Oct 15, 2026 earnings call — Leading-edge node ramp and capacity expansion
Company guided fourth-quarter revenue to be between $52.3 billion and $54.3 billion.
Revenue $46.1B (est $45.7B) · EPS $4.60 (est $4.45) · Next-quarter revenue outlook $52.3B–$54.3B
Key quotes
"Compared to second quarter, our third quarter gross margin decreased by 130 basis points sequentially to 66.4%, primarily due to the margin dilution from the initial ramp-up of our 2 nm technology."
"We are building 13 leading-edge and advanced packaging fab in Taiwan over the next several years, and another five to six fabs overseas."
"We observe increasing AI demand fueled by the rising adoption of agentic AI as we transition from chatbot to AI assistance."
Analyst Q&A 37
AI chips demand and supply
Q (Laura Chen · Citi): Laura asked how TSMC sees AI revenue CAGR and how TSMC reconciles customer capacity needs with limits like data center power, funding, HBM, advanced packaging and laser tightness. She asked how TSMC would avoid overbuilding capacity.
A: No management answer appears in this exchange; Laura finished with her first question and asked for follow-up.
Capacity planning and demand gap
Q (Laura Chen · Citi): Laura asked how TSMC assesses AI demand vs risks (power, components) and decides appropriate CapEx and capacity to avoid overbuilding.
A: Management said they have a sophisticated capacity planning process and consider many factors, including CSPs, CapEx, data center build-out, finance, power and component supply. They judged demand is still higher than supply and are working hard to narrow the gap.
"We make our own judgment, and then based on our assessment the demand, we still think the demand is higher than the supply, and we are working very hard to narrow the gap." — C.C. Wei (CEO·CHAIR)
AI accelerator CAGR reference
Q (Laura Chen · Citi): Jeff summarized that part of Laura's question was asking about the AI accelerator CAGR update versus prior guidance. Laura confirmed the reference.
A: Jeff restated that part of the question referred to their prior AI accelerator CAGR guidance. There is no new management update given in this exchange.
AI accelerator CAGR number update
Q (Laura Chen · Citi): Laura asked if TSMC has any update to its prior guidance of AI accelerator CAGR growing mid to high 50s from 2024 to 2029.
A: Jeff reiterated the prior phrasing of the guidance as mid to high 50s percentage from 2024 to 2029. Management did not provide a revised CAGR number in this exchange.
High NA EUV and profitability
Q (Laura Chen · Citi): Laura asked whether adopting High NA EUV around 2030 will improve productivity and accelerate long-term revenue growth, and how to think about pricing power, long-term gross margin, and capital efficiency.
A: Management said they continue to improve long-term productivity and that High NA is a very good tool, planned for use in 2030. They said there is no ceiling for TSMC's profitability but dramatic increases will not happen overnight, and they work with customers to create value.
"Yes, for the long-term productivity, we continue to improve it, and High NA definitely is a very good tool." — C.C. Wei (CEO·CHAIR)
Wafer capacity growth outlook
Q (Gokul Hariharan · JPMorgan): Gokul asked whether wafer capacity growth could reach or exceed 10% annually given the accelerated build-out and current shortages, and what TSMC plans for wafer capacity growth in coming years.
A: No management answer appears in this exchange; Gokul asked about next-year tightness and planned wafer capacity growth.
U.S. expansion and customer talks
Q (Gokul Hariharan · JPMorgan): Gokul asked about TSMC's accelerated overseas expansion, especially in the U.S., reports of a possible Texas cluster, and any engagement with SpaceX or Tesla and the cooperation framework.
A: No management answer appears in this exchange; Gokul asked whether any U.S. expansion details or customer discussions could be shared and how they would be structured.
TSMC-run fabs and partnership terms
Q (Gokul Hariharan · JPMorgan): Gokul asked if the overseas fabs would be TSMC-run (not part of other projects) and what principles guide such overseas expansions, comparing to JASM and ESMC.
A: No management answer appears in this exchange; the analyst sought clarification that overseas fabs would be TSMC-run and asked about guiding principles.
