Taiwan Semiconductor beat its own guidance on every line that matters: revenue, gross margin, and operating margin. Then the stock barely moved in the regular session and slipped in after-hours trading. If you’re trying to figure out what that disconnect says about the AI-capex trade, or how the options market handled it, here’s the actual scorecard from Thursday’s report.
Key Takeaway
- TSMC’s Q2 2026 revenue landed at $40.20 billion and gross margin at 67.7%, both above the top end of the company’s own April guidance.
- Despite the beat, TSM shares closed roughly flat in the regular session on July 16 and fell about 1.6% in after-hours trading as investors focused on heavier 2026 capital spending and softer Q3 margin guidance.
- CEO C.C. Wei said advanced-packaging (CoWoS) capacity is sold out through 2026, with Nvidia alone reported to have booked roughly 60% of that capacity, a direct signal on how tight the AI chip supply chain remains.
- Pre-earnings options pricing had implied a move of roughly 5% to 6.65% for TSM shares. The realized move was smaller than the higher end of that range, a good reminder that expected move is a probability band, not a prediction.
- TSMC raised its full-year 2026 revenue growth outlook to “slightly above 40%” and lifted capex guidance, alongside news of an additional $100 billion planned investment in Arizona.
What TSMC Actually Reported
TSMC posted consolidated Q2 2026 revenue of NT$1,270.38 billion, translating to $40.20 billion, at the very top of the $39.0 billion to $40.2 billion range management had guided to back in April. Gross margin came in at 67.7%, above the guided 65.5% to 67.5% range. Operating margin hit 60.3%, well above the 56.5% to 58.5% guided range. Net profit margin was 55.6%.
Revenue grew 36.0% year over year and 12.0% sequentially. Net income jumped 77.4% year over year and 23.4% from the prior quarter, reaching NT$706.56 billion, a new quarterly record. Diluted EPS was NT$27.25 per common share, or $4.31 per ADR unit (each TSM American depositary receipt represents five common shares, so ADR EPS runs roughly five times the local-share figure before currency conversion).
| Metric | Guided (April 2026 call) | Actual (reported July 16, 2026) |
|---|---|---|
| Revenue | $39.0B – $40.2B | $40.20B (top of range) |
| Gross margin | 65.5% – 67.5% | 67.7% (above high end) |
| Operating margin | 56.5% – 58.5% | 60.3% (above high end) |
| Net profit margin | Not separately guided | 55.6% |
| YoY revenue growth | Implied roughly mid-30s% | 36.0% |
Every metric TSMC guides to came in above the high end of its own range. That is a clean beat by any conventional measure, and it is the kind of quarter that would normally send a stock higher, not sideways.
The Q3 guidance that mattered more than the Q2 beat
Management guided Q3 2026 revenue to $44.6 billion to $45.8 billion, a roughly 12% sequential and 37% year-over-year increase at the midpoint. Gross margin guidance for Q3, however, came in at 65% to 67%, about 1.7 percentage points lower than Q2’s actual result, attributed to dilution from ramping 2nm production. Operating margin guidance for Q3 sits at 56% to 58%.
That single data point, a margin step-down tied to the cost of ramping the next process node, appears to be the main reason the stock did not rally on an otherwise record quarter. When a company clears the bar on the quarter just reported but signals a tougher margin picture ahead, the market tends to price the forward guidance, not the trailing beat.
Why the Stock Fell After Beating
TSM shares closed the regular session on Thursday, July 16 at $419.48, essentially unchanged, then fell about 1.55% in after-hours trading to $412.99. Given the size of the beat, that reaction surprised some traders. Three things appear to explain it.
First, the Q3 margin guidance discussed above signals near-term cost pressure from the 2nm ramp. Second, TSMC lifted its full-year 2026 capital expenditure guidance to a range of roughly $60 billion to $64 billion, the largest capex commitment in company history, and announced an additional $100 billion planned investment in Arizona. Heavier spending compresses free cash flow in the near term even when it is aimed at capturing demand that clearly exists. Third, and probably most important, TSMC had already become priced for perfection. Shares had run up meaningfully into the print, and when a stock is trading on the assumption that every quarter will be a blowout, a merely excellent quarter with a soft spot in the outlook can read as a disappointment relative to whisper numbers, even though every headline metric beat the official guide.
