to revenue by technology. Three-nanometer process technology contributed 15% of wafer revenue in the second quarter, while five-nanometer and seven-nanometer accounted for 35% and 17%, respectively. Advanced technology, defined as seven-nanometer and below, accounted for 67% of wafer revenue. Moving on to revenue contribution by platform. HPC increased 28% quarter over quarter to account for 52% of our second-quarter revenue, surpassing 50% for the first time. Smartphone decreased 1% to account for 33%. IoT increased 6% to account for 6%. Automotive increased 5% to account for 5% and DCE increased 20% to account for 2%. Moving on to the balance sheet, we ended the second quarter with cash and marketable securities of TWD 2 trillion or USD 63 billion. On the liability side, current liabilities increased by TWD 23 billion mainly due to the increase of TWD 16 billion in accounts payable. Long-term interest-bearing debt increased by TWD 9 billion mainly as we raised TWD 12 billion in corporate bonds. On financial ratios, accounts receivable turnover days decreased by three days to 28 days. Days of inventory decreased by seven days to 83 days primarily due to higher N3 wafer shipment. Regarding cash flow and capex. During the second quarter, we generated about TWD 378 billion in cash from operations, spent TWD 206 billion in capex, and distributed TWD 91 billion for third quarter '23 cash dividend. Overall, our cash balance increased TWD 101 billion to TWD 1.8 trillion at the end of the quarter. In U.S. dollar terms, our second-quarter capital expenditures totaled TWD 6.36 billion. I finished my financial summary. Now, let's turn to our current quarter guidance. Based on the current business outlook, we expect our third quarter revenue to be between USD 22.4 billion and USD 23.2 billion, which represents a 9.5% sequential increase or 32% year-over-year increase at the midpoint. Based on the exchange rate assumption of USD 1 to TWD 32.5, gross margin is expected to be between 53.5% and 55.5%, operating margin between 42.5% and 44.5%. This concludes my financial presentations. Now, let me turn to our key messages. I will start by talking about our second quarter '24 and third quarter '24 profitability. Our second quarter gross margin was 53.2%, slightly ahead of the high end of our guidance, mainly as we saw a higher-than-expected overall capacity utilization rate as compared to our forecast three months ago. We have just guided our third-quarter gross margin to increase by 1.3 percentage points to 54.5% at the midpoint. This is primarily due to the higher overall capacity utilization rate in the third quarter and better cost improvement efforts, including productivity gains, partially offset by continued dilution from N3 ramp-up, N5 to N2 conversion costs, and higher electricity prices in Taiwan. Excluding the impact of foreign exchange rate, of which we have no control over, and factoring in the margin impact from our global manufacturing footprint expansion plans, we continue to forecast interim gross margin of 53% and higher is achievable. Next, let me talk about our 2024 capital budget. Every year, our capex is spent in anticipation of the growth that will follow in the future years, and our capex and capacity planning is always based on the long-term market demand profile. As the strong structural AI-related demand continues, we continue to invest to support our customers' growth. We are narrowing the range of our 2024 capital budget to be between USD 30 billion and USD 32 billion as compared to USD 28 million to USD 32 billion previously. Between 70% and 80% of the capital budget will be allocated for advanced process technologies. About 10% to 20% will be spent for specialty technologies, and about 10% will be spent for advanced packaging, testing, mass making, and others. At TSMC, a higher level of capital expenditures is always correlated with the higher growth opportunities in the following years. Now, let me turn the microphone over to C.C.

