TSMC (TSM.US) Q3 Conference Call: Customer demand is extremely strong and far exceeds supply; capital expenditures for 2026-2028 will be significantly higher than the previous three years

According to Zhitong Finance APP, TSMC (TSM.US) released its Q3 2026 financial report this Thursday. The data showed that TSMC's Q3 consolidated revenue reached NT$1.4942 trillion (US$46.72 billion), an increase of 50.9% year-on-year and 17.6% quarter-on-quarter, exceeding the company’s guidance range of US$44.6–45.8 billion. Net income attributable to shareholders amounted to NT$763.06 billion, up 68.7% YoY and 8.0% QoQ, hitting a new single-quarter record for the sixth consecutive quarter and achieving profit growth for the 11th straight quarter; diluted EPS was NT$29.43, higher than the market estimate of around NT$28.3. Q3 gross margin stood at 66.4%, down 1.3% QoQ, mainly due to the ramp-up dilution of the 2nm process.

By process node, the most advanced 2nm shipments accounted for 8% of wafer sales in Q3 (versus 3% last quarter), 3nm made up 33%, 5nm took 28%, and 7nm comprised 10%. Advanced nodes (7nm and below) contributed a combined 79% of sales— the growing volume and price of advanced nodes were the key drivers for revenue surpassing expectations.

TSMC expects Q4 revenue to be between US$52.3 billion and US$54.3 billion, up about 14% quarter-on-quarter, with a gross margin of 65% to 67%; full-year 2026 revenue in US dollars is expected to grow by approximately 44%. On capital expenditures, the company reiterated that its 2026 capital budget is between US$60 billion and US$64 billion.

TSMC management emphasized multiple times during the earnings call that "demand exceeds supply" and is "very, very high," with a significant gap remaining. Regarding AI demand, management stated that AI-related demand is supported by a strong base requirement for AI accelerators, while the emergence of agent AI CPUs and the continued growth in networking chip demand are further driving chip demand.

TSMC CEO C.C. Wei said on the call that previously, the company anticipated a compound annual growth rate for AI accelerators in the upper mid-50% range between 2024 and 2029, but current demand growth very likely exceeds this estimate; TSMC "has no intention of leaving opportunities for others" and aims to capture every possible piece of business.

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 guidance for 2027 capital expenditures (which is what the market wants to know), the company remains highly confident in the AI supercycle trend.

Below are the details from TSMC's Q3 earnings call.

I. Executive Remarks

1. Wendell Huang – Chief Financial Officer

Good afternoon, everyone. Thank you for joining our conference call today. My presentation will start with key financial highlights for Q3 2026, followed by performance guidance for Q4 2026.

In the third quarter, calculated in New Taiwan Dollar, our revenue grew 17.6% quarter-on-quarter, mainly supported by strong demand for advanced process technologies. In US dollar terms, revenue increased 16.2% QoQ to US$46.7 billion, exceeding the upper end of our Q3 guidance. Gross margin declined 1.3 percentage points QoQ to 66.4%, mainly due to profit dilution from the 2nm node ramp. Total operating expenses accounted for 7.8% of net revenue, and the operating margin was 58.7%. Overall, our Q3 EPS was NT$29.43, and return on equity (ROE) was 45.5%.

Now, let's look at revenue by technology node. The 2nm process contributed 8% of wafer revenue in Q3, while 3nm, 5nm, and 7nm accounted for 33%, 28%, and 10%, respectively. Advanced node processes—defined as 7nm and below—comprised 79% of wafer revenue.

Next, let's examine platform contributions to revenue. High-Performance Computing (HPC) revenue grew 22% QoQ, comprising 68% of Q3 revenue. Smartphone business revenue increased 10%, accounting for 21% of sales. Internet of Things (IoT) revenue was up 5%, contributing 4%. Automotive business revenue rose 23%, accounting for 4%. Consumer Electronics (DCE) revenues declined 12%, making up 1%.

Turning to the balance sheet. At the end of Q3, our cash and marketable securities totaled NT$4 trillion, or US$126 billion. On the liability side, current liabilities increased NT$110 billion QoQ, mainly due to NT$46 billion growth in the current portion of bonds and bank loans and a NT$26 billion increase in dividends payable. Days sales outstanding reduced by 1 day to 28, and inventory turnover days decreased by 5 to 82, mainly due to increased wafer shipments.

