TSMC Q3 Earnings Preview: Can Record AI Demand Sustain TSM's Margin Growth?

 

TSMC holds its third-quarter earnings call on October 15, but revenue is no longer the question. The company's 2026 monthly revenue disclosures show July, August and September sales of NT$467.58 billion, NT$514.81 billion and NT$511.86 billion, for a quarterly total near NT$1.494 trillion. Against the same three months of 2025, which totaled roughly NT$989.9 billion, that is growth of about 50.9%. Revenue for the first nine months stands at NT$3.899 trillion, up 41.1%.

 

With the top line already public, the open variables are margin, capital spending and the outlook for next year. According to Reuters' earnings preview, an LSEG SmartEstimate drawn from 16 analysts puts third-quarter net profit near NT$746.3 billion, about $23.4 billion, a 65% increase from a year earlier. That would mark a sixth consecutive record quarter and an eleventh consecutive quarter of profit growth. The harder question is whether profitability can keep pace when revenue is expanding at fifty percent.

 

 

 

Revenue cleared the top of the company's own guidance. July guidance called for $44.6 billion to $45.8 billion, and at the NT$32 exchange rate assumption in that release, NT$1.494 trillion converts to roughly $46.7 billion.

 

The implied net margin is heading lower. Dividing the consensus profit by reported revenue gives an implied net margin near 50%, against the 55.6% TSMC actually posted in the second quarter.

 

Customer demand has not cooled. NVIDIA reported $96.2 billion of revenue in its most recent quarter, up 106% year over year, with data center revenue up 117% and next-quarter guidance of $108 billion.

 

Packaging capacity is still the binding constraint. TrendForce, citing market estimates, expects the CoWoS supply gap to narrow from around 20% to roughly 10% by the end of 2026, with capacity set to expand more than 60% further by 2027.

 

Capital spending has already been raised once. The 2026 capex range moved to $60 billion to $64 billion from $52 billion to $56 billion, while cumulative Arizona investment now stands at $265 billion.

 

The stock has priced in a great deal. Taipei-listed shares are up 63% this year against 69% for the broader market, and the US-listed ADR closed at $453.31 on October 9 for a market value near $2.06 trillion.

 

 

 

Because TSMC reports revenue every month, the quarterly top line arrives well before the call. September revenue rose 54.6% year over year, August 53.3% and July 44.7%, an acceleration within the quarter rather than a seasonal pull-forward. That pattern points to capacity coming online and to a mix shifting toward higher-value nodes.

 

The more informative comparison is against guidance. The second-quarter results release put third-quarter revenue guidance at $44.6 billion to $45.8 billion. Converted at the NT$32 assumption in the same document, actual revenue lands near $46.7 billion, roughly 2% above the top end. For a company approaching NT$1.5 trillion in quarterly sales, that is not a rounding difference.

 

 

The second quarter set the benchmark: revenue of NT$1,270.38 billion, net income of NT$706.56 billion, diluted EPS of NT$27.25, gross margin of 67.7%, operating margin of 60.3% and net margin of 55.6%. Management then guided third-quarter gross margin to 65% to 67% and operating margin to 56% to 58%, both below what the second quarter delivered.

 

Setting the NT$746.3 billion consensus against reported revenue implies a net margin near 50%, more than five points below the prior quarter. Part of that reflects the guided step down in gross margin, and part may come from non-operating items, particularly currency effects. What matters on the call is not the profit figure but how management attributes the compression. Early-stage 2nm depreciation and yield costs plus overseas fab dilution would persist for several quarters. Currency swings and one-off items would not.

 

 

 

The cleanest read on TSMC's visibility comes from what its largest customers report. NVIDIA's most recent quarterly results show revenue of $96.2 billion, up 18% sequentially and 106% from a year earlier, with data center revenue of $89.0 billion, up 117%. Guidance for the current quarter is $108 billion give or take 2%, and it assumes no data center compute revenue from China. Jensen Huang said demand is accelerating.

