There are ordinary earnings reports, and then there are Taiwan Semiconductor Manufacturing Company earnings reports—the quarterly moment when much of the technology industry gathers around one company’s numbers and attempts to determine whether the artificial-intelligence boom is still a historic infrastructure cycle or merely the most expensive group project in corporate history.
TSMC trades in the United States under the ticker TSM, but I do not think of it as just another semiconductor stock. I see it as the industrial foundation beneath an astonishing amount of modern computing. Other companies design the glamorous chips, announce them beneath theatrical lighting, give them aggressive names, and explain how they will transform civilization. TSMC performs the slightly less cinematic task of actually manufacturing many of them.
Without that manufacturing capacity, the grand AI revolution becomes an attractive slide presentation waiting for hardware.
That is what makes the next earnings report so important. I am not merely watching whether TSMC beats an earnings estimate by a few cents. I am watching the physical limits of the technology boom: chip demand, advanced-node capacity, packaging constraints, manufacturing costs, overseas expansion, and the uncomfortable question of how long a company can spend tens of billions of dollars annually before investors begin asking whether every new factory will earn an adequate return.
After a powerful second quarter, expectations are already standing on a ladder. TSMC is guiding for another sharp increase in revenue, AI demand remains intense, and management has raised its 2026 capital budget to an extraordinary $60 billion to $64 billion.
That is not a casual investment plan. That is the sort of spending normally associated with governments, energy systems, or villains building elaborate underground facilities.
TSMC is spending it because customers want more leading-edge capacity, and they apparently want it yesterday.
As I look toward the company’s third-quarter results, three issues dominate my thinking: whether advanced-chip demand remains as strong as management expects, how the 2-nanometer ramp and overseas factories affect margins, and whether TSMC’s enormous capacity expansion is building the next decade of growth or simply constructing a very expensive monument to current enthusiasm.
The Starting Point Is Already Ridiculous
Before previewing the next quarter, I need to appreciate what TSMC just accomplished.
In the second quarter of 2026, revenue reached $40.2 billion, up 12% sequentially and landing at the high end of management’s guidance. Gross margin climbed to 67.7%, while operating margin reached 60.3%. High-performance computing, which includes much of the AI-related demand investors care about, rose 20% from the previous quarter and accounted for 66% of total revenue.
Advanced process technologies—defined by TSMC as 7 nanometers and below—generated 77% of wafer revenue. The 3-nanometer node contributed 30%, 5-nanometer accounted for 33%, and even the new 2-nanometer process contributed 3% as it began ramping.
Those figures come from TSMC’s second-quarter earnings conference, and they tell me something more important than whether the company cleared a spreadsheet estimate: the revenue mix is moving decisively toward the most advanced, strategically important, and technically demanding parts of semiconductor manufacturing.
TSMC is not growing because the world suddenly developed an urgent shortage of chips for digital alarm clocks. Its growth is being driven by leading-edge computing.
The company’s June revenue provided an especially dramatic illustration. TSMC reported NT$442.68 billion in revenue for the month, up 67.9% from June 2025. Revenue for the first half of 2026 totaled approximately NT$2.4 trillion, an increase of 35.6% from the previous year, according to the company’s June revenue report.
A 67.9% monthly increase from a company of this scale is the sort of number that causes investors to check the decimal point.
The difficulty is that spectacular performance does not lower expectations. It feeds them.
Once a company demonstrates that it can grow at this pace while producing a gross margin near 68%, investors do not politely applaud and return to reasonable assumptions. They immediately ask what the company plans to do for an encore.
Preferably while juggling.
What TSMC Has Already Told Us About the Third Quarter
TSMC did not leave investors entirely dependent on imagination. Management expects third-quarter revenue between $44.6 billion and $45.8 billion.
At the midpoint, that would represent approximately 12% sequential growth and 37% year-over-year growth. Gross margin is expected between 65% and 67%, while operating margin is projected between 56% and 58%.
These are remarkable targets.
A company guiding toward approximately $45 billion in quarterly revenue is not participating at the edge of the AI infrastructure cycle. It is standing in the middle of the highway collecting tolls from nearly everyone trying to drive through it.
Still, the gross-margin guidance introduces the first real tension in the preview.
The midpoint of 66% would be lower than the 67.7% achieved in the second quarter. Ordinarily, rising revenue and higher utilization can support stronger margins. TSMC, however, is simultaneously ramping a new process technology and expanding manufacturing outside Taiwan—two activities that have a habit of consuming money with the quiet efficiency of teenagers discovering online shopping.
