Full Coverage: TSMC 1Q25 Financial Results
TSMC, the world’s semiconductor foundry service leader, reported first quarter 2025 revenue of NT$839.25 billion, representing a 41.6% year-over-year increase but a 3.4% sequential decrease in New Taiwan dollars (NTD), or a 5.1% decrease in US dollars, as business was impacted by typical smartphone seasonality. This decline was partially offset by robust growth in AI-related demand, which remains a key driver for the company.
The January 21 earthquake in Taiwan, along with subsequent aftershocks, led to wafer losses due to scrapped in-process inventory. Despite these disruptions, TSMC worked diligently to recover much of the lost production. As a result, the first quarter revenue was slightly above the midpoint of the company’s previous guidance range. Gross margin for the quarter decreased by 0.2 percentage points sequentially to 58.8%, primarily due to the earthquake’s impact and the beginning of margin dilution from overseas operations, though this was partially offset by ongoing cost improvement efforts.
Looking ahead, TSMC expects its second quarter gross margin to decrease by 80 basis points to 58% at the midpoint, mainly as the margin dilution from its Arizona fab becomes more pronounced. The impact from overseas fabs is expected to grow throughout the year as the company ramps up further in Kumamoto (Japan) and Arizona (USA), with a forecasted 2–3% margin dilution for the full year 2025 due to the fragmented globalization environment.
TSMC reiterated that a long-term gross profit margin of 53% or higher is achievable. The company also maintained its 2025 capital budget guidance at US$38–42 billion, with approximately 70% allocated for advanced process technologies, 10–20% for specialty technologies, and 10–20% for advanced packaging, testing, mask making, and other areas.
Regarding tariffs, TSMC acknowledged uncertainties and risks from potential tariff policies but noted it has not observed any changes in customer behavior to date. The company continues to expect full-year 2025 revenue growth to approach the mid-20% range in US dollar terms, and remains confident that revenue from AI accelerators will achieve a compound annual growth rate (CAGR) in the mid-40% range over the next five years, starting from 2024.
TSMC’s Arizona fab has already entered high-volume production using advanced process technology, with yields comparable to those in Taiwan. Construction of the second Arizona fab, which will utilize 3nm technology, is complete, and TSMC is accelerating its production schedule in response to strong AI-related demand. Plans for third and fourth Arizona fabs, which will use 2nm and A16 process technologies, are underway, with construction expected to begin later this year. Upon completion, around 30% of TSMC’s 2nm and more advanced capacity will be located in Arizona, establishing a leading-edge semiconductor manufacturing cluster in the US.
In Japan, TSMC’s first specialty technology fab in Kumamoto began production in late 2024 with strong yields, and the construction of a second specialty fab in Dresden, Germany, is scheduled, contingent on local infrastructure readiness. In Taiwan, TSMC plans to build 11 wafer manufacturing fabs and four advanced packaging facilities over the next several years, with volume production of N2 (2nm) expected to start in the second half of 2025 in both Hsinchu and Kaohsiung Science Parks.
TSMC expects the number of new tape-outs for 2nm technology in its first two years to exceed those of 3nm and 5nm nodes, driven by both smartphone and high-performance computing (HPC) applications. N2 is projected to deliver a 10–15% speed improvement at the same power, or a 20–30% power reduction at the same speed, and over 15% chip density improvement compared to N3E. N2 volume production remains on track for the second half of 2025, with a ramp profile similar to N3. The N2P extension will offer further performance and power benefits, with volume production scheduled for the second half of 2026. The A16 node, featuring Super Power Rail (SPR) technology, will provide an additional 8–10% speed improvement or 15–20% power reduction, and a 7–10% chip density gain, with production also set for the second half of 2026.
TSMC emphasized that it is not engaged in any joint venture, technology licensing, or technology transfer discussions with other companies.
In summary, despite short-term headwinds from smartphone seasonality and the January earthquake, TSMC delivered strong annual growth, driven by surging AI demand, and remains committed to aggressive investment in advanced technology and global capacity expansion.
Q&A Session Transcripts (edited):
Q: AI-related demand: there's been a lot of noise around CoWoS and order cuts and such. So, what is the strategy for TSMC, CoWoS still doubling this year? Is the demand still exceeding the supply? And how is the supply and demand? Still see demand exceeding supply by 2026?
A: The last time we talked about CoWoS, the demand is almost insane and much, much higher than we can prepare. And now it's a little bit better. I think we still need to build a lot of capacity to meet the demand. As I said, we have to double our CoWoS capacity, still fully loaded for 2026.