 

The relevance for TSMC is that its capacity is being consumed by confirmed orders rather than anticipated ones. AMD's data center accelerators, Apple's silicon and the custom ASICs commissioned by hyperscalers all route through the same advanced nodes and the same packaging lines. That concentration is the source of the revenue and, as discussed below, the source of the risk.

 

 

The second-quarter release puts 3nm at 30% of wafer revenue, 5nm at 33%, 7nm at 11%, and advanced technologies at 77% in aggregate, with 2nm contributing 3% in its first quarter of revenue. Sequentially, 3nm rose from 25%, 5nm fell from 36% and 7nm fell from 13%.

 

That migration supports pricing, since leading-edge wafers carry higher prices. It also carries a cost, because depreciation and yield losses on a new node land before the revenue does. That is precisely the role 2nm plays over the next several quarters.

 

 

 

For two years the constraint on AI chip shipments has been advanced packaging rather than wafer supply. Per the market estimates compiled by TrendForce, TSMC's monthly CoWoS capacity could reach a record 120,000 to 140,000 wafers in 2026, and adding 50,000 to 60,000 wafers from outsourced partners brings industry capacity close to 200,000 per month, narrowing the supply gap from roughly 20% to about 10% by year end. The same analysis notes plans to expand CoWoS capacity by more than 60% by 2027 and cites Reuters on the company's forecast at its May technology symposium for a compound annual growth rate above 80% between 2022 and 2027.

 

A narrowing gap cuts both ways for TSMC. More orders convert into shipments, but once packaging stops being the sole bottleneck, customer tolerance on pricing shifts too. Both effects belong in any view of 2027 margins.

 

 

2nm began contributing revenue in the second quarter, with the volume ramp concentrated in the second half of 2026. According to MacroMicro's summary of the earnings call, the company built inventory ahead of mass production, overseas fabs diluted gross margin, and management offset part of that through cost improvement and higher utilization.

 

In effect, the gross margin of the third and fourth quarters embeds a prepayment for 2027. Whether that prepayment pays off depends on how quickly customers migrate to 2nm and whether the cost can be passed into price. The first shows up in commentary about customer adoption on the call. The second only becomes visible in the margin curve several quarters from now.

 

 

The Reuters preview notes that 2026 capex guidance has been raised to $60 billion to $64 billion from $52 billion to $56 billion. In Arizona, an additional $100 billion announced in July takes cumulative investment in that state to $265 billion, and management has said capex over the next three years will rise by more than it did over the past three, citing stronger than expected AI demand.

 

Texas is the new variable. The same report notes that Reuters reported in September that the company was evaluating a Texas investment, with the possibility of subleasing part of the complex under Elon Musk's Terafab plan, and analysts expect the topic to come up. Morgan Stanley's Charlie Chan expects full-year revenue growth in the low-to-mid 40% range, above guidance of more than 40%, and sees Texas expansion becoming necessary within five years given limited space in Arizona. Morningstar's Phelix Lee sees less urgency, noting some planned Arizona fabs are not yet built and describing Texas as a good location for a 2030s or 2040s expansion plan.

 

For investors, the composition of capex matters more than the headline. Spending directed at advanced packaging and 2nm is backed by orders already in hand. Spending on new overseas fabs shows up first as depreciation and margin dilution, with returns years away. The company's final wording on full-year 2026 growth, and any directional comment on 2027, will carry the most information of anything said on the call.

 

To watch the price reaction the moment the numbers land, the TSM tokenized stock futures market is where the order book responds first

 

 

 

Customer concentration comes first. AI accelerators account for most of the incremental revenue, and those orders sit with a handful of buyers, so any change in one customer's capex cadence is amplified through TSMC's capacity planning. Currency is second, since the company reports in New Taiwan dollars while selling largely in US dollars, which affects both gross margin and non-operating items, and is one plausible explanation for the implied margin compression this quarter. Overseas expansion is third, because Arizona and any Texas project generate depreciation long before scale benefits. Geopolitics and export controls remain difficult to quantify but are visibly real, as NVIDIA's decision to exclude China data center compute revenue from guidance demonstrates.