Secondary wafer market trend
Q (Krish Sankar · TD Cowen): Krish asked whether a secondary wafer market (customers passing allocations to others) is a small, customer-specific activity or an emerging trend that TSMC allows.
A: No management answer appears in this exchange; Krish asked about prevalence and TSMC's stance on wafer allocation resale.
High NA timing vs competitors
Q (Krish Sankar · TD Cowen): Krish asked if adopting High NA EUV around 2030, while competitors may already be using it earlier, is a disadvantage for TSMC.
A: No management answer appears in this exchange; Krish raised concern about competitive timing for High NA adoption.
Follow-up acknowledgment
Q (Krish Sankar · TD Cowen): Krish thanked management and had no new question.
A: No management answer appears; the exchange is an acknowledgment from the analyst.
No new content
Q (Krish Sankar · TD Cowen): No substantive question in this short exchange.
A: No management answer appears; brief exchange with no new content.
IDM customer insourcing risk
Q (Charlie Chan · Morgan Stanley): Charlie asked how TSMC will handle a big IDM customer expanding capacity and possibly insourcing more, and whether prior comments that A14 could be larger than N2 assume that IDM outsources more server CPU at A14.
A: No management answer appears in this exchange; Charlie asked whether the A14 sizing assumption included potential outsourcing of server CPUs by that IDM.
Clarify A14 vs N2 assumption
Q (Charlie Chan · Morgan Stanley): Charlie asked to confirm whether A14 being a bigger node than N2 assumes that the big IDM will outsource more server CPU at A14.
A: No management answer appears in this exchange; the analyst sought to confirm understanding about outsourcing assumptions.
IDM outsourcing answer needed
Q (Charlie Chan · Morgan Stanley): Charlie noted TSMC would likely need to ask that IDM directly about its plans and had no further questions in this exchange.
A: No management answer appears; Jeff suggested the analyst ask the IDM directly about its outsourcing plans.
Advanced packaging backup options
Q (Charlie Chan · Morgan Stanley): Charlie asked whether SoW 3.5D and CoWoS can support customers with very large reticle chips and whether TSMC would support customers if they couldn't get sufficient demand from a competitor's EMIB-T solution.
A: No management answer appears in this exchange; Charlie asked if TSMC could be a backup and support CoWoS for customers who initially chose another supplier's solution.
Backup for competitor packaging
Q (Charlie Chan · Morgan Stanley): Jeff paraphrased Charlie's question asking if TSMC would be willing to act as a backup if a competitor cannot deliver their packaging solution.
A: No management answer appears in this exchange; the analyst reiterated the backup-support question for advanced packaging.
End of Charlie's questions
Q (Charlie Chan · Morgan Stanley): Charlie thanked the panel and ended his turn.
A: No management answer appears; this exchange is a closing remark from the analyst.
CapEx beyond 2026 and tool price
Q (Sunny Lin · UBS): Sunny asked how to think about CapEx beyond 2026 — whether CapEx will grow in line with sales or faster — and whether higher tool prices will drive higher CapEx in 2027.
A: No management answer appears in this exchange; Sunny asked whether tool price inflation should be factored into 2027 CapEx expectations.
Clarify tool price question
Q (Sunny Lin · UBS): Jeff asked Sunny to repeat the last part; Sunny clarified she meant tool price inflation as a variable for 2027 CapEx.
A: No management answer appears; this is clarification of the tool price question.
CapEx size question
Q (Sunny Lin · UBS): Sunny confirmed she was asking whether 2027 CapEx could be much higher than this year, given price and expansion.
A: No management answer appears in this exchange; the analyst reiterated the CapEx magnitude question.
Overseas margin dilution and timing
Q (Sunny Lin · UBS): Sunny followed up asking if the overseas margin dilution of 3%-4% (from earlier remarks) implies a faster expansion of Arizona phase two or pulling in other phases of U.S. fabs.
A: Wendell said TSMC is accelerating capacity everywhere, including the U.S., and reiterated that the margin dilution from overseas fabs in the latter part of the five-year period will be 3 percentage point to 4 percentage point. He confirmed that remains the case.
Follow-up interrupted
Q (Sunny Lin · UBS): Sunny attempted another follow-up but was told her turn had reached two questions and the call would move on.