This is a pattern worth remembering for any high-multiple bellwether stock: the setup for the print (how much good news is already in the price) often matters more to the immediate reaction than the print itself.
The AI-Capex Read-Through: What TSMC’s Call Signals for NVDA, AMD, and Hyperscalers
TSMC does not just make chips, it makes almost everyone else’s advanced chips, which is why its quarterly numbers function as a proxy for demand across the entire AI hardware stack. On the call, CEO C.C. Wei described AI demand as “extremely robust” and characterized the current buildout as the emergence of “a new industry called AI industry,” one he expects to eventually touch automotive and robotics applications as well as data centers.
The more concrete signal came on advanced packaging. Wei said CoWoS (chip-on-wafer-on-substrate) capacity, the packaging technology that lets multiple chips be combined into a single AI accelerator package, is “extremely tight and sold out through 2026.” TSMC has been expanding CoWoS capacity at roughly 80% per year, from about 35,000 wafers per month at the end of 2024 to an estimated 75,000 by the end of 2025, with a target of 125,000 to 130,000 wafers per month by the end of 2026. Total CoWoS demand for 2026 is projected at roughly one million wafers, nearly triple the 2024 level. Nvidia is reported to have booked around 60% of that capacity through 2026, with additional claims on the 2026-2027 expansion.
For traders who watch NVDA, AMD, and the hyperscalers as a group, the read-through is straightforward: the bottleneck in the AI buildout right now is not chip design or GPU demand, it is packaging capacity at TSMC. That keeps TSMC in a strong pricing position, but it also means any slip in TSMC’s packaging ramp becomes a supply constraint for every AI accelerator maker downstream, not just for TSMC’s own numbers.
ADR Mechanics Retail Traders Often Miss
A few structural quirks of trading TSM specifically (versus TSMC’s Taiwan-listed common shares) trip up retail traders:
It’s an ADR, not a direct listing
TSM trades on the NYSE as an American depositary receipt. Each ADR represents five common shares of the Taiwan-listed stock. That ratio matters when comparing per-share figures. TSMC’s headline EPS of NT$27.25 is a per-common-share number; the $4.31 figure often quoted for the ADR reflects both the 5-to-1 ratio and currency conversion, not a simple exchange-rate translation of the local EPS.
Reporting currency is New Taiwan Dollars
TSMC reports in NT$ under Taiwan-adopted IFRS, then provides a USD convenience translation. The USD figures that circulate in US financial media (like the $40.20 billion revenue number) are translations, and the exact USD figure can shift slightly depending on the exchange rate used at the time of reporting.
The earnings call timing is unusual for a US-listed stock
TSMC reports on Taiwan’s clock. The Q2 2026 call was held Thursday afternoon in Taipei, which lands in the very early morning in New York, well before the US market opens. That means the “reaction” US traders see at the open already reflects a full round of Asian-market and pre-market digestion of the numbers, unlike a typical US company that reports after the 4pm close.
Quarterly cadence, not a fiscal-year offset
TSMC’s fiscal quarters line up with the calendar year (Q1 = Jan-Mar, Q2 = Apr-Jun, and so on), unlike some large-cap tech names with offset fiscal years. That makes it easier to compare TSMC’s results directly against calendar-quarter data from customers like Nvidia and AMD when you’re doing the read-through analysis above.
What the Options Market Priced In vs. What Actually Happened
Ahead of the print, TSM options were pricing an implied move of roughly 5% to 6.65%, depending on which model and expiration you looked at, with some estimates putting nearly $150 billion of TSMC’s market value in play. Implied volatility and IV rank were both elevated into the report, meaning options were relatively expensive by the stock’s own recent history, which is typical the week of an earnings release.
The expected move (sometimes called implied move) is generally derived from about 85% of the value of the at-the-money straddle for the nearest expiration. It is a probability range, not a forecast: it tells you the market’s estimate of a roughly one standard deviation move, not which direction the stock will go or how far it will actually travel.