Thank you, Wendell. Good afternoon, everyone. I would like to begin with our demand outlook for the near term. We ended our second quarter with revenue of USD 20.8 billion, surpassing our guidance in U.S. dollar terms. Our business in this quarter was bolstered by robust demand for our industry-leading three-nanometer and five-nanometer technologies, although we did face some challenges due to ongoing smartphone seasonality. Looking ahead to the third quarter of 2024, we anticipate strong support from smartphone and AI-related demand for our advanced process technologies. For the full year 2024, we project a 10% increase in the overall semiconductor market, excluding memory, which is consistent with our forecast from three months ago. We are also redefining our original definition of the foundry industry to what we now call foundry 2.0. This new definition encompasses packaging, testing, mass making, and all integrated device manufacturers, excluding memory manufacturing. We believe this change better captures TSMC's expanding market opportunities. However, I want to stress that TSMC will remain focused on the most advanced back-end technologies to assist our customers in developing leading-edge products. According to this new definition, the foundry industry size was approximately USD 250 billion in 2023, compared to USD 115 billion previously. We forecast that the foundry industry will grow nearly 10% year over year in 2024. TSMC held a 28% share of this market in 2023, bolstered by our technology leadership and broad customer base; we expect this share to increase in 2024. In the past three months, we have witnessed a significant rise in AI and high-end smartphone-related demand from our customers compared to three months ago, resulting in an increased overall capacity utilization rate for our advanced three-nanometer and five-nanometer process technologies in the first half of 2024. Therefore, we continue to anticipate a strong growth year for TSMC in 2024. We are raising our full-year guidance, now projecting our 2024 revenue to slightly exceed mid-20% growth in U.S. dollar terms. Next, I will cover TSMC's capacity planning and investment strategies, particularly crucial given the high forecasted demand from AI-related businesses. TSMC aims to be the reliable technology and capacity provider for the global logical integrated circuit industry for the foreseeable future. The ongoing surge in AI-related demand underlines a strong structural demand for energy-efficient computing. As numerous AI applications emerge, the value of our technological capabilities is increasingly recognized, and our customers depend on us for the most advanced process and packaging technologies at scale in an efficient and cost-effective manner. Consequently, TSMC implements a disciplined approach to align with this long-term demand driven by AI, high-performance computing, and 5G trends. We collaborate closely with our customers for capacity planning and maintain a thorough system evaluating market demand from both top-down and bottom-up perspectives to decide on appropriate capacity investments. Our capital investment choices hinge on four guiding principles: technology leadership, manufacturing flexibility and responsiveness, maintaining customer trust, and ensuring sustainable and healthy returns. To achieve worthwhile returns from our investments, both pricing and costs matter. TSMC's pricing strategy is designed to provide value rather than being opportunistic. Currently, we are heavily investing in leading-edge, specialty, and advanced packaging technologies to bolster our customers' growth and facilitate their success. When our customers succeed, TSMC also thrives. For instance, we are pleased to observe that many of our customers have seen improvements in their profitability over the past few years. However, we've also encountered rising cost pressures due to increased process complexity, higher electricity costs in Taiwan, global fab expansion in higher-cost areas, and other inflationary challenges. Therefore, we will continue to work closely with our customers to demonstrate our value and diligently collaborate with our suppliers to optimize cost performance. We believe these efforts will enable TSMC to achieve sustainable and healthy returns, allowing us to continue investing in technology and capacity to support our customers' growth and our mission as a trusted foundry partner while achieving profitable growth for our shareholders. Lastly, I will discuss the status of our N2 and N16 technology introductions. Our two-nanometer and 16 technologies address the essential need for energy-efficient computing, attracting nearly all AI innovators to work with TSMC. We expect the number of new tape-outs for two-nanometer technologies in their first two years to outpace that of both three-nanometer and five-nanometer technologies during their initial two years. We aim to deliver full node performance improvements of 10% to 15% in speed at the same power, or a 25% to 30% power reduction at the same speed, along with a more than 15% increase in chip density compared to N3E. Development of N2 technology is on track, with device performance and yield meeting or surpassing expectations, leading to potential volume production trends in 2025 with a ramp profile similar to N3. In our ongoing strategy of continuous improvement, we have also introduced N2P, an extension of the N2 family, with an additional 5% performance boost at the same power or a 5% to 10% power advantage at the same speed beyond N2. N2 supports both smartphone and HPC applications, with volume production slated for the second half of 2026. Additionally, we rolled out N16 as our next nano chip-based technology, featuring a Super Power Rail (SPR) as a new offering. TSMC's SPR is a cutting-edge power delivery solution, being the industry's first to include a novel backside contact scheme that maintains gate density and device flexibility. Compared to N2P, N16 delivers an additional 8% to 10% speed enhancement at the same power rate, or a 15% to 20% power reduction at the same speed, along with a 7% to 10% increase in chip density. N16 is ideally suited for specific HPC products with complex signal routing and dense power delivery requirements, with volume production anticipated in the latter half of 2026. We are confident that N2, N2P, N16, and their derivatives will solidify our technology leadership and enable TSMC to seize growth opportunities well into the future. This concludes my key points, and thank you for your attention.