Regarding cash flow and capital expenditures, in Q3, we generated about NT$920 billion in cash from operations, had capital expenditures of NT$443 billion, and paid NT$156 billion for Q4 2025 cash dividends. Overall, our ending cash balance increased by NT$488 billion to NT$3.6 trillion. In US dollar terms, Q3 capital expenditures were US$13.8 billion.

This concludes my financial summary. Now, let’s turn to the current quarter’s performance guidance.

Based on our current business outlook, we expect Q4 revenue between US$52.3 billion and US$54.3 billion. At the midpoint, this means 14% revenue growth QoQ and 58% YoY. Assuming an exchange rate of 31.8 NT dollars to the US dollar, we expect gross margin to be in the 65% to 67% range and operating margin between 56.5% and 58.5%.

Now let me share our core messages. First, a word on profitability for Q3 and Q4 2026.

Compared with Q2, our Q3 gross margin declined by 130 basis points QoQ to 66.4%, mainly due to gross profit dilution from the early ramp-up of 2nm technology. Our Q4 gross margin guidance is 65% to 67%, mid-point 66%, a 40 basis point sequential decline, mainly due to continued rapid 2nm ramp-up pressure and profit margin dilution from overseas fabs. These effects should be partially offset by continued cost improvement measures, including productivity improvements and cross-node capacity optimization.

Although overseas fabs still have higher costs, thanks to the company’s overall scale expansion and cost improvement measures, we now expect the gross margin dilution from overseas ramp-ups to near 2 percentage points. To support strong, structural customer demand, we are accelerating fab construction, particularly in the US. As a result, over the next few years, gross margin dilution from overseas fabs is expected to widen to 3% to 4%. We will leverage the scale of our Arizona fab and continue to improve operations to optimize the cost structure, while closely working with customers and suppliers to manage the related impact.

Overall, with fundamental competitive advantages in leading manufacturing technology and large-scale production, we are confident that TSMC will continue to be the most cost-efficient manufacturer in every region we operate.

Moving to the 2026 capital budget. At TSMC, elevated levels of capital expenditure have always been tied to higher growth opportunities in coming years. Given persistent, multi-year robust customer demand including the emerging agent AI market, we reaffirm our full-year 2026 capital budget at US$60 billion to US$64 billion.

We continue to expect capital expenditure investment for 2026 to 2028 to be significantly above the past three years, as we continue to invest heavily to support customer growth. Furthermore, despite tight equipment supply, we always work closely with suppliers in advance to prepare capacity, just as our customers also collaborate with us early for capacity planning.

Even if we invest at this level over the next few years to support future growth, we remain committed to delivering profitable growth for shareholders. We are also committed to maintaining a sustainable and steadily rising cash dividend per share, both annually and quarterly.

2. C.C. Wei – Chairman and Chief Executive Officer

Good afternoon, everyone. Let me first start with our near-term demand outlook.

Third-quarter revenue hit US$46.7 billion, above the previous outlook in US dollars, mainly driven by strong demand for advanced process technologies. Moving into Q4 2026, we expect business to remain supported by robust advanced node demand, including the rapid ramp of 2nm.

Looking ahead, we continue to see that consumer and price-sensitive end markets are under challenge due to rising component prices and macroeconomic uncertainty. Thus, as we approach 2027, we remain prudent in our business planning and focused on core business fundamentals to strengthen our competitive position further.

That said, demand for our advanced process technology remains exceptionally strong, underpinned by our technological differentiation and expanding customer base. Therefore, we now anticipate full-year 2026 US dollar revenue will grow about 44% year-on-year.

Next, let’s talk about AI-related demand outlook.

The AI market remains extremely strong. Recent developments are highly positive for long-term AI demand. As we move from chatbots to AI assistants, we are seeing rising adoption of agent AI, which is fueling growth in AI demand. Additionally, we observe strong adoption of open-source AI models. This will improve efficiency, lower barriers to future AI usage, and further drive demand for advanced chips.

Thus, AI-related demand is underpinned by the base-level requirement for AI accelerators. Furthermore, the emergence of agent AI CPUs and relentless growth in networking chip demand are further fueling chip demand. Hence, we believe semiconductors will remain indispensable in the AI supercycle.

In our view, semiconductors’ value in the global supply chain will keep rising. This will create more value capture opportunities for our customers and for TSMC. Regardless of the solution adopted, the most advanced semiconductor process technology is necessary. With our leading technology, world-class manufacturing, international capacity growth, and customer trust, we are fully ready to seize all opportunities arising from multi-year structural AI demand. That’s why all AI players are working with TSMC.