In this case, the realized move (essentially flat in the regular session, then down about 1.55% after hours) landed well inside the smaller end of the pre-earnings implied range. That is a common outcome. Expected move estimates are built to cover a wide range of scenarios, including the tail cases, so a print that clears every guided metric but still disappoints on forward margin commentary can easily resolve toward the middle of that range rather than the edges.
A hypothetical, illustrative example only
To be clear, none of the following is a trade recommendation, a price target, or advice to buy or sell TSM. It is a purely illustrative walkthrough of how a defined-risk options structure could have been sized around this specific event, for educational purposes only.
Suppose a hypothetical trader, watching the pre-earnings implied move of roughly 5% to 6.65%, wanted defined risk on the idea that the realized move would come in smaller than the options market was pricing. One illustrative approach: sell a short strangle with both legs placed outside the expected-move boundaries (for example, roughly 6% to 7% out of the money on each side) and simultaneously buy further out-of-the-money options against each leg to cap risk, forming an iron condor. The position would collect premium up front and profit if the stock stayed inside the wider band by expiration, while the long legs define the maximum possible loss if the move proved larger than expected. This is a defined-risk structure precisely because both the maximum gain (net premium collected) and maximum loss (width of the wings minus premium) are known before the position is opened.
A trader evaluating a structure like this would typically want a platform with solid options-chain analytics and probability-of-profit tools built in; a platform like tastytrade is built specifically around that kind of options-first workflow. Whatever platform is used, the sizing decision (how far out of the money to place the strikes, and how much capital to risk on the wings) is what actually determines the risk profile, not the platform itself.
Bottom Line
TSMC’s actual Q2 2026 results beat guidance on revenue, gross margin, and operating margin, and management raised its full-year growth outlook. The stock’s muted-to-negative reaction was driven by heavier 2026 capex plans and softer Q3 margin guidance tied to the 2nm ramp, not by any miss on the quarter itself. For the broader AI-capex trade, the clearest signal was CoWoS packaging capacity being sold out through 2026, which points to the AI hardware bottleneck sitting at the packaging stage rather than at chip design or end demand.
FAQ
Q: Did TSMC beat or miss Q2 2026 earnings?
A: TSMC beat its own guidance on every headline metric. Revenue of $40.20 billion came in at the top of the guided $39.0 billion to $40.2 billion range, gross margin of 67.7% beat the guided 65.5% to 67.5% range, and operating margin of 60.3% beat the guided 56.5% to 58.5% range.
Q: Why did TSM stock fall if the quarter beat guidance?
A: Shares were roughly flat in the regular session and fell about 1.55% after hours, largely because Q3 gross margin guidance (65% to 67%) came in below Q2’s actual 67.7%, reflecting cost dilution from the 2nm ramp, alongside a large increase in planned 2026 capital spending. The market reacted to the forward outlook, not the trailing quarter.
Q: What is TSMC’s expected move and how is it calculated?
A: The expected (or implied) move is typically calculated as roughly 85% of the value of the at-the-money options straddle for the nearest expiration. Ahead of this report, estimates for TSM ranged from about 5% to 6.65%. It represents a probability range around a roughly one standard deviation outcome, not a directional prediction.
Q: How many shares does one TSM ADR represent?
A: One TSM American depositary receipt represents five common shares of TSMC’s Taiwan-listed stock. Per-share figures reported in New Taiwan Dollars need to be adjusted for both this ratio and the currency conversion before comparing them to the ADR’s US-dollar EPS.
Q: What does TSMC’s report mean for Nvidia and AMD?
A: TSMC’s commentary that CoWoS advanced-packaging capacity is sold out through 2026, with Nvidia reported to have booked roughly 60% of it, suggests the AI accelerator supply chain remains packaging-constrained rather than demand-constrained. That is generally read as a supportive signal for underlying AI chip demand, though it also means packaging capacity, not GPU design, is the bottleneck to watch.
Want more on how this quarter fits into the broader AI-capex story? Check TRDC’s other Q2 2026 earnings coverage in the Market Analysis section for related reads on how bellwether reports like this one ripple through the rest of the semiconductor and AI trade.