Today's event is being webcast live through TSMC's website at www.tsmc.com, where you can also download the earnings release materials. Wendell Huang, will summarize our operations in the second quarter 2026, followed by our guidance for the third quarter 2026. Wendell Huang, for the summary of operations and the current quarter guidance. After that, I will provide the guidance for the third quarter of 2026.

2 nm process technology contributed 3% of wafer revenue in the second quarter. Advanced technology, defined as 7 nm and below, accounted for 77% of wafer revenue. HPC increased 20% quarter-over-quarter to account for 66% of our second quarter revenue. Based on the current business outlook, we expect our third quarter revenue to be between $44.6 billion and $45.8 billion, which represents a 12% sequential increase or a 37% year-over-year increase at the midpoint.

Based on the exchange rate assumption of $1 to TWD 32, gross margin is expected to be between 65% and 67%. We have just guided our third quarter gross margin to decrease by 1.7 percentage point to 66% at the midpoint. Primarily as we expect the steep ramp-up of our 2 nm technology to dilute our gross margin by about 3 percentage points-4 percentage points. This dilution is expected to be partially offset by very strong demand for our leading-edge technologies and continued cost improvement efforts, including productivity gains and across-node capacity optimization.

What went well
  • Second-quarter revenue reached $40.2 billion in U.S. dollar terms, at the high end of guidance, driven by strong demand for leading-edge process technologies.
  • Gross margin rose 150 basis points sequentially to 67.7%, slightly ahead of guidance, on cost-improvement efforts and higher capacity utilization partially offset by overseas-fab dilution.
  • HPC revenue grew 20% quarter-over-quarter to 66% of total revenue, and advanced technologies (7nm and below) accounted for 77% of wafer revenue.
  • Management raised the full-year 2026 revenue growth outlook to slightly above 40% year-over-year (USD) and guided Q3 revenue to $44.6-$45.8 billion, up about 37% year-over-year at the midpoint.
  • TSMC raised its 2026 capital budget to $60-$64 billion (from $52-$56 billion in January) and announced an additional $100 billion investment in Arizona (roughly four more fabs), reflecting very high conviction in the multi-year AI megatrend.
  • The A14 node is on track and ahead of schedule on tape-outs, showing ~90% device performance and ~90% 256Mb SRAM yield, with a 10-15% speed or 25-30% power benefit and ~20% density gain versus N2; the 2026 dividend rises 33% to TWD 24 per share.
What went wrong
  • Q3 2026 gross margin is guided down 1.7 points to 66% at the midpoint, as the steep N2 ramp is expected to dilute gross margin by about 3-4 points.
  • The N2 ramp is expected to dilute second-half gross margin by 3-4 points, and overseas-fab dilution is forecast at 2-3% in early stages widening to 3-4% in later stages.
  • Consumer and price-sensitive end-market segments are being challenged by rising component prices and macroeconomic uncertainty, prompting prudent business planning.
  • Smartphone revenue declined 4% quarter-over-quarter, and mature-node demand outside AI-related areas (power-management IC, image sensors) remains soft on weak consumer demand.
  • Inventory days rose seven days to 87 days, primarily due to the N2 technology ramp, and packaging capacity remains so tight that it is currently limiting customer growth.

Guidance Changes

MetricPeriodCurrent guidance
Q3 2026 revenueQ3 2026$44.6B-$45.8B (+12% QoQ, +37% YoY at midpoint)
Q3 2026 gross marginQ3 202665%-67% (66% midpoint; -1.7 pts) at $1=TWD 32
Q3 2026 operating marginQ3 202656%-58%
Full-year 2026 revenue growth (USD)FY2026slightly above 40% year-over-year
2026 capital budgetFY2026$60B-$64B (70%-80% advanced, ~10% specialty, 10%-20% packaging/testing/mask)
Cash dividend per shareFY2026TWD 24 (up 33% YoY), with continued increases in 2027
N2 gross-margin dilution2H 2026~3-4 points
Multi-year AI accelerator CAGR5-year (from Jan)stronger than before (no new number given)