A: No substantive management answer appears; the moderator noted limits to questions and moved on.
Question continuation attempt
Q (Sunny Lin · UBS): Sunny asked to follow up on a point but was cut off due to time and question limits.
A: No management answer appears in this short exchange; it records the analyst trying to continue but being stopped.
Question limit enforcement
Q (Sunny Lin · UBS): Jeff noted Sunny had already asked two questions and the call would move on to accommodate others.
A: No management answer appears; the moderator enforced the two-question limit.
AI demand mix shift and packaging
Q (Haas Liu · BofA): Haas asked whether customer demand is shifting toward more efficient compute solutions like CPUs and networking switches as agentic AI grows, and how that affects TSMC's manufacturing and packaging opportunity.
A: No management answer appears in this exchange; Haas asked whether AI demand is moving from accelerators to CPUs/switches and the packaging implications.
Yield learning curve for dual tracks
Q (Haas Liu · BofA): Haas asked whether operating two process architectures (conventional power and Super Power Rail) at the same time will make yield learning and manufacturing more challenging and affect gross margins.
A: No management answer appears in this exchange; Haas asked about yield learning, complexity, and margin implications for dual technology tracks.
Clarify margin implication question
Q (Haas Liu · BofA): Jeff restated Haas' second question about yield learning, ramp complexity and margin impact for new nodes supporting diverse applications.
A: No management answer appears; this is a restatement to clarify the analyst's point.
Power grid as constraint
Q (Arthur Lai · Macquarie): Arthur asked whether power grid constraints have ever caused TSMC to reduce a customer's forecast and whether tighter grids could accelerate customer adoption of Super Power Rail.
A: No management answer appears in this exchange; Arthur asked if power limits have forced forecast cuts and whether they push customers to Super Power Rail.
Clarify AI definition and CPU inclusion
Q (Arthur Lai · Macquarie): Arthur asked if the previous AI accelerator growth framework includes CPUs (agentic AI) and when TSMC would refresh its AI demand growth numbers if CPU forecasts increase.
A: C.C. Wei is indicated as responding, but the transcript records no substantive content beyond an acknowledgment. The analyst later thanked C.C., implying some reply but no text is provided.
When to revise AI numbers
Q (Arthur Lai · Macquarie): Arthur asked when TSMC would revise AI demand forecasts if clients increased CPU forecasts significantly.
A: Jeff later clarified that the AI accelerator definition is a narrow one and that CPUs are not part of that definition, though CPUs are part of TSMC's business. No timing for a revision was given in this exchange.
Clarify CPU inclusion again
Q (Arthur Lai · Macquarie): Arthur probed again about whether the mid to high 50s framework included CPUs and when a refresh would occur.
A: Management reiterated that CPUs are not part of the AI accelerator narrow definition, and they did not give a timetable for any refresh of the AI demand number in this exchange.
Acknowledgement of definition
Q (Arthur Lai · Macquarie): Arthur acknowledged management's clarification that CPUs are outside the narrow AI accelerator definition.
A: No further management answer appears; the analyst acknowledged the clarification.
Pricing increases and scope
Q (Sahej Singh · Stifel Financial): Sahej asked about reported price increases across mature and advanced nodes and whether pricing moves apply to existing contracted book or mainly to incremental capacity above commitments.
A: No management answer appears in this exchange; Sahej asked whether price increases apply to current contracts or only to incremental capacity.
Clarify pricing question
Q (Sahej Singh · Stifel Financial): Jeff confirmed Sahej was asking specifically about pricing application to contracted book versus incremental capacity.
A: No management answer appears; this is a moderator clarification of the pricing question.
Outsourcing specialty components strategy
Q (Sahej Singh · Stifel Financial): Sahej asked about TSMC's strategy in procuring specialty items (like interposers), referencing a GlobalFoundries announcement, and how TSMC decides to partner vs make in-house.
A: No management answer appears in this exchange; Sahej asked for color on the approach to strategic sourcing versus in-house production for specialty items.
Closing remarks
Q (Sahej Singh · Stifel Financial): Sahej thanked the team and closed his questions.
A: No management answer appears; this exchange records the analyst closing his turn.