Taiwan Semiconductor Manufacturing Company delivered a powerful third-quarter earnings report on October 15, 2026, but the more consequential number may be the fourth-quarter forecast. The foundry reported US$46.72 billion in quarterly revenue, a 66.4% gross margin and NT$763.06 billion in net income, while projecting US$52.3 billion to US$54.3 billion in revenue for the final quarter of the year. That combination makes the results a test of how rapidly artificial-intelligence infrastructure demand is becoming manufacturing revenue, rather than merely another record quarter. All reported numbers come from TSMC’s official Q3 results and Q4 guidance.
For investors, three questions now matter more than the headline earnings beat. Can 2-nanometer manufacturing scale without materially damaging margins? Can advanced packaging and overseas expansion support the next wave of AI accelerator orders? And can a fundamentally strong chip supply chain outperform when higher inflation raises discount rates for technology equities? The last question became more pressing after the September US consumer inflation report showed headline CPI running at 3.7% year over year. These are related market forces, not proof that every move in TSM shares is caused by inflation.
Key takeaways
- TSMC generated US$46.72 billion of Q3 2026 revenue and NT$763.06 billion of net income, with year-over-year growth of 41.2% in US-dollar revenue and 68.7% in local-currency profit.
- Its 66.4% gross margin was inside management’s prior 65%–67% range; the 58.7% operating margin exceeded the prior 56%–58% range.
- Advanced technologies represented 79% of wafer revenue, including an 8% contribution from the newest 2-nanometer node.
- Q4 revenue guidance of US$52.3 billion–54.3 billion is a forecast, not a completed sale; capacity, yield and customer deployment remain crucial.
- Strong foundry earnings do not remove valuation, energy-cost, interest-rate or geopolitical risks for AI-linked stocks.
TSMC Q3 2026 earnings: the numbers that matter
The cleanest way to read this report is to separate reported financial performance from the next-quarter outlook. Revenue denominated in New Taiwan dollars grew 50.9% from a year earlier, while the US-dollar presentation grew 41.2%. These two growth rates are not contradictory. The reporting currency and exchange-rate translation differ. Net income rose 68.7% in local currency, faster than local-currency sales, suggesting that operating leverage and product mix helped profitability during the period. That is a historical observation; it does not imply identical leverage in subsequent quarters.
| Q3 2026 metric | Reported result | Comparison / definition |
|---|---|---|
| Revenue | US$46.72 billion | +41.2% YoY in USD; +16.2% QoQ in USD |
| Revenue, reported local currency | NT$1,494.24 billion | +50.9% YoY under TIFRS |
| Net income | NT$763.06 billion | +68.7% YoY |
| Diluted earnings per ADR | US$4.60 | Equivalent to NT$29.43 per common share |
| Gross margin | 66.4% | Within 65%–67% guidance |
| Operating margin | 58.7% | Above 56%–58% guidance ceiling |
| Net profit margin | 51.1% | Reported consolidated result |
| 7nm-and-below wafer revenue mix | 79% | 2nm 8%; 3nm 33%; 5nm 28%; 7nm 10% |
| Q4 2026 revenue guidance | US$52.3–54.3 billion | Forward-looking management range, not realized revenue |
| Q4 2026 gross margin guidance | 65%–67% | Forward-looking management range |
Source: TSMC Q3 2026 investor results and Q4 guidance and TSMC October 15 earnings announcement. Quarterly figures and forecast ranges are different categories of evidence; currency comparisons use TSMC’s stated basis.
A comparison with the company’s Q3 2025 release adds context. A year earlier TSMC reported a 59.5% gross margin, a 50.6% operating margin and 74% of wafer revenue from 7-nanometer and more advanced nodes. The latest 66.4%, 58.7% and 79% figures therefore show a higher advanced-node mix and a materially stronger margin profile. The causal explanation is not exclusively AI: smartphones, product mix, foreign exchange, utilization and manufacturing efficiency can also affect margins. It is safer to describe the combination than to assign an invented dollar contribution to any single customer.
Why the Q4 guidance may matter more than the earnings beat
The company’s investor page places actual Q3 revenue of US$46.72 billion next to its earlier Q3 guidance of US$44.6 billion–45.8 billion. The reported figure exceeded the upper boundary by US$0.92 billion, or approximately 2.0% of that upper bound. Investors should not confuse this calculation with a sell-side consensus surprise, which would require a separate, timestamped analyst estimate. Management’s Q4 outlook of US$52.3 billion–54.3 billion implies a midpoint of US$53.3 billion, roughly 14.1% above actual Q3 revenue. The implied sequential increase is a calculation, not an official growth promise.