Thank you, C.C. Thank you, Wendell. This concludes our prepared remarks. Before we start the Q&A session, I would like to remind everyone to limit questions to two at a time to ensure that all participants have the chance to ask their questions. We will take questions from both the floor and the call. If you would like to ask your question in Chinese, I will translate it to English before management responds. Now, we will begin the Q&A session. I would like to take the first few questions from the floor, then we will move on to the call. Let's begin. The first question will be from Gokul Hariharan from JPMorgan. Thank you.

Thanks, Jeff. Good afternoon, C.C., and Wendell. I appreciate the insights on your future capacity plans. Regarding AI accelerator and related capacity, both front-end and advanced packaging, it's clear that demand is increasing significantly, with numerous customers seeking capacity at TSMC. In our previous discussions a couple of quarters ago, C.C., you mentioned that we anticipated supply and demand to balance by the end of this year. I would like to hear your thoughts on the current supply-demand balance for AI accelerator and CoWoS advanced packaging capacity. Additionally, during your symposium, you mentioned a 60% compounded growth rate for CoWoS capacity over the next four to five years. Could you provide details on how much CoWoS capacity you plan to build next year? Last year, you stated that you would be doubling the capacity this year. Now that we are halfway through the year, could you share your outlook on capacity expansion for next year? That's my first question. Thank you.

Gokul, I also tried to reach the supply and demand balance, but I cannot today. The demand is so high. I had to work very hard to meet my customer's demand. We continue to increase. I hope sometime in 2025 or 2026, I can reach the balance. You're talking about the CAGR of growth on rental increase of the CoWoS capacity. Now, it's out of my mind. I mean, we continue to increase whatever, wherever, whatever I can, okay? The supply continued to be very tight all the way to probably and how we can be eased in 2026. That's today's situation.

Any thoughts on next year capacity? Like are you going to double your capacity again next year for CoWoS?

The last time I said that this year, I doubled it, right, more than double, okay? So, next year, if I say double, probably, I will answer your question again next year, is more than double, okay? We're working very hard, as I said, wherever we can, whenever we can.

Thank you. My second question is about gross margins. The guidance for the second half seems to be more optimistic than initially expected, with gross margins potentially increasing instead of decreasing. It seems many of the challenges affecting gross margins are occurring this year. How should we view gross margins moving forward for TSMC? Are we likely to return to the high 50% to 60% gross margins seen in 2022, considering you're selling more of your value and benefiting from yield improvements? Additionally, how should we assess the impact of subsidies and ITC credits as you begin expanding your overseas operations? What effects will that have on costs and gross margins? Also, TSMC is primarily discussing gross capital expenditures and gross spending right now.

OK. So, let me summarize Gokul's second question is around gross margin and profitability. He notes second half '24 gross margin seems to be better than the expectation. So, his question is really, how should we think about gross margin in the next several years? He notes as we said, we will sell our value and the dilution of N3 will gradually reduce. So, where can the gross margins go back to a high 50s or 60% that kind of level that we saw a few years ago in 2022? Maybe that's the first part of this question. I'll stop here, then I'll get to the second.

Sure. Let me share some thoughts on gross margin for 2025 and beyond. There are both positives and negatives to consider. One positive is the decrease in dilution from N3. We are effectively selling our value while also driving down costs and increasing productivity. Conversely, we might convert more N5 to N3 due to strong demand, which would have a negative impact in the year we do it, but ultimately be beneficial in the future. We are also dealing with cost challenges, particularly due to inflation and rising electricity prices. Additionally, we will start producing at our overseas fabs next year, including Phase 1 of the Arizona fab and Phase 1 of the Kumamoto fab. We anticipate these overseas fabs will reduce our gross margins by two to three percentage points in the coming years. Nonetheless, taking all this into account, we're confident that we can achieve a gross margin of 53% or higher, especially with our cost management efforts in Taiwan. That addresses the first part of your question.

Yeah, that's right.

And then maybe also just Gokul asked if it's possible to get back to the high 50%, 60% level that we saw in 2022.

Yeah. If we have a very high utilization rate, everything else stays the same. Possible.

OK. And then the second part of his question was, what is the impact from the different government incentives, including the CHIPS Act, ITC credits in the U.S., etc., to the financials? And also, I think partly gross capex and net capex.

Generally speaking, when subsidies are received, then you see that on the cash flow statement, it will be used to offset the asset value that will be on the balance sheet. When this fab begins to production, the P&L impact will come in. So, generally speaking, it's like that. Different governments have different approaches in providing the grants. So, that's a different story. But you can look at our financial statements; there will be actual subsidy received in the period of the previous quarter and previous year. For example, 2023, we received total subsidies of slightly higher than USD 1.5 billion equivalent, and we received that mainly in Japan. Yeah.