Next, let's discuss TSMC’s capacity expansion strategy.

As a foundry, our top priority is to provide customers with the best technology and effective capacity support to unleash their innovation. To address long-term structural growth in the semiconductor market, we work closely with customers and customers’ customers on capacity planning. We’ve now gone a step further, sending our own team to check US data center construction progress and power supply status. The feedback is very good so far, and cloud service providers (CSPs) are well prepared.

Internally, TSMC adopts a rigorous capacity planning system, assessing market demand from both top-down and bottom-up perspectives. It’s a continuous process. Considering that advanced node technology is very complex, with long design and delivery cycles, we have very clear visibility into customer roadmaps and production plans for coming years. Thus, we remain highly confident in the AI supercycle.

We are rapidly expanding capacity to meet our evaluated demand and are increasing capex investment to support future customer growth. TSMC is scaling up production at an unprecedented pace, with about 20 fabs under construction globally. Over the next few years, we will build 13 advanced node and advanced packaging fabs in Taiwan and another five to six fabs overseas.

Our new N2 process node entered large-scale production in Q4 2025 and is ramping smoothly with good yields in both Hsinchu and Kaohsiung. By 2027, N2 family annual capacity will be more than double its current level, driven by strong smartphone, HPC, and AI demand. We are leveraging our manufacturing strength to boost N2 wafer output and further enhance capacity support for our customers.

Additionally, we have started preparing capacity for A14 technology. Risk trial production in 2027 is progressing well, with volume production starting in 2028. Therefore, TSMC’s semiconductor technology and manufacturing capabilities will continue to play a pivotal role in supporting the global semiconductor industry and customer innovation.

Now, the latest on TSMC’s global N3 capacity expansion.

As mentioned, historically, once a process node reached its target capacity, we typically did not further add incremental capacity. However, per our evaluation, we are rolling out our global expansion plan as planned and building three new fabs to support the robust, multi-year demand pipeline for 3nm technology.

In Taiwan, our new 3nm fab in the Tainan Science Park is progressing well, with plans for mass production in the first half of 2027. Additionally, due to strong customer demand, we have decided to add another 3nm fab in Tainan, scheduled to begin mass production in the first half of 2028.

In Arizona, construction of our second fab has been completed, with tool move-in and installation now underway. Volume production is slated to begin in Q3 2027, slightly ahead of schedule. In Japan, our second fab using 3nm technology is on track and will commence volume production in 2028. On top of these new fabs, we are also upgrading 5nm equipment in Taiwan to support 3nm capacity.

We are also leveraging our manufacturing strength to boost output efficiency at all our fabs globally. To increase wafer output, we are focusing on cross-node capacity optimization between N7, N5, and N3. Thus, we are doing everything possible, everywhere possible, to expand capacity as quickly as possible for customer support. We do not intend to leave any business opportunity to others— we aim to capture every deal we can.

That concludes my core remarks. Thank you, everyone.

II. Q&A Session

1. Laura Chen, Citi Analyst

Good afternoon. Thank you, C.C., for sharing TSMC’s strategy for capacity expansion. I would like to follow up further. Given the increasingly aggressive demand for AI semiconductors, what is TSMC’s view on the AI revenue CAGR growth?

Moreover, how does TSMC balance client capacity demand with potential concerns over data center power supply? Recently, we have seen some financing issues and challenges with HBM, APS, advanced packaging, and other critical components. On the networking equipment front, there’s tightening supply for lasers. If these trends result in a mismatch between semiconductor supply and actual AI development, how will TSMC adjust its capacity to maximize mitigation of any supply chain overbuild risk?

That is my first question.

C.C. Wei

That is a very, very long question. Laura, let me share with you— we have a highly advanced capacity planning process and a very strict evaluation approach. We take all possibilities into account, including CSPs, capital expenditures, data center construction, financial status, power supply, and other component supply. We make our own evaluations.

Based on our assessments, we still believe that demand exceeds supply. We are working very hard to narrow this gap.

Laura Chen

So far, based on our comprehensive research into the supply chain, we remain quite optimistic about long-term demand. In this context, could you elaborate further? Has there been any recent change in AI CAGR growth? I remember it being a little over 50% previously.

C.C. Wei

Laura, all I can tell you now is that demand remains extremely strong. We see no signs of slowdown. Whether or not the CAGR is a bit over 50% is not the point— it may be even higher.