Performance Breakdown

MetricYoYNote
Q2 revenue (USD) $40.2B (high end of guidance) Strong demand for leading-edge process technologies.
Gross margin 67.7% (+150 bps QoQ) Cost-improvement efforts and higher capacity utilization, partly offset by overseas-fab dilution.
HPC platform revenue +20% QoQ (66% of revenue) Robust AI-driven demand for leading-edge silicon.
Smartphone platform revenue -4% QoQ (22% of revenue) Seasonal/mix softness relative to HPC strength.
Automotive platform revenue +15% QoQ (4% of revenue) Recovery in automotive demand.
Advanced technology (7nm and below) 77% of wafer revenue 5nm 33%, 3nm 30%, 7nm 11%, 2nm 3% of wafer revenue as N2 begins ramping.
Cash from operations TWD 783B Strong profitability; CapEx of TWD 496B ($15.7B) and TWD 156B of dividends distributed, cash balance up TWD 99B to TWD 3.1T.
Inventory days 87 days (+7) Ramp of N2 technology.

Earnings Call Themes & Trends

TopicPrevious mentionCurrent periodTrend
AI megatrend / demand convictionStrong multi-year AI demandAI-related demand extremely robust; CSP customers giving strong positive signals; demand seen very strong through ~2029-2030 with a very large demand-supply gap; conviction described as very high, driving the CapEx and revenue-outlook increases.
Capital expenditure ramp$52B-$56B (Jan); next-3-year CapEx significantly higher than prior 3 yearsRaised to $60B-$64B for 2026; next-3-year CapEx now expected 'even more significantly higher'; increase driven by rising demand (all AI-related) plus tool-price inflation; no formal 3-year CapEx number provided.
Arizona / global capacity expansionExisting Arizona and overseas fabsAdditional $100 billion Arizona investment (~4 more front- and back-end fabs, total ~$265B); 13 leading-edge and advanced-packaging fabs being built in Taiwan; three new N3 fabs (Taiwan, Arizona, Japan) plus 5nm-to-3nm tool conversion; schedule flexed to market and customer demand, being expedited.
Agentic AI / CPU resurgenceEmergence of agentic AI is reviving CPU demand in AI data centers alongside accelerators; positive for TSMC since x86, Arm and RISC-V CPUs are almost all TSMC customers; wafer allocation balanced across CPU/GPU/XPU.
A14 technology roadmapN2 leading edgeA14 (2nd-gen nanosheet) on track and ahead of schedule (pre-production 2027, volume 2028) with strong smartphone and HPC/AI interest; A13 (~6% die-area saving via 97% optical shrink) and A12 (Super Power Rail) both targeting 2029 volume; A14 family expected to be larger and longer-lasting than N2.
Competition (Samsung, Intel, packaging)No shortcuts for newcomersAcknowledged Samsung's strong memory profits and Intel's U.S. policy support (TSMC also has support) but reiterated technology, manufacturing and customer trust as the durable moat; welcomed alternative advanced-packaging options like EMIB-T given tight back-end capacity, since they can help TSMC's larger front-end wafer business.
Mature-node strategyFocus on higher-value-added specialty segmentsOnly AI-related mature nodes (power-management IC, image sensors) are in shortage; commodity/consumer mature-node demand is soft; TSMC continues to add higher-value mature capacity (JASM Japan for CIS, ESMC Germany for auto/industrial).
Pricing / profitability philosophyWants gross margin 'higher the better' but as a trusted partner will not impose extreme increases; targets margins sufficient for sustainable expansion; C.C. Wei jokingly envied memory peers' ~86% margins, saying ~68% would make him happy.