OK. All right. Great. Thank you. Let's move on. We'll take the next one from Charlie Chan from Morgan Stanley, and then we'll go to Bruce Lu from Goldman. Thank you.

I see an opportunity to meet you in person again. My first question is about your progress in demonstrating the value of what you offer. I'm curious about the current status and whether you anticipate a capacity shortage for next year. If that is the case, do you think it will enhance your ability to sell more value to your customers? Thank you.

OK. So, Charlie's first question is around pricing, and he wants to understand the progress of, I guess, selling our value. And also, next year, looking at next year, particularly for the leading-edge nodes, do we expect that in terms of the demand to be very full?

Charlie, this pricing strategy is very strategic. You are inquiring about the current status. So far, everything is going well. This is an ongoing process, and we are continuously communicating our value. Additionally, my customers are also performing very well. Therefore, we should also do well.

Yes. So, that is actually my follow-up question on this first question. For different segments, for example, HPC customers are doing very, very well. But for smartphone customers, probably more sensitive to the cost. Do you expect the kind of difference of kind of value increase for different customers, even at the same node?

So, Charlie is asking, how will we do the pricing? Will it be different between, for example, HPC customer versus a smartphone customer at the same node? And also, his question earlier was, do we expect the demand for the leading nodes to be very high next year?

Since the pricing is strategic, so it won't be flat for average product sector. So, it will be different, OK? That all I can share with you. And all my customers, they are looking for leading-edge as a capacity for the next few years, and we are working with them. And so, far, we try our best to support them, both in pricing and in capacity.

Thank you. And second topic is definitely over the past two days, the geopolitical risk. So, Mr. Donald Trump talked about, maybe a few weeks ago, right, Taiwan/TSMC took 100% chip business from the U.S. So, congrats on the bounce back pretty high market share. However, the concern is growing, right, that the U.S. continues to depend on our Island TSMC and the chip production. So, our question is for shareholders, right, how TSMC is going to mitigate this potential geopolitical risk? For example, whether you are going to further expand your U.S. capacity or even share the ownership, right, with the U.S. government? And maybe a technical question to Wendell, for today, right, if we are shipping a chip to the U.S. customers, do we need to pay for the U.S. tariff?

Charlie’s second question is about overseas expansion and geopolitical risk, specifically regarding comments made by former President Trump that TSMC has taken 100% of the business. He wants to know how TSMC plans to address geopolitical risks, including whether this involves expanding capacity in the U.S. He also asked about the possibility of joint ventures or partnerships, whether with the government or other partners. Lastly, there’s a related question for Wendell about tariffs on chips shipped to U.S. customers.

OK, Charlie. So, far, we did not change any of our original plan of expansion of our overseas fab. We continue to expand in Arizona, in Kumamoto, and maybe future in Europe. No change to our strategy. We continue our current practice. You mentioned about the JV, No. OK.

On the tariff, not that we know of. Normally, if there's an import tariff, the customers will be responsible for that, but no discussion. Nothing.

OK. Thank you. Thank you, Charlie. All right. We'll take the next question from Bruce Lu from Goldman Sachs in the front, then we'll move online.

Thank you for my question. I would like to know why we aren't increasing our gross margin or structure for the PBT target. TSMC has been emphasizing the value of its services for the past few quarters without adjusting the margin target, which suggests that you are likely absorbing all the costs. I recall in 2021, TSMC raised its gross margin target to support future growth alongside increased R&D as technology improves. One of your customers supports this by suggesting that you should charge even more. So, my question is, why aren't you raising your gross margin target while trying to demonstrate your value, which we believe warrants a higher rate?

OK. So, thank you, Bruce. So, Bruce's first question is about profitability and value. Bruce seems to agree that TSMC is providing value to our customers. He also notes in 2021, indeed, a few years ago, our gross margin target, long-term gross margin target was about 50%, and we're able to increase that to 53% and higher. So, his question is really with everything that is going on today with the value of our technology, enabling our customers more and more, why doesn't TSMC increase or revise up our long-term gross margin target from the current 53% and higher? Is that the essence?