Laura Chen

My second question also relates to the very strong demand and the capital expenditure cycle. We see that this capex cycle is clearly much higher than before. TSMC previously announced plans to adopt High-NA EUV around 2030. Should we think that this will further boost production efficiency, and thus could be a more significant driver for revenue growth in the longer run? Furthermore, as capital intensity rises, how should we understand the company’s pricing power, long-term margin, and capital return efficiency?

C.C. Wei

Let me answer this. Yes, regarding long-term production efficiency, we are continually making improvements. High-NA is undoubtedly an excellent tool. Currently our plan is to use it in 2030. Again, our capex will be much higher than it is today.

You asked whether margin or profitability can improve. Let me put it this way: TSMC profitability has no ceiling, but a large increase in margin will not happen overnight. Our philosophy or strategy is always to work with our customers. We want to help our customers succeed, and then capture value for ourselves as well.

2. Gokul Hariharan, JPMorgan Analyst

My first question is also about capacity expansion. It seems that there is a supply shortage in the market. C.C., I’d like your thoughts— I believe we haven’t seen TSMC grow capacity by more than 10% per year for quite some time. If you look at the annual report, it’s been about a decade since capacity grew by more than 10%. Given these accelerated capacity builds, over the next few years should we expect wafer capacity growth to approach or exceed 10% annually? Can you comment on your planned wafer capacity increases in the coming years?

That is my first question.

C.C. Wei

As I mentioned, demand is extremely strong and keeps rising. Of course, we have our judgments and observe from all angles. As I said, we communicate with customers, also customers’ customers, and additionally check data center construction status ourselves.

We take all results into our judgment and evaluation. The result is: demand is very, very high. The gap between demand and supply remains significant. That’s why we mobilize all forces to increase capacity, as I mentioned.

As for whether the gap will widen or shrink, we are working hard to address it. I can't answer that now because our customers are still raising demand forecasts.

Wendell Huang

Okay, Gokul. I think it’s still a bit early to talk about future capacity growth rates. I just want to remind everyone that even if the overall growth numbers may not seem big every year, growth in advanced process nodes is actually the main driving force for capacity increases.

Gokul Hariharan

Understood. Thank you. Next, I think you mentioned an even more aggressive overseas plan, especially in the US. There’s been a lot of market talk about possible new fab clusters in Texas, for example. I believe one of your customers also mentioned relevant discussions with TSMC. C.C., can you speak to the overseas plans, such as more accelerated capacity building? Also, considering that one of your customers mentioned some cooperation last week with TSMC, is there any interaction with SpaceX or Tesla? What might be the general framework of such cooperation? Is it similar to arrangements like in Japan or ESMC with some customers?

C.C. Wei

Okay, Gokul. We are working very hard to increase US capacity. It is necessary, because most of our large customers are in the US, and they are asking TSMC to expand capacity. The US government is also very supportive. It’s a place where we want to expand.

Recently, we also expanded our evaluation to see if there are more opportunities for expansion outside of Arizona. Texas is one of them. We have already sent teams to Texas for all relevant assessments. We have also communicated with a customer trying to build a fab, specifically Terafab. We are discussing with them to see if we can also utilize actual capacity from that project. It’s still under discussion. If we reach any firm conclusions, we will share them with everyone.

Gokul Hariharan

Okay. But all these fabs would end up operated by TSMC, right? I assume you wouldn’t be part of the Terafab project itself, or something like that. Can you help us understand what principles you’re applying here? Is it similar to what you do at JASM or ESMC?

C.C. Wei

We’re still discussing. Let me put it this way: both options are possible. But the fab must be operated by TSMC.

3. Krish Sankar, TD Cowen Analyst

I have two questions. First, C.C., some customers have talked about the so-called “secondary wafer market,” where one customer transfers their wafer allocation to another. Is this just a very small part of the business? Is this specific to individual cases, or do you see this becoming a new trend where weaker-demand customers shift part of their allocation to stronger ones?

C.C. Wei

If you’re asking whether this is a new strategy or approach, the answer is no. TSMC’s principle is not to allow customers to sublet or auction capacity. In some cases, customers may cooperate via mask authorization agreements. However, this is not a common practice and is very, very small in scale.

Krish Sankar

Then a brief follow-up. You did mention the High-NA agreement with ASML starting in 2030. Historically, TSMC has led in EUV adoption. But by the time TSMC uses High-NA EUV in 2030 or 2031, some competitors could already be using it. Is this a competitive disadvantage, or are you unconcerned about competitors leading in High-NA EUV?