Guidance can change because customers may pull forward or delay deliveries, process yields may shift, foreign exchange can alter translated revenue and the timing of equipment installation matters. The guidance also assumes NT$31.8 per US dollar, compared with the actual Q3 reference exchange rate of NT$31.98 shown in the company table. Analysts should compare revenue in the same currency and use the stated assumptions rather than combine different exchange rates casually. A beat relative to management’s range says something about execution; the next test is whether the company can deliver that higher fourth-quarter revenue while maintaining the forecast 65%–67% gross margin.
What 2-nanometer revenue actually tells us
The 2-nanometer process contributed 8% of Q3 wafer revenue. That is a significant early indicator because leading-edge manufacturing depends on more than announcing a process node: commercial customers must qualify designs, the foundry must produce acceptable yields and production volumes must rise without unacceptable scrap or rework. An 8% share shows that the technology has moved into meaningful reported shipments. It does not reveal individual customer identities, the profitability of the node or the number of chips sold. TSMC did not publish those figures in the summary release.
The remaining mix is also informative. Three-nanometer accounted for 33%, five-nanometer for 28% and seven-nanometer for 10% of wafer revenue. Together with 2-nanometer, those figures produce the reported 79% advanced-technology mix. The denominator is wafer revenue, not total corporate revenue, so the percentages should not be multiplied mechanically by US$46.72 billion to claim exact product-node sales. Advanced packaging, other manufacturing services and accounting classifications complicate that shortcut. The mix nevertheless shows how much of TSMC’s economics now depends on complex manufacturing technologies with high barriers to entry.
The AI accelerator demand chain is longer than a GPU order
An AI cloud provider can announce a multiyear infrastructure commitment without creating an equivalent amount of current-quarter foundry revenue. The path from demand to recognized manufacturing revenue includes chip design, wafer starts, fabrication cycle time, testing, advanced packaging, memory integration, delivery and customer acceptance. At each stage, a different bottleneck can delay the next shipment. For that reason, a strong revenue guide is useful evidence of near-term delivery expectations, while headlines about future cloud spending should be treated as demand signals rather than realized semiconductor sales.
This distinction matters for investors who compare Nvidia, AMD, Broadcom and TSMC. The first three sell different mixtures of chips, systems, IP and networking products; TSMC is a manufacturing supplier that can benefit from several design winners. But it is not economically neutral. The foundry must fund expensive capacity before every wafer is sold, and its margin depends on utilization, pricing, yields and process mix. A customer can report booming bookings while TSMC experiences a temporary packaging constraint, or TSMC can book strong wafer sales before the customer fully monetizes an AI service. Cross-company numbers must therefore be reconciled by revenue recognition rather than treated as one synchronized growth rate.
Margins: the quality test behind the revenue surge
A 66.4% gross margin and 58.7% operating margin are striking, but their interpretation requires discipline. Gross margin is revenue minus cost of goods sold divided by revenue; operating margin also includes operating expenses. A high gross margin can reflect favorable pricing, manufacturing yields, product mix and utilization. A rising operating margin can reflect scale and cost discipline. Neither measure is identical to free cash flow, and neither guarantees that capital spending is low. Semiconductor manufacturing is capital intensive even during unusually profitable quarters.
TSMC’s Q4 gross-margin guidance of 65%–67% overlaps the reported Q3 outcome, while operating-margin guidance of 56.5%–58.5% is slightly below the Q3 actual at its upper edge. One interpretation is that management expects high utilization but still faces cost pressure from technology ramps and international expansion. Another is that guidance incorporates prudent assumptions rather than signaling deterioration. The release alone cannot distinguish these explanations conclusively. Investors should compare the actual Q4 mix and margins when reported, instead of declaring a decline merely because the midpoint is lower.
Why currency translation can mislead a growth comparison
Investors reading English-language financial coverage may encounter both 50.9% revenue growth and 41.2% revenue growth for the same quarter. The former is the local-currency year-over-year increase; the latter is the US-dollar increase disclosed by the company. Currency conversion matters because the New Taiwan dollar and US dollar do not remain fixed over a year. The proper comparison depends on the question: local-currency growth helps assess the reported accounting business, while US-dollar revenue is convenient for comparing the global AI hardware supply chain.
For earnings modeling, using one currency consistently is more important than choosing the most impressive percentage. A model that forecasts US-dollar sales but applies a local-currency growth rate without an exchange-rate bridge may overstate or understate revenue. The same issue affects ADR earnings: NT$29.43 per common share and US$4.60 per ADR unit are not interchangeable per-share figures. ADR structures and currency translation must be respected. The official investor presentation hub is a better starting point than a social-media earnings graphic that omits the basis.