 

 

In a strong case, fourth-quarter guidance beats expectations, the full-year growth framing moves above the mid-40% range, and gross margin holds above 65%. The market then reads the margin dip as 2nm prepayment and the valuation re-rates higher.

 

In a neutral case, guidance lands inside expectations with gross margin near the lower end of the 65% to 67% band and capex unchanged. The share reaction hinges on management's language about 2027 rather than on the third-quarter print.

 

In a pressured case, gross margin slips below the guided floor or management hedges on 2027 capital spending, and the market begins questioning the durability of AI-related capex. Given how much the shares have already moved, with stockanalysis data showing a $453.31 close on October 9, a market value near $2.06 trillion and a forward multiple around 20 times, the downside in that case is not trivial.

 

 

Fourth-quarter revenue and gross margin guidance rank first, since they set the full-year landing point. Second is the final framing of 2026 growth and any willingness to point toward 2027. Third is progress on CoWoS capacity and 2nm customer adoption, which determines next year's delivery flexibility. Fourth is the split of capital spending between domestic and overseas capacity. Fifth is the wording on Texas, where even a refusal to comment carries information. Semiconductor earnings rarely stand alone, and memory and logic cycles tend to confirm one another, a dynamic explored in our analysis of Micron's results, while the tokenized US stock market page makes it straightforward to compare how the sector reprices over the same window.

 

 

For James Mitchell, the number that matters is not the 65% profit growth but the widening gap between revenue and profitability. Revenue grew 50.9% year over year, while the consensus profit implies a net margin near 50% against 55.6% in the prior quarter. Revenue came in above the company's own guidance ceiling and margin still compressed. For a foundry with near-monopoly position at the leading edge, that combination deserves to be broken down line by line, because it determines whether this expansion is generating free cash flow or recycling it into the next capacity cycle.

 

Two readings look vulnerable. The first treats higher capex as unambiguously bullish. The $60 billion to $64 billion range does reflect genuine demand, but capital spending is depreciation first and capacity second, with returns measured in years. The $265 billion committed to Arizona contributes nothing but cost during construction. The signal is not the headline number but how much of it goes to CoWoS and 2nm, where orders already exist. The second treats a narrowing packaging gap as purely positive. Closing the gap from roughly 20% to 10% lets more orders ship, but it also relieves the scarcity that supports pricing, which pushes margin in the opposite direction from volume.

 

Three variables deserve tracking from here. The first is the distance between gross margin and the guided floor, since 65% is a clean psychological line and only two consecutive quarters at or below it would establish structural cost pressure. The second is the direction of blended wafer pricing: with 3nm rising from 25% to 30% of wafer revenue and 2nm starting to contribute, a richer mix that fails to lift margin would say costs are eating the benefit faster than expected. The third is customer capex cadence, where NVIDIA's $108 billion guidance is the hardest forward indicator available and any change there reaches TSMC's capacity plans within a quarter or two.

 

The cross-asset lesson runs beyond semiconductors. TSMC, the memory makers and the hyperscalers form one chain of AI infrastructure spending, and that chain is expanding capital investment at a pace with few historical parallels. Anyone holding technology assets needs to separate revenue growth from return on capital: revenue can be bought with capacity, while returns depend on whether that capacity runs full through its depreciation life. TSMC's gross margin curve is among the earliest places in the chain where that answer shows up.

 

 

 

The earnings call is scheduled for October 15 at 2pm Taipei time, which is 0600 GMT. Because the company discloses revenue monthly, the quarterly top line is already known at roughly NT$1.494 trillion, up about 50.9% year over year. The incremental information will come from margin, fourth-quarter guidance, the final framing of full-year growth and capital spending plans.