The third quarter therefore appears likely to deliver strong growth with some deliberate margin pressure.
That is not necessarily bad. I would rather see a company accept near-term costs to support durable demand than protect one quarter’s margin by refusing to build what customers need. But “temporary investment” is one of corporate finance’s most flexible phrases. It can describe a sensible growth program or an expense that plans to remain until everyone stops asking about it.
I will be listening for evidence that the margin headwinds are proceeding as expected rather than expanding.
AI Demand Remains the Main Event
The central question is straightforward: Is the demand real?
So far, every available signal suggests that it is.
Cloud providers, chip designers, and technology companies continue investing heavily in AI infrastructure. The leading AI accelerators require advanced manufacturing processes, sophisticated packaging, high-bandwidth memory integration, and enormous supporting ecosystems. TSMC occupies a critical position across much of that chain.
Management’s confidence has not been subtle.
During the second-quarter call, CEO C.C. Wei described AI demand as becoming “stronger and stronger and stronger.” I would not normally treat repetition as detailed financial guidance, but executives rarely raise capital spending by billions of dollars because customers sent them encouraging emojis.
TSMC increased its full-year 2026 capital budget to between $60 billion and $64 billion. At the beginning of the year, the range had been $52 billion to $56 billion. In April, the company indicated spending would likely come near the high end of that range. By July, even the high end was no longer high enough.
Management said the increase reflected rising customer demand as well as inflation in equipment costs. Customers are pushing for additional capacity, particularly in advanced technologies.
That is the bullish interpretation: TSMC has direct visibility into customer road maps, orders, product development, and capacity requirements. It does not have to guess about semiconductor demand by counting mentions of artificial intelligence on earnings calls. It speaks directly with the companies designing the chips.
If TSMC is spending more, it is probably because major customers are asking for more.
The cautious interpretation is that customers can also become overly enthusiastic. The semiconductor industry has a proud historical tradition of ordering too much capacity near the top of a cycle and then developing a sudden philosophical interest in inventory reduction.
AI demand may be structurally different from earlier cycles, but the phrase “this time is different” has personally escorted many investors into disappointing places.
That is why I will be looking beyond the headline growth. I want to hear whether demand is broadening across multiple customers and product categories or becoming increasingly concentrated among a small group of hyperscalers and AI-chip designers.
High demand is good.
High demand dependent upon a handful of customers spending extraordinary amounts indefinitely is still good, but it comes with a footnote large enough to require its own desk.
Customer Concentration Deserves Attention
TSMC’s position gives it extraordinary leverage, but it also ties its results to the ambitions of a relatively small group of massive technology companies.
When a few customers become larger, their orders can fill factories, drive advanced-node adoption, and support excellent economics. They can also create concentration risk.
If one major customer changes its architecture, delays a product, reduces spending, or shifts part of its sourcing strategy, the effects can move through TSMC’s plans.
I do not expect management to provide a detailed customer-by-customer forecast. Semiconductor companies treat customer identities and volumes with the discretion normally found in international espionage. Still, the language surrounding demand matters.
I want to hear whether strength extends beyond training accelerators into inference chips, custom silicon, networking processors, CPUs, and edge-computing applications. I want evidence that AI demand is becoming an ecosystem rather than remaining a narrow spending surge centered on a few products.
TSMC’s high-performance-computing platform already represents 66% of revenue. That is an enormous share, and it demonstrates how dramatically the business mix has changed.
Smartphones represented 22% of second-quarter revenue. Automotive was 4%, Internet of Things was 5%, and digital consumer electronics contributed only 1%.
A decade ago, smartphone cycles often dominated leading-edge semiconductor discussions. Today, AI and high-performance computing have moved to the center of the room, taken the best chair, and started adjusting the thermostat.
The smartphone business remains important, especially as new devices adopt advanced nodes. But it is no longer the primary growth story.
For the next earnings report, I will pay close attention to any commentary about non-AI markets. A recovery in smartphones, automotive, or industrial demand could provide additional support. Weakness in those areas may not derail TSMC while AI remains strong, but it would make the company more dependent on a single powerful theme.
Diversified growth is healthier than heroic growth being carried by one extremely muscular customer category.
The 2-Nanometer Ramp Is Both Opportunity and Expense
TSMC’s 2-nanometer technology is beginning to contribute revenue, and its expansion will be one of the most important parts of the upcoming report.
New process nodes are central to TSMC’s competitive position. Customers move to smaller, more advanced nodes to gain improvements in performance, power efficiency, and transistor density. Those advantages matter enormously for AI accelerators, smartphones, and high-performance computing.