TSMC versus its own prior-quarter expectations
Investors often ask whether TSMC “beat estimates,” but the word estimate can refer to several different benchmarks. The official Q3 guidance was US$44.6 billion–45.8 billion of revenue, a 65%–67% gross margin and a 56%–58% operating margin. Actual revenue exceeded the range; gross margin landed inside; operating margin was 0.7 percentage point above the upper boundary. These are comparisons with management’s own published guidance, not a claim about analyst consensus.
The distinction is valuable because markets price expectations before a release. A company can outperform its own guidance and still decline if traders expected an even bigger beat, or if the broader equity market is de-risking. Conversely, a modest reported beat can drive a rally if it resolves a major uncertainty about capacity or demand. Reports that the ADR traded lower around the release should be read with an exact session and quote timestamp. This article does not assign a definitive closing percentage before the US trading session is complete; intraday movement is not a final closing return.
How a hotter CPI report changes the valuation discussion
The BLS September 2026 CPI release reported a 0.6% monthly headline increase and 3.7% annual inflation, with gasoline prices up 9.1% month over month. Core CPI increased 0.3% month over month and 2.6% year over year. Those are consumer price data, not direct measurements of TSMC input costs. Nevertheless, higher-than-expected inflation can influence interest-rate expectations, Treasury yields and the discount rates investors use to value long-duration growth companies. That channel is particularly relevant to richly valued AI infrastructure shares.
Suppose investors forecast the same cash flows from a semiconductor company but require a higher return to hold its shares. The present value of those cash flows can decline even if the revenue outlook improves. This explains how good company news and a falling stock price can coexist. It does not establish that every price decline is rational or that inflation must always overwhelm earnings. The impact depends on valuation, the maturity of cash flows, changes in expected growth and the wider market’s risk appetite. For broader context, KCEX Blog’s analysis of dollar strength, Treasury yields and Bitcoin separates several of these transmission channels.
Three ways the AI supply chain can absorb a macro shock
First, firms with durable demand and pricing power may continue growing revenue despite higher financing costs. TSMC’s advanced-node mix is relevant here, because customers have limited near-term substitutes for certain leading-edge production capabilities. Second, a prolonged increase in real yields can compress equity valuation multiples even when earnings rise. Third, a macro shock can delay downstream data-center construction, electricity connections or customer procurement, which would eventually affect the timing of chip orders. These are distinct mechanisms and should not be collapsed into a single “AI is recession-proof” or “AI is doomed” narrative.
Investors should also separate the chip industry from the digital-asset market. Bitcoin does not have a semiconductor earnings statement, while AI-related stocks have revenue, margins and capital spending. Both can respond to liquidity and interest-rate expectations, but their fundamental cash-flow profiles are different. Correlations can rise during a forced deleveraging event without implying that the underlying businesses have become identical. A thoughtful cross-asset portfolio therefore measures exposure to the same macro factor instead of assuming diversification simply because ticker symbols belong to different categories.
Capacity expansion: why a backlog is not a free lunch
Advanced semiconductor production requires fabs, specialized tools, materials, trained staff, utilities and reliable logistics. New capacity can expand the addressable market, but the spending often precedes revenue by years. Investors should look for the timing of equipment installation, depreciation, production qualification and customer commitments. A fabrication facility can be strategically important while depressing near-term returns on invested capital during its ramp. Equally, delaying investment can protect near-term cash flow while surrendering future market share.
TSMC’s reported margin profile suggests substantial current earnings power, but it does not by itself settle the optimal pace of future investment. Management must balance advanced-node demand with construction cost, international diversification and customer needs. The trade-off becomes sharper if global interest rates remain elevated. Investors should not confuse a management capital-expenditure budget with a contractual guarantee of future revenue. A useful follow-up is to compare capex to operating cash generation and subsequent depreciation, rather than judging the spending figure in isolation.
Advanced packaging: a second manufacturing constraint
AI accelerator economics depend on packaging as well as transistor density. Modern high-performance systems combine logic, memory and high-bandwidth connections; integrating those components at scale can be challenging even when leading-edge wafers are available. If packaging capacity lags, a customer may have finished wafers but insufficient complete products. If memory supply tightens, an otherwise available accelerator may not ship on the expected timetable. These constraints are different from wafer yield and can move independently.