C.C. Wei

We always assess new equipment. If we want to use it for high-volume manufacturing, it must have the best technology, very reasonable costs, and, most importantly, top maturity. We say 2030 for adoption of High-NA because we think it’ll be mature then. We are also working closely with ASML and plan to have 12-inch mask capabilities by 2031. By then, we expect the cost of High-NA to be more reasonable and suitable for TSMC mass production.

Let me emphasize: It’s an excellent tool. At smaller line widths, it does offer some advantages. As for whether using it later than competitors will put TSMC at a disadvantage, no, it will not. We are confident that we are still the best company to provide customers with the most advanced technology.

4. Charlie Chan, Morgan Stanley Analyst

Glad to hear AI semiconductor demand remains extremely strong despite recent market corrections. My first question relates to your unprecedented capacity expansion strategy. As I understand, one IDM customer is aggressively expanding its own capacity. There is market concern that this customer might increase internal production rather than outsourcing to TSMC.

Firstly, I’m curious how you’ll address such business opportunities. Secondly, I recall C.C. mentioning last time that A14 could be a larger node than N2. Does that assume this large IDM will outsource more CPUs at the A14 node, such as server CPUs?

C.C. Wei

Charlie, we treat every customer equally, whether IDM or pure fabless. Of course, we consider the customers’ own internal capacity. We make our own judgments to calculate the support we provide and the scale of capacity build.

Regarding A14, so far, we see more and more customers participating in A14 activities, now exceeding activity at 2nm. Our expectation is that A14 will become a bigger node than current N2 technology.

Charlie Chan

Understood. So you do not necessarily have to assume that the IDM will also increase outsourcing; you still expect A14 to be larger than N2. Is that correct?

C.C. Wei

Charlie, as I said, frankly, I will not comment on specific customers. But let me assure you: we take all factors into account.

Charlie Chan

Second question is a follow-up on advanced packaging. From our observations, reticle sizes are getting larger. C.C., you mentioned before, either last or two quarters ago, that TSMC’s solutions, including SoW and 3.5D, are being pushed forward.

My question is: Do you think these SoW, 3.5D options can truly meet the needs of customers who are developing larger reticle size chips, such as 14x reticle chips? Secondly, if a customer initially adopts Intel’s EMIB-T but can’t secure enough supply, will TSMC offer CoWoS options or capacity for those clients?

C.C. Wei

Okay, Charlie. I actually still think TSMC’s advanced assembly— the so-called advanced packaging technology— is very competitive. Most importantly, we deploy these technologies to help our customers get their products to market. That’s our top priority: to offer best-in-class technology and enough capacity.

Currently, I can share a real case: we have already achieved 5x reticle size in real production. I don’t think anybody else has done that, and our yields are excellent. Our customers are very satisfied. We are moving towards 9x reticle size, and advancing to 14x as well. Of course, we are also pushing CoWoS, to ensure reasonable cost reduction in the future.

At present, CoWoS remains the first choice for customers. Over time, yes, competitors offer very competitive alternatives, but I want to stress: as of now, we still are the only company mass-producing advanced packaging at 5x reticle size, and we remain highly competitive. I want to emphasize that.

5. Sunny Lin, UBS Analyst

My first question is hoping to further understand how we should view capital expenditure after 2026. I recall a few quarters ago you mentioned that if things are done well, capex growth can be slightly below revenue growth. But now, customer activity may be further increasing, and you are willing to support larger expansions.

So, currently, could capex keep pace with, or even be faster than, revenue growth as in 2026? Meanwhile, since supply is so tight, equipment prices are rising. Will higher equipment prices become a major upward driver for capex in 2027?

Wendell Huang

Okay, Sunny. It’s too early to comment on post-2026 capex figures. As we said in the prep remarks, we have very strong confidence in the AI supercycle. Therefore, we expect capex from 2026 to 2028 to be significantly higher than the past three years. This raises investment levels to support customer growth. We do not intend to leave business opportunities for others.

Equipment price inflation is real. Prices for equipment, materials, labor, etc. are all rising. However, most capex is always tied to long-term demand, and future capex composition will be similar. In terms of capex intensity, as we’ve always said: if we execute well, revenue growth will outpace capex growth. Based on what we see now, this still holds true.

Sunny Lin

If I may, a quick follow-up on capex. Previously Wendell mentioned that with accelerated overseas expansion, overseas business will dilute margins by 3–4 points starting from 2027. For the US, you’re mainly focused on Arizona. Does this mean Arizona phase 2 expansion will be very rapid, or are you also fast-tracking phase 3 or additional phases, leading to greater margin dilution next year?