Thank you for acknowledging TSMC's value, Bruce. I'm collaborating with our customer on this. As I mentioned, our pricing strategy is deliberate, and we aim to showcase our value. At this point, I want to focus on the target of 53% and above, and I won't be changing that number for now. Once I have further discussions with my customer, I may be able to provide a higher target. Thank you.

OK. My next question is for advanced packaging. So, management used to mention that Advanced Packaging margin was lower than the corporate age, but with higher ROICs. But given the recent progress for CoWoS and everything, do we see a much better profitability for the CoWoS? And given that it's so difficult to expand the capacity, are you planning to work with more partners to increase your CoWoS supply, which will start your current supply and demand issues?

OK. Thank you, Bruce. So, Bruce's second question is around advanced packaging. Part of it is in terms of the profitability. He notes we used to say, which is true, it's lower than the corporate average profitability but can earn a similar return or ROE, but his question is now with more and more CoWoS was demand and greater scale is the profitability of advanced packaging, I think, approaching or at or above the corporate average? And also, given the tight supply, would we consider to work with more partners to help increase the capacity for CoWoS to support our customers' growth?

You are correct that the gross margin for advanced packaging used to be significantly lower than the corporate average. However, it is now getting closer to that average. We are improving it due to economies of scale and our efforts to reduce costs. The gross margin has seen substantial improvements over the past two years. Regarding our collaboration with OSAT partners, we are indeed pursuing that because the current CoWoS capacity is insufficient and in high demand, which has limited customer growth. Therefore, we are working with our OSAT partner to increase capacity for our customers, allowing them to expand. This will also help in selling TSMC's wafer.

OK. Thank you, C.C. Thank you, Bruce. Operator, can we move to the first participant online for their questions, please?

Yes. First one, we've got Brett Simpson, Arete. Go ahead, please.

Yeah. My question was really about your capacity plans for the next node at N2, including N16. We're hearing that AI chip makers are looking to migrate more aggressively from N1 to the leading edge, particularly due to backside power because they're trying to lower their power budgets going forward. So, my question: can you support this move? And if so, should we be expecting N2, N16 to be structurally a much bigger node than we've seen in the past few nodes?

OK. Brett, thank you. So, Brett's first question is on capacity planning, particularly at the leading-edge N2 and N16. So, he notes rightly that AI customers are migrating aggressively from N1 in the past to the most leading node. He notes particularly N16 driven by the interest in backside power. So, his question is can we support this move in terms of capacity to support the customers and whether thus N2 and N16 will be a much bigger node than our nodes in the past?

Brett, you are right. All the people want to move into kind of a power-efficient mode. And so, they are looking for the more advanced technology so that they can save power consumption. And so, a lot of my customers want to move into N2, N2P, N16 quickly. We are working very hard to build the capacity to support them. Today, it's a little bit tight, not a little bit; actually, today is very tight. I hope in next year or the next two years, we can build enough capacity to support this kind of demand. Today, yes, we are working hard to support them. And enough, not yet, but we are working hard to get it.

Does that answer your first question, Brett? OK. Thank you. Yeah, that's great, Jeff. My follow-up question was for Wendell. I wanted to just dig into the gross margin dilution from N3; where is that at today? And does the introduction of N3E structurally improve your N3 returns? I guess N3 is less capital-intensive. There's fewer EUV layers, so I'm keen to understand that this drives better economics for TSMC, particularly as you start to ramp more entry capacity in the second half of this year. Thank you.

All right. Thanks, Brett. So, Brett's second question is on the gross margin dilution from N3. He notes that N3 uses fewer EUV layers, less capital intensity. So, his question is, as we ramp N3 more and more, does N3E structurally improve the returns and gross margin of N3 as a whole?

OK. Brad, we don't break it down between the different nodes within the family. But I can share with you, overall speaking, as we said before, N3E takes a longer while to reach the copper mark. In the past, it was about eight to 10 quarters. For N3, we're looking at 10 to maybe 12 quarters. But it is improving, and we expect it to continue to improve.

OK. Thank you, Brett. Operator, let's take the next set of questions from the next participant on the call, please.

Next one to ask a question, Charles Shi from Needham. Go ahead, please.

Hi. Thanks for taking my questions. Maybe the first one, just want to follow up. Wendell, I think I heard you talking about that potentially more implies to N3 conversion. Maybe what you're trying to do right now, the conversion. I just want to understand the overall philosophy here because I think in the past, TSMC does do this node-to-node conversion quite actively, let's say, 10-nanometer to seven-nanometers, probably even earlier 20-nanometer to 16-nanometer. And I think you told us basically treat 10 and 7 as one large node, 20 and 16 as one large node. Should we start to really think about maybe 5 and 3 as just one big node and maybe more conversion, we should think about more of the N3 capacity growth will come from conversion going forward less from the greenfield investment? That's the first question.