 

 

Reuters reports that an LSEG SmartEstimate from 16 analysts points to net profit near NT$746.3 billion, about $23.4 billion, up 65% year over year. Delivering that would make it a sixth consecutive record quarter and an eleventh consecutive quarter of sequential profit growth. For comparison, second-quarter net income was NT$706.56 billion with diluted EPS of NT$27.25.

 

 

Dividing the consensus profit by reported revenue implies a net margin near 50%, against 55.6% in the second quarter. Guidance itself called for gross margin of 65% to 67%, below the 67.7% actually achieved. Likely contributors include early-stage 2nm depreciation and yield costs, dilution from overseas fabs, and currency effects on non-operating items. The weighting of each should become clearer on the call.

 

 

Market estimates compiled by TrendForce put TSMC's monthly CoWoS capacity at 120,000 to 140,000 wafers in 2026, with industry capacity approaching 200,000 per month once outsourced partners are included, narrowing the supply gap from roughly 20% to about 10% by year end. Capacity is set to grow more than 60% further by 2027. Easier supply means more shipments, though it also reduces the scarcity premium in advanced packaging.

 

 

2nm contributed revenue for the first time in the second quarter at 3% of wafer revenue, with the volume ramp concentrated in the second half of 2026. A richer node mix generally supports pricing, but a new node carries depreciation and yield costs before it carries profit. That makes 2nm a margin headwind over the next few quarters and a key swing factor for 2027 earnings power.

 

 

Reuters reported in September that the company was evaluating a Texas investment, with the possibility of subleasing part of the complex under Elon Musk's Terafab plan. Morgan Stanley's Charlie Chan argues limited space in Arizona could make Texas necessary within five years, while Morningstar's Phelix Lee sees less urgency and frames Texas as a 2030s or 2040s proposition. No formal announcement has been made, so the wording of any answer on the call is what to watch.

 

 

Taipei-listed shares are up 63% this year, slightly behind the broader market's 69%. The US-listed ADR closed at $453.31 on October 9 for a market value near $2.06 trillion and a forward multiple around 20 times. With that much appreciation behind it, the post-earnings move is more likely to hinge on guidance and capex language than on the third-quarter numbers themselves.

 

 

The information above is provided for general market information and analysis only and does not constitute investment advice, financial advice, legal advice, tax advice or a recommendation to trade. Prices of equities, crypto assets and other related financial assets can fluctuate sharply, and past performance, analyst estimates and financial metrics do not guarantee future results. The financial data, consensus figures, capacity estimates and share price information cited here reflect publicly available sources at the time of publication and may change at any time, so the latest official disclosures from the relevant companies and institutions should be treated as authoritative. Readers should conduct their own research and make decisions based on their own financial circumstances, investment objectives and risk tolerance, consulting a qualified professional where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect loss arising from the use of this information.

 

 

James Mitchell specializes in technical analysis, market trends, and trading strategies for both Bitcoin and altcoins. Based in London, he has over 10 years of experience in financial markets. Before joining MEXC Learn, James worked as a senior analyst at a leading European investment firm, where he developed expertise in risk management and quantitative trading. His transition to cryptocurrency markets began in 2017, and he has since become recognized for his data-driven approach. He holds a Master's degree in Financial Economics from the London School of Economics. His analytical approach combines traditional technical analysis with on-chain metrics to provide readers with actionable insights.

 

Areas of Expertise: Technical Analysis, Market Trends and Cycles, Trading Strategies, Bitcoin and Altcoin Analysis, Risk Management.

 

 

 

Want the fastest access to MEXC's latest updates? Join our official Telegram group now!

MEXC Platform: Download App | Zero Fee Trading | Buy USDT Now

Join MEXC Community: X (Twitter) | Telegram | Discord

Get Insights: MEXC Learn | Terms of Service | MEXC Blog | MEXC News

Account Verification: Understand KYC | How to Complete KYC

External Content Platforms: Substack | Medium | Paragraph | LinkedIn | X(News)