Wendell Huang

As we said, we are doing everything possible, everywhere possible, to expand capacity for strong customer demand— including in the US. We previously noted that overseas fab dilution will reach 3 to 4 percentage points in the second half of a five-year cycle. That still holds. So in the back half of the five-year cycle, overseas fab dilution will be 3 to 4 percentage points.

6. Haas Liu, Bank of America Analyst

My first question is about your view on potential structural changes in AI demand. In your prepared remarks, you mentioned AI accelerator chip demand is expanding towards agent AI demand, including CPUs and networking switching chips. We understand AI investment is still very strong and even strengthening in all fields.

Can you elaborate: are you seeing demand shift toward more efficient computing solutions, especially as commercialization becomes more important? How should we think about incremental opportunities for manufacturing and packaging?

C.C. Wei

All I can say is that AI demand is extremely strong. As to whether demand comes more from accelerators or CPUs, I can’t comment. However, we do see continued growth in agent AI CPU demand— that is a current trend. Other than that, everything is very strong.

Let me reiterate: based on our judgment and all parameters, demand still far exceeds our current capabilities. We are working very hard to close this gap. That’s all I can say.

Haas Liu

My follow-up concerns the yield learning curve under the future dual-process-node strategy. Historically, you’ve improved new node yields by producing smaller and simpler chip designs. But going forward, with process node families supporting both traditional and Super Power Rail power delivery— the former for mobile/consumer, the latter for HPC/AI— will running both architectures in parallel make the learning curve steeper, or increase complexity? How should we think about the impact on margin, including yield and cost control?

C.C. Wei

Actually, it’s a bit different from past years. Yes, we now face very strong demand from both smartphone and AI/HPC applications, simultaneously.

Let me say: this is an advantage for TSMC. We are good at this, prepared for it, and customers are happy to work with us. There is a learning curve, but we’re doing quite well. Customers are satisfied so far.

7. Arthur Lai, Macquarie Analyst

My question is about power as a constraint. Historically, have grid constraints ever led you to lower customer forecasts? Also, as power constraints become more severe, might customers adopt Super Power Rail products faster for better power performance?

C.C. Wei

Without doubt, power supply is something we pay attention to. That’s why we checked US power supplies. So far: things are good.

As for whether this changes customer behavior to adopt backside power rail and such metallization technologies, the answer is: no. Adoption of backside power and related metal approaches depends on product requirements. Some accelerators will use it; others, like phones, won’t. It depends on the customer’s product.

Arthur Lai

Returning to the earlier discussion on agent AI CPU demand. Based on our previous demand framework, does the 50%+ growth include CPU? If customers significantly raise CPU demand forecasts, when would AI demand growth data be updated?

C.C. Wei

CPU is not included in our AI demand calculation. For a simple reason: we cannot determine whether a CPU user is using it for a classic server or an AI server. It also depends on the customer's use case and allocation. So, for now, no. As of today, we stick to a very narrow definition of AI demand.

8. Stifel Financial Analyst

Moving into 2027, N3 starts to contribute incremental profit and that cycle is done. N2 is still in the margin-dilutive ramp-up phase and the margin dilution from overseas business is widening. Regarding offsetting factors, there have been reports of significant price increases for both mature and advanced nodes. Can you elaborate? If there is a price change, does it apply to existing orders under contract, or mostly to new capacity beyond what customers have already committed?

C.C. Wei

As we have discussed before: our pricing is strategic. Prices also depend on segment, geography, technology, among other factors. Once we offer physical location flexibility, such as US production, pricing differs from production in Taiwan. Regardless, we work to get customers to recognize our value and maximize the value our technology generates for shareholders.

Sahej Singh

My second question is about a point you made in your prepared remarks that impressed me: “We will not leave any business opportunity on the table.” About a week or two ago, GlobalFoundries announced they will supply silicon interposers to TSMC. How does TSMC weigh professional foundry supply versus internal supply for such strategic decisions?

C.C. Wei

We have drafted this cooperation agreement with GlobalFoundries as part of our overall backend supply strategy. Actually, our main goal is to help our customers succeed in the market. We want customers to be well supported in all supply links, both front-end and back-end.

Currently, the backend is quite tight in terms of supply. Therefore, we want to cooperate with partners like GlobalFoundries. In this way, both TSMC and GlobalFoundries can support TSMC customers’ products, ensuring they have sufficient capacity support when they go to market.