They also do not materialize because someone reduced the font size on a product road map.
Developing and manufacturing a new node requires enormous investment, years of work, expensive equipment, process refinement, and a gradual improvement in production yields. Early in the ramp, costs are high and volumes are lower. As production matures, yields improve, utilization rises, and the economics become more attractive.
Management expects the steep N2 ramp to reduce gross margin by approximately three to four percentage points during the second half of 2026. That pressure is a major reason third-quarter gross-margin guidance falls below the second-quarter result.
I do not view this dilution as a sign that the core business is deteriorating. It is the cost of moving the technology frontier forward.
The important questions involve execution.
Is customer adoption proceeding as planned?
Are yields improving?
Is the ramp creating any unexpected production difficulties?
Will the revenue contribution rise meaningfully in the third and fourth quarters?
How quickly can the margin burden moderate?
If management sounds confident about N2 yields and customer demand, I can tolerate a temporary margin decline. If the ramp experiences delays or higher-than-expected costs, the market may become less forgiving.
TSMC has earned a reputation for manufacturing execution. Investors effectively pay for that reliability. When a new node ramps smoothly, the company strengthens its competitive advantage. When execution slips, the problem is not limited to one quarter’s profitability. Customers build multiyear product plans around the availability of these technologies.
A delay can disrupt entire road maps.
No pressure.
It is only one of the most complicated manufacturing transitions humanity knows how to perform.
Gross Margin Is the Number Everyone Will Pretend Not to Obsess Over
Revenue receives the headlines, but margin quality often determines how investors interpret the quarter.
TSMC’s second-quarter gross margin of 67.7% was exceptional. The third-quarter midpoint of 66% remains exceptional. Most manufacturers would celebrate those economics by commissioning a statue.
Investors, however, evaluate companies relative to expectations, not emotional gratitude.
If TSMC reports revenue near the high end of guidance but gross margin slips below 65%, the market may focus on the cost pressure. If revenue is strong and margin remains near 67%, investors may conclude that leading-edge demand, utilization, pricing, and manufacturing efficiency are offsetting the expensive N2 and international expansions better than anticipated.
Several forces will shape the outcome.
Higher factory utilization should help. Cost improvements and cross-node capacity optimization should help. Strong demand for advanced processes should support pricing and product mix.
Against those benefits, the N2 ramp will weigh on profitability. Overseas factories will dilute margins. Foreign-exchange movements can help or hurt. Electricity, labor, materials, and equipment costs remain relevant.
Management expects overseas expansion to reduce gross margin by two to three percentage points during the early stages, with dilution potentially widening to three to four points later as the international manufacturing footprint grows.
That is not a minor rounding error.
TSMC’s global expansion improves geographic diversification and places more advanced manufacturing closer to major customers. It also means operating in regions with higher construction, labor, regulatory, and supply-chain costs than Taiwan.
Investors want geographic resilience, domestic manufacturing, and endless capacity. They would simply prefer all three to arrive without affecting margins, capital spending, or free cash flow.
This is the traditional investor request: please build the future, but do not let the future become expensive.
Arizona Is About More Than One Factory
TSMC’s United States expansion has become one of the largest industrial projects in the semiconductor industry.
The company previously announced plans to increase its total planned investment in the United States to $165 billion. The broader Arizona strategy includes three additional fabrication plants, two advanced-packaging facilities, and an R&D center, on top of projects already underway, according to TSMC’s expansion announcement.
The scale is difficult to overstate.
TSMC is attempting to replicate portions of a highly developed semiconductor ecosystem outside Taiwan. A fabrication plant does not operate in isolation. It needs suppliers, engineers, technicians, chemicals, gases, equipment, utilities, logistics, process knowledge, and a workforce trained to perform incredibly precise work.
Constructing the building is the visible part.
Creating the manufacturing culture and supplier network is the hard part.
The first Arizona fab entered high-volume production in the fourth quarter of 2024. Construction of the second fab has been completed, with tool installation planned during 2026. TSMC has also indicated that it is accelerating aspects of its Arizona expansion in response to customer demand.
I will be listening for updates on timelines, tool installation, production yields, customer commitments, staffing, and advanced-packaging development.
Arizona is strategically important because the United States wants more domestic semiconductor manufacturing, and TSMC’s leading customers want supply-chain resilience. But these plants must eventually generate acceptable returns. Strategic significance does not exempt a factory from arithmetic.
The optimistic case is that strong demand fills the fabs, customers accept pricing that reflects higher costs, subsidies and partnerships reduce the burden, and TSMC builds a valuable second manufacturing center.
