TSM.US (TSM.US) FY26Q2 conference call: Q2 gross profit margin of 67.7% exceeded the guideline and raised annual capital expenditure to US$60-64 billion

The Zhitong Finance App learned that TSM.US (TSM.US) held an FY26Q2 earnings conference call. The company announced a total cash dividend of TWD 467 billion in 2025, an increase of 28.6% over the previous year, and a cash dividend of TWD 18 per share for the whole year; in 2026, the cash dividend per share will rise to TWD 24, an increase of 33% over the previous year, and management promised to continue to increase in 2027.

In terms of performance guidance, Q3 2026 revenue is expected to be 44.6 billion to 45.8 billion US dollars, corresponding to a median increase of 12% month-on-month and 37% year-on-year; gross margin is expected to be 65% to 67%, and operating margin is expected to be 56% to 58%. The company raised the dollar-caliber revenue growth rate for the full year 2026 to “slightly above 40%” (previously over 30%).

In terms of gross margin, Q2 gross margin increased 150 bps to 67.7% month-on-month, slightly exceeding the guideline; the median value of the Q3 gross margin guide decreased by 1.7% to 66% month-on-month, mainly due to the steep 2nm slope dilution gross margin of about 3-4%. In the second half of the year, 2nm climbing is expected to dilute the gross profit margin of 3-4%; overseas fabs dilute 2% to 3% in the early stages and expand to 3% to 4% in the later stages.

In terms of capital expenditure, the company raised its capital budget for the full year 2026 to US$60 billion to US$64 billion (the previous guidance was approximately US$52 billion to US$56 billion in January and US$56 billion in April). Of this, about 70% to 80% is used for advanced manufacturing processes, and about 10% to 20% is used for advanced packaging, testing, etc. Q2 capital expenditure in a single quarter was about US$15.7 billion; cash and marketable securities TWD 3.5 trillion (approximately US$110 billion) at the end of the quarter; inventory days increased by 7 to 87 days month-on-month due to N2 climbing.

Q&A Q & A

Q: Will this upward cycle provide capital expenditure guidance for the next three years (26-28) like 2021?

A: There are currently no specific figures to share. We are investing capital expenses this year for future business opportunities, and as long as there are business opportunities, we won't hesitate to invest. As we said in our prepared statement, we are very confident in the AI megatrends over the years. We are increasing our capital expenditure and raising our capital budget for this year. The last time we said that capital expenditure in the next three years will be significantly higher than in the past three years, the current situation is: capital expenditure in the next three years will increase significantly compared to the past three years.

Q: After adding 100 billion US dollars, Arizona's cumulative investment reached about 265 billion US dollars. What is the plan and schedule for production capacity implementation in Arizona in the next few years?

A: The schedule depends on market conditions. Looking at the current situation, the general trend is very strong, so we are announcing additional investment in Arizona. How many factories will be built? A lot. Specifically, it is likely that around 4 more fabs will be built. (Confirmed to include previous funding arrangements.)

Q: How to deal with competition from Foundry rivals such as Samsung (which obtains high profits from the storage business) and Intel (supported by US policy)? Some US customers are in touch with these peers, and ASML has announced that it will expand EUV production capacity in 2028. Are you worried that your rivals will seize more production capacity and compete with you?

A: Yes, one of my competitors in Korea made a lot of money, and I'm very envious; another competitor in the US received strong support from the US government — by the way, we also received government support, but we didn't announce it to the outside world. But as we said, there are no shortcuts. This means a return to fundamentals is essential in the semiconductor industry. We welcome and appreciate the government's support; of course, money is important. But the most important thing has always been technology, manufacturing, and customer trust, a fundamental aspect that hasn't changed in my 37-40 year career, and it has always been TSMC's secret to winning business. From a competitive perspective, choosing a technology and mass-producing it is not going to 7-Eleven to buy milk — it's not today that you see this bottle of milk is better and you go to the next store and don't like to switch to another one. You have to actually use it, verify it with test chips, work collaboratively, prepare production capacity, and then climb the hill. This will take about 5 years.

Q: Will the 5-year (semiconductor/AI) CAGR be revised since we have seen strong signals from customers and their customers and raised the full-year guidelines? Previously, it was around 50% higher; at the same time, rising storage costs accounted for a significant share of AI capital expenditure. How should TSMC's contribution to AI-related growth be viewed?

A: If you read our message — we continue to invest more and raise capital expenses, there are good reasons. If you ask about the CAGR of AI, I won't give you a number, but the answer is getting stronger, stronger, stronger. The reason we are not giving numbers today is because it continues to rise and is stronger than what we said before.

Q: Advanced packaging competition, especially EMT (Embedded/Partial Interconnect Packaging), is receiving market attention. How is TSMC responding?

