NT$1.49tn Already Booked, 65–67% Margin Guided (TSMC Earnings Preview, 15 October 2026): What to Expect From Q3 — and Why Margin and Capex Decide It

TSMC reports third-quarter 2026 results on Thursday 15 October at 14:00 Taiwan time (06:00 UTC), and the headline number is already public: monthly filings put quarterly revenue at a record of about NT$1.49tn, above the top of the company's own guidance. What is still unknown is how much of that revenue became profit while the 2-nanometer ramp weighs on margins, and what TSMC guides for the fourth quarter and for capital spending. Those are the figures that reach other chip stocks and the indices.

When TSMC reports, and why the timing matters

The earnings conference is scheduled for Thursday 15 October 2026 at 14:00 Taiwan time, according to TSMC's investor relations calendar. In other time zones that is 06:00 UTC, 07:00 in London and 02:00 in New York. TSMC observed a quiet period from 5 to 14 October, so the last word from management on this quarter was the July call.

The timing means the reaction comes in a fixed order. The Taipei market closes at 13:30 local time, before the call, so the Taiwan-listed shares first trade on the results on Friday. The New York ADRs react in US pre-market trading on Thursday. Before Taipei reopens, two other markets will already have priced the news: European chip-equipment makers trading in the European morning, and US futures overnight.

What is already known: the revenue is in

TSMC is unusual among mega-caps because it publishes revenue every month. On 8 October it reported September revenue of NT$511.86bn, up 54.6% from a year earlier and 0.6% below August's record of NT$514.81bn, according to Focus Taiwan. Adding up the three months gives a quarter of about NT$1.49tn.

The US-dollar figure is about 2% above the top of the guided range and about 3.3% above its midpoint. The guidance came from TSMC's 16 July results release and assumed an exchange rate of NT$32 to the dollar. The Reuters conversion, at about NT$31.9, is close to that assumption, so the currency did little to flatter or hurt the quarter. Analysts quoted in the coverage attributed the beat to strong demand for AI chips and to smartphone makers building inventory ahead of new launches.

The margin question: 2nm dilution against operating leverage

TSMC guided the gross margin down to 65–67% from 67.7%, even while guiding revenue up 12% sequentially at the midpoint. The July release explains why: chief financial officer Wendell Huang said the third quarter would be supported by "the steep ramp-up of our 2-nanometer technology." In the second quarter, 2nm was only 3% of wafer revenue. A new node is expensive to bring up and does not yet run at mature yields or full utilisation, so its early wafers earn less than wafers from the established 3nm and 5nm lines. The more 2nm production grows, the more it dilutes the average margin.

Operating leverage works in the other direction. Most of a foundry's cost base is fixed: depreciation, staff and fab overheads. When revenue comes in about NT$30bn above the estimates, much of the extra revenue becomes gross profit. That is why the beat makes the top of the margin range plausible. It does not guarantee it.

On NT$1.49tn of revenue, the guided range translates into these figures:

• At 65% gross margin: about NT$969bn of gross profit. At a 56% operating margin, about NT$834bn of operating income.
• At 66% / 57% (midpoints): about NT$983bn and NT$849bn.
• At 67% / 58%: about NT$998bn and NT$864bn.

For comparison, the second quarter produced NT$860.3bn of gross profit and NT$766.6bn of operating income. Even the bottom of the range would be a large step up in absolute profit. What markets will judge is where in the range the quarter lands, because that shows whether the 2nm ramp is costing more or less than planned.

There is one complication in the net profit line. The second quarter's pre-tax income of NT$862.4bn was well above its operating income of NT$766.6bn, a gap of about NT$96bn from non-operating items. Non-operating income is not part of the guidance, so net profit can differ from the LSEG estimate for reasons that have nothing to do with how the fabs performed. Gross margin, which TSMC guides directly, is the cleaner test.

The three scenarios and what each would mean

Margin at or above 67%, firm Q4 guide. This would mean operating leverage outweighed the cost of the 2nm ramp. A rising fourth-quarter revenue guide would confirm that leading-edge orders from AI-chip and smartphone customers have not slowed. The read-through would go to the customers whose orders fill those fabs, including Nvidia and Apple, and to the suppliers of the tools.

Margin inside the range, guide in line. This is the case already reflected in the estimates. With revenue known and the shares up about 65% this year before the release, according to the Reuters coverage, a result that only matches the estimates offers little new. In that case the commentary would decide the reaction: 2nm yields, advanced-packaging capacity, overseas fab costs and any change to the full-year growth outlook, which was raised in July to just above 40% in US-dollar terms.

Margin below 65%, or a cautious guide. A margin below the range would mean the 2nm ramp, the overseas fabs or other costs absorbed more than the revenue beat added. A cautious fourth-quarter guide would be more important than a soft margin, because it would be the first sign from the supplier that order growth at the leading edge is slowing. That signal reaches every company whose valuation assumes the AI build-out continues at its current pace.

Capex: the line that reaches the equipment makers

In July TSMC raised its 2026 capital spending plan to US$60–64bn from US$52–56bn. For a foundry, capex is not an abstraction. It becomes orders for lithography, deposition, etch and inspection tools, which make up revenue for equipment makers in the Netherlands, the United States and Japan.

That makes the capex commentary on 15 October as much about other companies as about TSMC. If TSMC confirms the range or raises it, equipment makers' order books for 2027 look more secure. If it trims it, the equipment makers absorb the cut directly. Our coverage of SpaceX's US$40bn of debt for Nvidia chips follows the same chain from its other end: money committed to AI computing becomes chip orders, chip orders become wafer demand, and wafer demand becomes capacity spending.

How it reaches US500, NAS100 and the majors

TSM is an ADR on the New York Stock Exchange and belongs to neither the S&P 500 nor the Nasdaq-100. Its effect on those indices comes through read-through. A small number of mega-cap technology companies now account for most of a US500 or NAS100 move, and several of them design chips that TSMC manufactures. When the main manufacturer guides on demand, the market updates its view of the designers. We tracked that concentration effect in Nvidia's August results, when technology was the only sector to rise and the S&P 500 still closed 0.72% higher.

The timing adds a rates layer. US September CPI is released the day before, and our CPI preview sets out what is expected. Long-duration growth stocks are among the assets most sensitive to the discount rate. A hot inflation print followed by a strong TSMC report means the two are pulling in opposite directions, and how the market weighs one against the other is part of the reaction.

The currency channel is thin. The Taiwan dollar is not one of the eight currencies the meter scores. The link to the majors runs through broad risk appetite, which is one of the five factors, and through Japan, where chip-equipment makers are significant index weights. One company's quarter rarely moves the dollar or the yen by itself. A guide that changes how the market views the AI investment cycle as a whole can. The about page explains how the factors are framed.

What would change the picture

What has to be true: for this cycle's margins to hold through the 2nm ramp, leading-edge demand has to keep fabs full while yields on the new node improve. Our note on mature-node chip prices shows the other side: the AI build-out is tightening older processes too, not only the leading edge.

What is already priced: a revenue beat, since it was disclosed a week before the call, and a large rise in profit. The LSEG estimate already assumes about 64% growth.

What to watch on 15 October: - The gross margin against 65–67%. It is the cleanest measure of what the 2nm ramp costs. - The fourth-quarter revenue guide. It is the supplier's read on customer orders for the rest of the year. - Capex for 2026 and any early 2027 indication. This is what equipment makers' revenue depends on. - The full-year growth outlook. In July it was raised to just above 40% in US dollars.

None of these is a forecast. They are the variables the reaction will depend on.

Educational macro context only — not investment advice.