A strong foundry earnings release does not quantify every packaging queue. Investors should resist attributing a particular percentage of revenue growth to a packaging technology unless the company discloses it. Instead, monitor public capacity commentary, customer shipment schedules and changes in lead times. The broader AI hardware debate is explored in KCEX Blog’s Broadcom AI semiconductor earnings analysis, which highlights the role of custom accelerators and networking alongside GPUs. That comparison is useful because the fastest-growing customer architecture may alter the mix of packaging requirements even when total compute demand continues expanding.
Customer concentration and the difference between demand and dependence
A concentrated set of large AI chip designers can be positive for manufacturing scale but can also increase negotiating leverage and timing risk. A few hyperscalers may accelerate spending simultaneously, lifting utilization and margins. The same customers can pause procurement, redesign chips or spread orders across suppliers when priorities change. TSMC’s summary release does not provide a customer-by-customer revenue table, so attributing the full quarter to any named technology company would be unsupported.
Investors can still evaluate the economic structure. A foundry benefits from being the production partner to multiple design firms, which diversifies design outcomes. However, several designers may ultimately depend on the same small group of cloud customers. This creates indirect concentration that is not visible from a list of chip brands. When assessing the sustainability of demand, track cloud capital expenditure, customer deployment and monetization together. The critical question is whether AI infrastructure spending becomes recurring economic output, not merely whether a record chip shipment quarter can be repeated.
What the share-price reaction can and cannot prove
An earnings-day stock move combines new information, existing expectations and the market environment. If the shares decline after an earnings beat, one possible explanation is that traders were already positioned for strong results. Another is concern about future margins or the required capital investment. A third is broad risk reduction after the US inflation shock. Without synchronized order-flow evidence, these explanations remain hypotheses rather than verified causes. This is especially important when comparing the Taiwan-listed shares with the New York ADR, which trade in different sessions and currencies.
Investors should capture the security, exchange, exact timestamp, price basis and benchmark before citing a performance figure. A premarket ADR quote is not a Taiwan close; a percentage from an intraday high is not a daily return. Comparing TSM against a semiconductor index can help isolate company-specific performance, but even relative returns may reflect different index weights and investor exposures. This methodological discipline is more valuable than a dramatic headline about a stock “ignoring” earnings.
A practical earnings scorecard for the next quarter
A useful monitoring framework starts with five indicators. Revenue delivery against the US$52.3 billion–54.3 billion Q4 range tests the near-term shipment outlook. Gross margin against 65%–67% tests the balance between mix and ramp costs. Operating margin against 56.5%–58.5% tests operating leverage. The share of wafer revenue from 2nm and 3nm tests the advanced-node transition. Finally, capex and cash generation test whether growth translates into sustainable returns after the manufacturing investment required to support it.
None of these indicators should be viewed alone. Revenue can exceed guidance because of currency or timing while underlying demand changes little. Margins can soften temporarily during a successful technology ramp. A rising 2nm share may be positive strategically but require learning-curve spending. Strong cash generation can coexist with a large capex cycle. Investors should compare actuals with the company’s prior assumptions and identify what changed, rather than simply color-coding every number green or red.
Bull, base and bear scenarios: conditional, not price targets
In a constructive scenario, Q4 revenue lands near or above the top of guidance, gross margin remains within the stated range and advanced-node adoption broadens. That would support the thesis that AI and other high-performance demand is translating into profitable wafer shipments. It would not guarantee a higher share price, because valuation may already anticipate this outcome. In a middle scenario, revenue remains within the guided range and margins normalize modestly as 2nm and overseas capacity ramp. The business could remain strong even if the stock trades sideways while investors reassess discount rates.
In a downside scenario, customer delivery schedules slip, manufacturing costs rise or macro tightening suppresses capital expenditure. Revenue could miss the range, or margins could fall below guidance. The earliest warning might be changing shipment commentary rather than an immediate collapse in headline AI spending. No numeric TSM price target is provided because a defensible valuation target would require a current market quote, diluted share count, cash-flow forecast, discount rate and peer assumptions measured on a consistent basis. Inventing a target from the earnings headline would offer false precision.
Why the results matter for Nvidia, AMD and Broadcom investors
TSMC is a common manufacturing node in the economics of several AI chip suppliers, but its results should not be mapped one-for-one onto any customer’s earnings. A design company can grow revenue faster than foundry wafer sales because of product pricing, system content, software or inventory timing. It can also grow more slowly if wafer output rises before final products are accepted. Some companies use different process nodes or packaging architectures. These distinctions help explain why a strong TSMC quarter is evidence of industry demand without being a complete earnings forecast for every AI-linked stock.