A: Our packaging production capacity is very tight, which has limited customer growth. More additional flexibility on the market will help TSMC's wafer business (which is the main body of TSMC's business). The technology reportedly looks good, and we hope it succeeds to share some of TSMC's orders. We are currently working to close the gap between demand and capacity, so we welcome more alternatives and provide customers with flexibility.

Q: As a new technology, if customers have problems after adopting it, how will TSMC handle such special needs and is there an alternative?

A: Our number one priority is supporting customer success. We are willing to do anything that helps our clients' business.

Q: When planning to expand production capacity, in addition to customers and customer customer needs, do they also consider the competitive pressure brought about by competitors to build production capacity (as a market leader, long-term supply shortages are not ideal)? About how long will it take to meet current needs? In addition, in addition to the shortage of chips, there are also issues such as data center extensions and power supply. How can they be incorporated into the planning framework?

A: Competition is the primary consideration every time we consider the business; then evaluating where we are and using a combination of bottom-up and top-down methods to assess requirements is the hardest part of our daily work. We make a lot of judgments, and we will be more careful in communicating with customers and their customers (i.e. CSP) to obtain all required inputs before making decisions. Note that I believe every customer has told me the truth, but if you add up the “truth” of all customers, it's not necessarily the truth — because all CEOs can be aggressive, this is their job. So we need to make a very careful judgment, which may not be entirely correct, but since this is a large amount of money (capital expenditure has been raised from about $52 billion to $60 billion to $60 billion to $64 billion this year, and will continue to increase), we are doing it very carefully. We are also reviewing the progress of AI data centers — construction, site selection, demand, rent, etc., to ensure that TSMC's chips are not backlogged into inventory.

Q: Even with such a large-scale production capacity expansion plan, do you still think there will still be a shortage of supply until the end of next year?

A: You want a guarantee from me. I believe demand will be very strong from now until around 2029 and 2030. This trend is so steady, and I think we're seeing a new industry — I call it the AI industry, which has penetrated into everyday life and will impact automobiles, humanoid robots, and all industries. As far as the amount of capital invested by everyone (including all CSPs) is concerned, this is a very important new industry for the world, and demand will always exist, and it is essentially semiconductor chips, and most of them are at TSMC.

Q: Regarding profitability: In the long run, the profit of foundry manufacturing (especially advanced manufacturing processes) should be higher than storage; now TSMC is no longer the most profitable semiconductor manufacturer, does this mean that the current pressure to transfer to customers and capture its own value is less than a year ago?

A: Your question is actually quite simple — what is TSMC's pricing strategy, and what kind of gross profit margin should it have? Of course the higher the better. But we are partners. I've said many times that customers must be successful, and I don't want to extract too much value from the market; moreover, we are a company that customers can trust, and we won't increase prices 4 or 5 times all of a sudden, so customers can't survive. We earn the value we deserve and ensure that profits and gross margins are sufficient to support long-term sustainable expansion, which is beneficial to our customers and TSMC. This is our philosophy. So I'm really envious of the storage company's 86% gross margin — I'd be happy if I had 68%. All in all, we are a very trustworthy company.

Q: As AI demand significantly surpasses other terminal markets, the share of the top five customers has reached an all-time high. How do you view the risk of customer concentration?

A: That's not our concern. Our customers are getting bigger and we're happy, and some of them are growing really fast. Moreover, it's not as simple as “big customers get bigger”; in fact, many new players have emerged in the AI industry.

Q: Your direct customers are financing/investing in their end customers to support AI needs. Will TSMC consider making such investments or financing arrangements for customer customers (end customers)?

A: Direct answer to you: Every company has different considerations and strategies. Until now, TSMC has not made this kind of financial arrangement because we have been working smoothly and successfully with our existing customers using the current model.

Q: How will the additional $100 billion US investment be distributed over the next 3-5 years, and at what pace?

A: We do have plans, but the schedule and schedule mostly depend on market conditions and customer needs. If you want me to give you a firm timeline, we don't have one today, but we have plans and will move forward as quickly as possible. (Note: Our new plant and facilities in Taiwan will also be promoted as soon as possible, and the new Japanese plant will be put into operation as soon as possible. Because there is currently a huge gap between supply and demand, we are working hard to close this gap.)

Q: When can the computing part of HPC — (CPO/Co-packaged Optics, etc.) platform make a substantial contribution to revenue?

A: Production has already started, and it will gradually climb over time. I think AI data centers need to reduce power consumption and increase communication channel bandwidth, so the demand for this technology will continue to grow, and it will become a very important technology in the next few years.

Q: Under Agentic AI and CPU growth potential, what is the growth potential and visibility of different AI chips (GPU accelerators vs. CPU/XPU)?