OK, Charles. So, Charles' first question is really look at our conversion strategy. He notes that we have always talked about building in tool commonality to provide us flexibility. We have done so in the past that certain nodes like 20 and 16, 10 and seven. So, his question is really we had said that we potentially convert more N5 tools to support the strong demand for N3 capacity. So, his question is should we, investors, start to think of N5 and N3 as one big node?

OK. Right. You mentioned about 12 and 16, they are a big foundry. Seven and 10 are a big family. But five and three are not a big family in our definitions. At the same time, there are node-to-node to commonality in TSMC is pretty high. So, for five and three, the commonality of tools is over 90%, and these two nodes are adjacent. They're all in Tainan Science Park. And so, it's very easy to do the conversions. Did I answer your questions?

Yes. I may ask a second question?

Certainly.

Thanks. Maybe a question about CoWoS. I think I heard, C.C., you said maybe there's some technical difficulty on my side. I just want to clarify, maybe you may double the CoWoS capacity again in 2025. But a little bit more technical question I do want to better understand the technology constraints because your customers to be migrating from CoWoS to the more advanced version CoWoS, and we learned that CoWoS ARM does not require TSV, does not require a large silicon interposer, does that help at least to some degree, the capacity constraints you're facing on overall CoWoS, and does that help to maybe to achieve that goal of maybe getting to the supply demand balance some point in 2025, 2026. That's a two-part question. Thank you.

OK, Charles. So, you're asking about CoWoS. First, let me clarify that we stated CoWoS capacity will more than double in 2024. You're correct in understanding that it is expected to double again in 2025. Additionally, as customers transition from CoWoS-S to CoWoS-L and CoWoS-R solutions, there are several changes that may help to alleviate capacity constraints. I’d like to address whether this will enable CoWoS to achieve a balance between supply and demand by 2025.

Well, all right. Charles, you really know all the details of the technology. The CoWoS-R, CoWoS-L blah blah blah. All these kinds of things because of our customers' requirements. So, even the same customer, they have different approaches for their different products. When I say that we doubled the capacity, this is some in or the different version of the CoWoS together. Which portion is really double which portion is much more than the other one, not going to share with you because this is related to my customer demand. So, from last year to this year, we have more than doubled. And as I said, from this year to next year, we want to double again or probably we want to more than double again. But still, I have to work with our OSAT partners to increase the overall supply to support my customers. Whether that this kind of different version of the CoWoS will give me some flexibility today, yes, and no, because different version has a different tool set. But in common, some of the tools can be used by all the CoWoS, okay, but different versions have different demands.

OK. Thank you, C.C. Thank you, Charles. We'll come back to the floor for the next two questions, please. We'll take the first one from Laura Chen from Citigroup, and then we'll go to Sunny Lin from UBS.

Thank you, Jeff. I appreciate you taking my question. My first question also concerns the advanced node. C.C., you mentioned earlier that every client is currently engaged with you on the two-nanometer migration. I’m curious if, as we look ahead to 2026, in the second year, we expect the revenue contribution to be greater than what we saw with N3. Additionally, given that the performance is significantly improved, should we anticipate the dilution period to be shorter than it was for N3?

OK. Sorry. So, Laura's first question is about basically that almost every customer is engaging with TSMC on two-nanometer technology. So, her question is do we expect the revenue contribution from N2 in 2026, therefore, to be larger than compared to N3 at a similar point in time of the ramp and also correspondingly with the N2 margin dilution be less or better than N3, essentially?

I will give this kind of monetary question to the CFO, OK?

All right. Laura, the revenue, yes, it’s going to be bigger, OK? Gross margin dilution, it will be faster to reach corporate average.

OK. That's very clear and helpful. Thank you. And also, my second question is also on the packaging side. So, we know that last time we also discussed that AI will also benefit for TSMC in terms of the advanced nodes, the die areas. Just wondering that do you also see your HAI device clients are moving to 3D IC or SoIC anytime in the next two years? Or before that happens, can we expect that more clients on the smartphone side, they will also adapt maybe information first? Because so far, our understanding is that info only have one advanced one single time. I'm just wondering if we see that more clients to move to on the AI side on advanced packaging. Thank you.