For portfolio analysis, TSMC offers exposure to manufacturing capacity and process leadership; chip designers offer exposure to architecture, customer relationships and sometimes software ecosystems. Those exposures overlap but are not identical. KCEX Blog’s AI compute and CPU-demand research discusses how workloads can expand demand beyond accelerators into CPUs and other infrastructure. That broader perspective matters when investors extrapolate a single GPU narrative into a forecast for the entire semiconductor sector.
Geopolitical and operational risks investors should not ignore
The foundry’s strategic importance also creates nonfinancial risks. Geographic concentration, trade restrictions, export controls, power and water availability, logistics and natural disasters can affect operations or customer access. Expansion into additional countries can diversify some operational exposure while raising costs and introducing new labor, permitting and supply-chain challenges. A higher valuation multiple is not automatically justified merely because the company occupies a critical point in global technology infrastructure.
Operational risks are difficult to price precisely because probabilities and consequences are uncertain. Rather than attaching an arbitrary geopolitical discount, investors can examine business continuity plans, capacity distribution, supplier redundancy and the sensitivity of earnings to production interruptions. These questions belong alongside growth and margin analysis. The latest earnings demonstrate what the company produced during the quarter; they cannot establish that future production will be immune to disruptions.
How to interpret the headline “AI demand is booming”
The evidence supports a narrower, stronger statement: TSMC reported exceptionally fast revenue growth, a high advanced-node mix and guidance for another sequential revenue increase. Those observations are consistent with strong demand for advanced chips, including AI applications. They do not quantify the proportion of sales attributable exclusively to AI, nor prove that every AI project will earn an adequate return on investment. The difference between consistent evidence and a complete causal breakdown is important for credible research.
A better research process traces four stages: announced customer spending, supplier orders, recognized supplier revenue and downstream utilization or revenue. Each stage has different accounting and economic meaning. Analysts who skip the intermediate steps risk double-counting the same investment across cloud providers, chip designers, foundries and equipment suppliers. TSMC’s earnings are one of the more concrete observations in that chain, but not the whole chain.
Frequently Asked Questions
FAQ: Did TSMC beat its own Q3 revenue guidance?
Yes. Actual Q3 revenue was US$46.72 billion, versus a management guidance range of US$44.6 billion to US$45.8 billion. The result was US$0.92 billion above the upper end. That comparison uses company guidance, not a separately surveyed Wall Street consensus. The relevant primary source is TSMC’s Q3 investor results page.
FAQ: What was TSMC’s Q3 2026 gross margin?
The company reported 66.4% gross margin, 58.7% operating margin and 51.1% net profit margin. Gross margin was within its previously indicated 65%–67% range, while operating margin exceeded the prior 56%–58% range. These are quarterly accounting metrics; they are not free-cash-flow margins or a forecast for the next year.
FAQ: How much of TSMC wafer revenue came from 2nm?
The 2-nanometer node accounted for 8% of third-quarter wafer revenue. Three-nanometer accounted for 33%, five-nanometer 28% and seven-nanometer 10%, for a combined 79% share from 7nm and more advanced technologies. The percentages refer to wafer revenue, not total company revenue or customer shipment units.
FAQ: What is TSMC forecasting for Q4 2026?
Management expects US$52.3 billion–54.3 billion in revenue, 65%–67% gross margin and 56.5%–58.5% operating margin. These are forward-looking ranges, conditional on manufacturing and market conditions. They should not be presented as already achieved financial results. Currency assumptions also matter: the published outlook uses NT$31.8 per US dollar.
FAQ: Why might TSM stock fall despite strong earnings?
A stock can fall when results fail to exceed what investors had already priced in, when future cost concerns rise or when a macro shock pushes required returns higher. The hotter September CPI release created an additional valuation consideration for technology stocks. None of these hypotheses alone proves the cause of a specific intraday move. A responsible market recap uses a dated, exchange-specific price and a suitable benchmark before drawing conclusions.
Methodology, sources and limits
This analysis treats TSMC’s October 15 earnings announcement and Q3 investor results as primary evidence for financial figures and Q4 guidance. The company’s earnings release provides reported consolidated results; the investor relations page provides guidance ranges and the currency assumptions. The BLS October 14 CPI release provides the separate macroeconomic backdrop. Calculations such as the Q4 guidance midpoint and the amount by which revenue exceeded the prior upper bound are straightforward arithmetic derived from those published figures. No unverified consensus estimate or unverified current share quote has been converted into a purported fact.
The analysis is informational rather than personalized investment advice. Financial results are historical; company guidance is forward-looking; scenarios are conditional reasoning. Semiconductor equities can experience sharp losses despite strong revenue growth because valuation, liquidity, technology execution and geopolitical exposure can change independently. Readers should consult the latest company filings and their own risk constraints before making a trading decision.