Q&A Summary

Sunny Lin (UBS) asked for a three-year CapEx outlook and the plan/timeline for the additional $100B Arizona investment.
Huang declined a specific number but said the next three years' CapEx will be even more significantly higher than the past three, given very strong AI conviction; Wei said the Arizona schedule depends on market conditions and would involve roughly four more front- and back-end fabs.
Charlie Chan (Morgan Stanley) asked about competition from Samsung and Intel and whether the multi-year AI CAGR guidance would be raised.
Wei acknowledged Samsung's memory profits and Intel's U.S. support but stressed technology, manufacturing and customer trust as fundamentals with no shortcuts ('not buying milk from 7-Eleven'); on the AI CAGR he said it is 'stronger and stronger' than the prior mid-to-high 50% but gave no new number.
Arthur (Macquarie) asked how TSMC would respond to advanced-packaging competition such as EMIB-T gaining traction.
Wei said packaging capacity is so tight it is limiting customer growth, so additional market flexibility is welcome and would help TSMC's larger front-end wafer business; TSMC's number-one priority is supporting customer success.
Gokul Hariharan (JPMorgan) asked about the philosophy of capacity expansion (including competitive pressure and data-center/power constraints) and long-term foundry profitability/pricing.
Wei said competition is always a factor and TSMC uses top-down/bottom-up judgment on customer input ('all the truths together is not the truth'), while checking AI data-center build progress so chips are not stockpiled; on pricing he emphasized being a trustworthy partner that earns its value without extreme increases, envying memory's ~86% margins while being happy with ~68%.
Jim Fontanelli (Arete) asked about customer-concentration risk and whether TSMC would finance or invest in end customers.
Wei said concentration is not a concern given many new AI-industry players are growing fast, and that TSMC does not currently make such financial arrangements because the current customer model works smoothly.
Mehdi Hosseini (SIG) asked for the timeframe of the $100B U.S. investment and when the COUPE (co-packaged optics) platform contributes materially.
Wei said there is a plan but the schedule depends on market and customer demand and is being expedited as fast as possible (as in Taiwan and Japan); COUPE is entering production now and will ramp as AI data centers need lower power and higher bandwidth, becoming important over the next few years.
Laura Chen (Citibank) asked about growth potential across GPU/accelerator versus CPU and the progress of glass substrate/core advanced packaging.
Wei said all of them (CPU, GPU, XPU) are at TSMC on the same leading-edge nodes with wafer supply balanced to demand; CoWoS remains the majority, with a glass-substrate pilot line about another year from production maturity.
Haas Liu (Bank of America) asked to quantify multi-year revenue growth, the demand drivers behind the CapEx raise, and whether back-end competition (EMIB-T) could cannibalize front-end value.
Wei said next few years will be very good business (declining a number), the drivers are all AI-related, and front-end wafer and back-end packaging are two different businesses, so back-end competition does not threaten front-end logic and TSMC welcomes the added packaging flexibility.
Robert Sanders (Deutsche Bank) asked about High-NA adoption/die-stitching challenges and whether unconstrained 3nm-and-below demand exceeds supply by more than 30%-50%.
Wei said High-NA is a high-performance tool but TSMC weighs cost and maturity (working with ASML), and declined to quantify the demand-supply gap other than to say it is 'very big.'
Evelyn Yu (Goldman) asked whether prior symposium capacity-CAGR figures (2nm ~70%, N3+ ~25%) still hold and whether TSMC should break out advanced-packaging CapEx separately.
Wei indicated the capacity numbers are now 'bigger,' and said CapEx is deliberately kept flexible between front-end and back-end (long-term back-end ~10%-20%), so a separate packaging line item is impractical as spend shifts to wherever the bottleneck is.
Felix Pan (KGI) asked what is driving the ~$10B year-to-date CapEx increase and about mature-node supply/demand and pricing.
Wei cited continually increasing customer demand plus tool-price inflation as the main reasons; on mature nodes only AI-related segments (power-management IC and sensors) are in shortage, while consumer and other commodity segments remain weak.

More on Taiwan Semiconductor Manufacturing Co Ltd

Reported 2026-07-16 · figures from the Taiwan Semiconductor Manufacturing Co Ltd Q2 2026 earnings call.

See how VectorShift works for your firm

Request Demo