Laura's second question is quite specific, but regarding advanced packaging and the increasing number of customers working on edge AI devices, what does this mean for the advanced packaging solutions we expect to see in the next two years? Will these edge AI customers begin to utilize SoIC or 3D IC, especially in smartphones? Will they continue to use information, or will they also explore these new solutions? Is that accurate, Laura?

Well, very technical questions. Let me share with you as my customers move into two-nanometer or N16, they all need to probably take in the approach of chiplets. So, once you use your chiplets, you have to use advanced packaging technologies. On the edge AI, for those kind of smartphone customers, as compared with the HPC customers, HPC is moving faster because of bandwidth concerns, latency of footprints, or those kind of thing. For smartphone customers, they need to pay more attention to the footprint as well as the functionality increase. So, you observe my big customers taking the info first, and then for two years, nobody catch it up. They are catching up. OK?

Thank you very much.

OK. Thank you, Laura. We'll take the next question from Sunny Lin from UBS, and then we'll go back to the call.

Thank you, Jeff. Good afternoon and thank you for taking my questions. So, my first question is on your business opportunities for smartphone and PC. Last few years, both were X growth for quite some time. And so, how we should think about the units and silicon content for the coming two, three years? First part, a lot of questions on the tight supply five- and three-nanometer. And so, are your customers engaging with you early on the planning into 2025 capacities for a better upgrade cycle? And then for silicon content, recall a few years back when 5G just started to ramp are used to provide the silicon content expectations of 5G high-end and low-end smartphones, so I wonder at this point in time, if you have any estimates for AI for smartphone going to the next two, three years?

That's quite a lengthy question. However, I'll address the main points first. AI implementation is challenging. Currently, all of my customers want to incorporate AI capabilities into edge devices, which will result in an increase in die size. The specific increase can vary from customer to customer, but a general estimate would be around a 5% to 10% rise in die size. As for unit growth, we haven't seen a significant increase yet, but we anticipate that the demand for AI functionality will lead to a quicker replacement cycle. Therefore, in a couple of years, we expect to see a substantial increase in edge devices, such as smartphones and PCs.

And will we have enough capacity to support?

That's one I try to avoid. The answer is it's very, very tight, and we are working very, very hard to get enough capacity to support my customers from now all the way to next year to 2026.

Got it. Thank you for the answer. So, my second question is try to look at the demand profile from different perspectives. If we look back in 2021 or early 2022, but then demand was also pretty high. customers. We're very aggressive on the demand forecast. Now, looking at GenAI, obviously, the technology has lots of great potential, but a new technology also has lots of volatilities where you start to ramp. And so, how are we managing the volatility of the demand? Why do you think this time around is different versus the COVID period? How do we get comfortable with our capacity planning?

Thank you, Sunny. Regarding your second question about TSMC's capacity planning and capital expenditure framework, you mentioned that although AI-related demand is currently very strong, a few years ago, specifically in 2021 and 2022, demand was also quite robust, with many customers optimistic about future needs. Given the current strong demand for generative AI, how is TSMC planning its capacity effectively? How do we handle the risk of not overbuilding in this kind of environment?

I thought I explained that our capacity premium process, right, and the investment we have, I put wording of discipline. That means we are not going to repeat the same kind of mistake that we have in 2021, 2022. Now, this time, again, we look at the overall very big demand forecast for my customer. And so, I look at it into actually the whole company with many people now examining and study that really is AI is so used for will be used by a lot of people or not. And we test our sales force inside TSMC, while using AI, we are using machine learning scale to improve our productivity, and we found out it's very useful. And so, I also in the line to buy my customer's product, and we have to form in the line, like I get no privilege, I'm sorry, but it's useful. And so, I believe that this time, AI demand is more real than two or three years ago. That timing is because people are afraid of a shortage, and so automotive, everything, you name it, they are all in shortage. This time, AI alone only AI alone. It will be a very useful tool for human beings to improve all the productivity in our daily life, be it in medical, industry, or in any product, manufacturing industry, or autonomous driving, everything you need the AI. And so, I believe it's more real. But even with that, we also have a top-down bottom-up approach and discuss with our customers and ask them to be more realistic. I don't want to repeat the same kind of mistake two or three years ago, and that's what we are doing right now.

Great to know. Thank you very much.