A: I can't give very specific numbers, but I can share them — they all use TSMC, and they all use the same advanced process. We are working with customers to allocate wafer supplies to balance the CPU, GPU, and XPU ratios.

Q: In terms of advanced packaging, a 14x mask size roadmap was previously announced to achieve a larger AI package, and the development of CoWoS glass substrates was recently shown in Japan; how are new technologies such as glass cores (glass cores), glass substrates, and glass carrier boards progressing now?

A: CoWoS is still the mainstream today. We are developing alternatives to reduce costs, and we are also working with substrate suppliers to help customers get their products to market. The pooling (related production line) we announced a few quarters ago is under construction and will take about a year to mature, at which time it can be put into mass production with customers.

Q: Can you quantify the sales growth outlook for the next few years? Also, please break down the key drivers of this round of capital expenditure increases and demand this year — is it still mainly from cloud computing, or is it spreading to edge computing? Are there any factors that have increased the price of the equipment supply chain?

A: Because revenue corresponds to investment - we first anticipate demand, make an assessment, and then decide on capital expenditure. The next few years will be a very good business for TSMC; that's all I can say. (Key driving aspect: Everything related to AI.)

Q: Back-end packaging competition is intensifying (especially Intel EMT). Are you concerned that TSMC's overall foundry value from manufacturing to packaging will be eroded?

A: The front-end wafer business and the back-end business are two different things; they are not the same. Furthermore, our back-end production capacity is in short supply, and the gap is even larger, so I welcome competitors to provide my customers with some flexibility so that front-end wafers can be packaged, which in turn helps TSMC's wafer sales. That's our attitude.

Q: You mentioned that high-NA equipment is too expensive; how do customers view the stitching (stitching) challenge brought about by a smaller exposure field? Even as technology improves, will this slow down the adoption of high-NA?

A: You have a deep understanding of high-NA. Currently, its exposure field is only half the size, and we have taken this into consideration such as manufacturing costs. Whether we use it or not, the High-NA is a very good tool with great performance. TSMC has clearly stated that it will cooperate with ASML in an effort to make it more suitable for manufacturing in terms of cost and maturity. We always consider technology maturity and cost before deciding whether to adopt it.

Q: The market generally assumes that unconstrained demand of 3nm and below exceeds your supply capacity by about 30% to 50%. Is the actual gap larger?

A: We don't have specific numbers to share. The gap is huge.

Q: The technical forum mentioned that 2nm family production capacity grew by about 17% CAGR in '26—28, and N3 and N5 grew by about 25% CAGR. Are these assumptions still true today? Have there been any changes in the past quarter?

A: The picture we showed on the tech forum — the numbers are bigger now, so I'll say that's all.

Q: The capital expenditure of advanced packaging has always been packaged with testing, mask manufacturing, etc. in about 10% to 20% of the total capital expenditure. How much of this is actually invested in advanced packaging? How do you view the gap between packaging's share of revenue and capital expenditure? Should it be considered as a separate capital expenditure item?

A: We work very hard to ensure that capital expenditure figures are accurate, but we need to maintain flexibility between the front and back ends. Sometimes bottlenecks occur, and we invest more money to buy bottleneck equipment, sometimes at the front end, and sometimes at the back end. Over the long term, this ratio is roughly the level I've always shared. The backend is about 10% to 20%. This range itself is very wide. Honestly, as time goes by, some customers' products require more testing, and as demand for testing increases, we will invest more capital in testing machines, packaging, or other fields, so it is impossible to specifically split the amount invested in each field.

Q: Since the beginning of the year, TSMC has raised its capital expenditure guidelines to close to $10 billion. Where does the room for improvement mainly come from? Compared to 6 months ago, is it a change in CPU/accelerator, storage vendors, or back-end capacity expansion?

A: The most important reason is that demand continues to grow. We feel pressure from customers to expand production (more accurately, cooperate with TSMC). This is one of the main reasons. The second reason is inflation — now we buy equipment at inflated prices.

Q: The market focuses on advanced AI processes, but mature processes also seem to be experiencing strong demand recovery and some supply constraints. How do you view the supply and demand dynamics and pricing of mature processes (both AI spillover effects and heavy reliance on weak consumer demand)?

A: The mature manufacturing process covers many different segments, and only the AI-related parts are in short supply — the most important is power management IC (PMIC), because all AI data centers require extensive power management, which is a mature process technology, which is really lacking; in addition, there is also a sensor part because a large number of sensors are needed to collect environmental information for AI data center analysis. In addition to this, as you pointed out, demand for consumer products is beginning to weaken, and demand in other segments is not strong, nor can we talk about large-scale shortages.