We work very hard to make sure that demand is not much, much higher than the capacity, and I believe that it will be more balanced next Year.
Q: From TSMC point of view, what do you think is required for more onshoring in the US? Can you share the perspective from the US government, or more directly, what customers are asking to do in terms of reshoring? Wendel also indicated that the margin dilution may be slightly bigger as we go along. So could you talk a little bit about how much of the value you can pass on to the customers as the expansion continues?
A: We have talked with US government officials, and the reason we are spending in Arizona, actually, let me say again, is all because of our customers’ request, and that's because of their very high demands. I announced it to in other occasions, say that very strong ai demand from us customers such as Apple, Nvidia, AMD, Qualcomm and Broadcom and so that we need to expand our capacity in the US and to support them, we talk with the US government and to ask for their help in getting the necessary permits so we can start the fab. And as a result, I would expect around 30% of our 2nm of capacity in Arizona, and that will also be an independent, leading-edge semiconductor manufacturing cluster.
We're in a very capital-intensive business, so we need to have a very high gross margin to earn a sustainable and healthy return, and that is why we set up our pricing strategy. Geographic manufacturing flexibility is an important part of our value proposition to the customers. We are already discussing this with our major customers, and the progress is so far so good.
Q: Is some of the recent rules and announcements, specifically the ban on H20, so his question is, how does this impact TSMC business? How does this impact on TSMC capacity planning and strategies?
A: No comment on specific product. However, we are mindful of the potential impact from all the recent tariff announcements, especially the potential impact to the end market demand will continue to watch it carefully. Having said that, we have not seen any change in our customers’ behavior so far, and so we stick to our forecast.
Q: Given that the capacity of mature nodes and 7nm are underutilized, are you concerned about slowing down expansions in Japan and Germany? And then number two, would you consider relocating equipment from Taiwan to overseas, rather than just building new fabs or Greenfield expansion?
A: Are we considering slowing down? The answer is, no, we execute our plan as scheduled. The reason is very simple, because this kind of mature node is a specialty technology in demand, which my competitor did not have the capacity or capability to support, so it's a free form. You mentioned the underloading of the mature node, and so again, I would emphasize no, we are not going to slow down our pen in Japan or in Germany. The second question is, how to do it? You have a good idea, but that is confidential information.
Q: With your total investment of 165 billion US dollars in Arizona, does TSMC believe its semiconductors will be exempt from these tariffs?
A: Tariff decisions are the government’s responsibility to decide. And as a private company, we are fully respectful of this, but we are not getting involved.
Q: Second quarter guidance, 13% at the midpoint in US dollar terms, Q on Q is very strong. So I wonder, are we seeing already some tariff pulling impact, or this part of that guidance? How come the dilution in the latter period widens to 3-4%? What are the drivers or reasons behind it?
A: We haven't seen any changes in customers’ behavior so far, our second quarter growth is driven mainly by strong demand for the 3nm and 5nm technologies, underpinned by the growth in the HPC platform. As I said, we haven't seen any changes in customer behavior in terms of pulling or due to the cost of tariffs.
The widening of dilution on the gross margin in this later part of the five-year period is mainly from inflation in cost and also potential tariff-related cost increases.
Q: CC has mentioned, all we want is fair treatment. So, what do you mean by fair or fairness? The R&D center in Arizona, what will be the purpose or the focus, and will it be involved at some point in ramping up new technologies? Will the R&D center over Arizona in the mid-to-long term, and can it also focus on things like new node development or pathfinding opportunities for long-term research?
A: It's very simple. If anybody gets the subsidy or the incentive, everybody should get the same, either we get all or we get zero, all right? So that's what we call fair. So again, I would like to assure you that we are, we are very competitive in either conditions.
TSMC is a firm that never stays stagnant. We always continue to improve it. And we need to establish a major R&D center in Arizona with about 1000 engineers. That's a big amount, but the focus will be to support our manufacturing cluster, improve its technology, and allow it to operate independently.
The first purpose is to that Arizona fab can operate in independently. But of course, we have done, and we are doing it right now to some kind of path finding, exploratory work and cooperate with universities.
Q: About the Gigafab cluster in Arizona. What’s the timeline of expansion, particularly given the strong AI-related demand? Can we pull in the timing for both the second fab and the third fab, and also, can we, at the same time, start the production of the third and fourth fab simultaneously?