OK. Thank you. Operator, can we move on to the next participant from the line, please? OK. If not, then maybe we'll take the last two questions from the floor or one or two. Let's start here and then here. So, we'll start with our third line from Arthur Lai from Macquarie.

Hi, C.C., Wendell, and Jeff. Thank you for taking my question. This is Arthur Lai from Macquarie. I previously covered downstream technology and especially data centers. I want to ask about the SPR because it’s crucial from a data center viewpoint. When you introduce the new technology, you can save about 20% on power. Can we also consider that total system power consumption could be reduced by another 20%? That’s a significant change. From the discussions you’ve had with customers, you mentioned that they can also see a reduction in their overall operating costs. So, it appears that the more you purchase, the more you save. Thank you.

OK. So, Arthur's first question, he would like to understand more about Super Power Rail or our best-in-class backside power solution as it relates to data center demand. He knows as we said that it brings greater power efficiency from the chip level. His question without specific numbers, but what does this mean for the system-level power consumption saving? What does it mean for our customers' ability in terms of total cost of ownership in terms of the power savings? And does it mean that the more you buy, the more you save?

The more you buy TSMC's wafer, the more you save. Regarding your question, Arthur, about the 20% sales in chip power consumption, this does not necessarily indicate that the system power consumption was reduced by 20%. The overall system involves various components, including connections, performance, and power consumption processes. Therefore, unless every component saves 20%, achieving a 20% reduction overall is unlikely. However, both the accelerator and the CPU are significant contributors to the overall system's power consumption. Even if it's not a full 20%, it's still a considerable portion. This is why all my customers want to utilize leading-edge technology and are eager to transition to two-nanometer technology.

Thank you. So, I also encourage the company to do the right thing. So, energy-efficient computing is definitely our goal for human beings. And so, I also, I would like to give more color about when you go into the N16 and when we expand the capacity, what do you think the biggest bottleneck would be?

OK. So, Arthur's second question is in terms of what would be the biggest bottleneck to expand our capacity of N16 to support our customers, if any.

We always say that when TSMC wants to expand the capacity, we need the land, we need the electricity, we need the talented people, and so all the above.

OK. Thank you. And then in the interest of time, we'll take questions from the last participant on the floor, which is Brad Lin from Bank of America Merrill Lynch.

Thank you for taking my question. I have two questions. The first one is regarding the compute test. We've noticed that several large tech companies have announced plans to accelerate their launch schedules. What does this mean for TSMC? Will it provide TSMC with better visibility into the pipeline and aid in capacity planning? Additionally, what major challenges might arise from this accelerated timetable?

Brad's first question is about recent announcements from several companies at Computex regarding their plans to speed up their product launches. He wants to know what this means for TSMC in terms of capacity planning and how it will support our customers. Is that correct? Yes.

We appreciate this trend since TSMC excels in leading-edge development. Each product takes about one and a half to two years to design. We received this information quite some time ago, and my customer is thrilled about it, which makes us happy as well because they recognize our value. I heed this advice. In response to your question, yes, we have been prepared. Although they announced it in June, we had already discussed these changes much earlier and made the necessary preparations.

Thank you very much. I would assume that would help us communicate the value more effectively. My second question is regarding the larger role of AI chips. There is a lot of activity surrounding fan-out panel-level packaging. Do you believe that solution will be relevant in the mid- to long term? Or does TSMC plan to make related investments? Thank you.

OK. So, Brad's second question is that, again, with AI-related chips, that they're larger and larger die sizes. So, his question is in terms of advanced packaging and specifically fan-out panel-level packaging, is this something that TSMC is looking at or exploring to do? Would this be something for TSMC in the mid- to long term?

Yes, we are exploring panel-level fan-out technology. However, its maturity is not yet sufficient, and I personally anticipate at least three more years before it becomes viable. In the meantime, we currently do not have a reliable solution for die sizes greater than 10 times the radical size. At present, we support our customers with sizes up to 5x and 6x. I believe that in two years, panel fan-out will begin to be introduced, and we are actively working on it. And we will be ready for it as well.

All right. Thank you, C.C. Thank you, Brad. Thank you, everyone. This concludes our Q&A session. Before we conclude today's conference, please be advised that the replay of the conference will be accessible within 30 minutes from now, and the transcript will become available 24 hours from now, both of which are going to be available through TSMC's website at www.tsmc.com. So, thank you, everyone, for joining us today. We hope everyone continues to stay well, and we hope you will join us again next quarter. Goodbye, and have a great day.