A: Since the second phase originally is planning for production in 2028 so Now should we assume it to be from maybe 2027 or even first of 2027 and for the third phase, since you are building the facade this year, so all the production start may be one year ahead, versus the original timeline of 2030?
A: Demand is strong. We have to really to speed it up, and building all the fab definitely will depend on our customers demand, of course. Yes, we are speeding up. How fast the second fab, as you said, it should be pulling, and this one, we are working hard to pull in at least a couple of quarters. That's at least on the third fab. The whole thing is also being constrained by the labor shortage of Arizona, and we need to get all the permits, everything, etc. So I cannot give you a very definitely date yet, but we are going to update you probably in the next quarter or one quarter after that.
Q: Overseas expansion in both pricing and margin, given the strong demand in terms of pricing, can we reflect even greater value to our customers and therefore, and also her question is given that the dilution from overseas will widen to three to 4% in the latter stages of the five-year period. What is your underlying wafer price assumption behind this?
A: These two things are actually go together as we said, reflecting our value is a continuous and ongoing process, and we, because of our business nature, we need very high gross margin to earn a sustainable, healthy return. Now, geographic manufacturing flexibility is an important part of our value proposition to the customers. Therefore, we are already discussing this with our major customers, and the progress is so far so good. Now, at the same time, the margin dilution from the overseas fabs, the additional dilutions come from the cost inflation as well as potential cost increases from the tariff policies. Of course, with that, we also want to reflect the value, and therefore the discussions with the customers are in continuous.
Q: Although there's a ban in China on certain AI chips or products that we reiterated our AI accelerator, growth will double this year. So the assumption is that implies a strong, non-China AI related demand. And wondering, what is the mechanics? Or can we comment beyond that, behind that?
A: Three months ago, we just cannot supply enough wafers to our customer. And now it's a little bit balanced, but still, the demand is very strong, and you are right. Other than China, the demand is still very strong, especially in the US. And so we are confident that we are going to double our AI's revenue this year.
The growth guidance in second quarter was primarily due to the demand from our three nanometer and five nanometer technologies, underpinned by the demand from the HPC platform.
Q: Why does TSMC management not adopt buyback a buyback framework, particularly with the strength of the cash position on your on your balance sheet at the moment?
A: We've done studies long time ago, and we continue to revisit that. We also talked to investors. Our conclusion stays the same, the sustainable and steadily increasing dividends is a better way of returning cash to the shareholders, so we're maintaining the policy.
Q: Do you see any changes in the chip design, particularly moving to chiplets with N3? What about the role of things like co-package optics and panel level packaging? Will TSMC continues to use advanced packaging technologies like, CoWoS or SOIC in Taiwan first, or is this also part of the plan for the expansion in Arizona?
A: Yes, our customers, they continue to use leading edge technology, and they also adopt the advanced packaging technologies more and more. This year is probably most of CoWoS-S, and then next year, CoWoS-L and etc, and we can see that customers start to picking up the SOIC and the more advanced packaging technologies. As for what we call these panel levels of packaging, we are aggressively developing it, and today it still is at feasibility study stage. Too early to say it will not be in Taiwan or in the US, but most likely in Taiwan first, we ramp it up and then bring it to the US later.
Q: About the capacity allocation. How do you allocate between Taiwan and the US? Is it duplicative or extra capacity? And then very specific, cc had mentioned that into and more advanced capacity, around 30% will be in Arizona once we scale up to the cluster, will that be kind of the percentage for the leading node in the future?
A: Right now, we plan a six fabs in Arizona, and in that six fabs, the two nanometer will be a major node, and that's what I say, 30% will be there as time goes by, after the 2 nanometer will be 1.4 and 1.0 that had not been discussed yet.
Q: In the near term, what is your visibility into the second half business outlook, and then also, how do we see the demand for end to progressing this year and also next year?
A: We're only at second quarter, so I think it's too early to talk about the second half. We did mention that the uncertainties and risks from tariffs exist, and we might get a better picture in the next few months, so we can probably update you in the next earnings call.
So far, (the demand for 2nm) is very strong, as we said, all the new tape out customers, the number of the tape out is exceeding that what we expected. And as we said, the number of the new tape out is much higher than the three nanometer in five nanometer in the same period of time.
Q: About first specialty technology fab in Japan. What is the capacity installment for this specialty technology fab, and also the revenue contribution from JASM?
A: The capacity for the fab will be 40k when it's ramped up. The revenue for this year compared to the whole company is really not significant at this moment.