The less cheerful case is that overseas fabs remain structurally more expensive, weigh on margins longer than expected, and require continuous investment before reaching efficient scale.
Both outcomes can contain successful production. The difference lies in profitability.
Advanced Packaging May Be as Important as Wafer Capacity
The semiconductor conversation often focuses on process nodes: 5 nanometers, 3 nanometers, 2 nanometers, and whatever increasingly tiny measurement appears next.
But manufacturing advanced wafers is only part of the challenge.
Modern AI systems require multiple components to be integrated into sophisticated packages. TSMC’s chip-on-wafer-on-substrate technology, commonly called CoWoS, has become especially important for combining advanced processors with high-bandwidth memory.
For much of the AI boom, packaging capacity has been a constraint. Producing more wafers does not solve the problem if those wafers cannot be packaged into finished systems at the required rate.
This is why I pay close attention to TSMC’s advanced-packaging investment.
The company plans to allocate roughly 10% to 20% of its 2026 capital budget to advanced packaging, testing, mask making, and related areas. Another 70% to 80% is expected to support advanced process technologies, while approximately 10% will go toward specialty technologies.
Those allocations show that TSMC understands the bottleneck is broader than wafer fabrication.
During the upcoming report, I want to know whether CoWoS supply is catching up with demand, whether customer requirements are continuing to increase, and how quickly additional packaging capacity can become productive.
There is an irony in spending billions to manufacture some of the world’s most sophisticated processors only to discover they are waiting in line for packaging. It is the semiconductor equivalent of building a jet and realizing the airport has one gate.
If packaging constraints ease while demand remains strong, TSMC can convert more customer orders into revenue. If demand continues outrunning capacity, the company may retain strong pricing and visibility, but near-term shipments could remain constrained.
Either way, advanced packaging is no longer a technical side note. It is central to the growth story.
Capital Spending Is a Signal—and a Risk
TSMC’s $60 billion to $64 billion capital budget may be the single most revealing number in the entire earnings setup.
Management has said that higher capital spending tends to correlate with stronger growth opportunities in subsequent years. That makes sense. Semiconductor capacity takes years to plan, build, equip, qualify, and ramp. TSMC must invest before the associated revenue arrives.
The spending therefore represents a statement about future demand.
It is also a commitment made under uncertainty.
At the midpoint, TSMC expects to spend approximately $62 billion in 2026. That is more than the annual revenue of many large public companies. It is being deployed across advanced nodes, packaging, specialty technologies, and international expansion.
For the bullish investor, this is evidence that TSMC sees a multiyear AI and high-performance-computing cycle. Customers are requesting more capacity, and the company is preparing to meet that demand.
For the cautious investor, the spending raises questions about free cash flow, depreciation, return on invested capital, and the possibility of overbuilding.
Both views deserve respect.
Semiconductor manufacturing punishes hesitation. A company that waits for demand to become undeniable will discover that the necessary capacity is still several years away.
It also punishes overconfidence. Capacity built for demand that fails to materialize becomes an extremely expensive collection of clean rooms.
TSMC must therefore make decisions today about markets that will exist years from now. It has to forecast customer needs, product cycles, AI adoption, competitive technology, and global policy while equipment prices rise and governments develop sudden, passionate opinions about chip geography.
I will watch whether management raises, maintains, or narrows its capital-spending expectations. Another upward revision would reinforce the demand signal but could also provoke questions about how much spending the growth cycle ultimately requires.
At some point, “customers want more capacity” must translate into durable cash generation.
The bulldozers cannot be the entire investment thesis.
The Competitive Position Still Looks Formidable
TSMC’s greatest advantage is not simply that it owns factories. It operates an ecosystem.
Customers rely on its manufacturing processes, design libraries, intellectual property partners, packaging technologies, production reliability, and years of accumulated expertise. Moving an advanced chip from one foundry to another is not like changing grocery stores because the avocados looked judgmental.
It can require extensive redesign, validation, and risk.
TSMC continues developing future nodes beyond N2. The company expects its A14 technology to enter risk production in 2027 and volume production in 2028. Management says A14 should offer meaningful performance, power-efficiency, and density improvements compared with N2.
The road map matters because the semiconductor industry does not award permanent leadership.
Competitors are investing aggressively. Governments are subsidizing domestic production. Customers have incentives to diversify supply. The cost of staying ahead continues increasing.
TSMC’s capital spending therefore serves two purposes. It meets demand, and it defends technological leadership.
The company must keep customers satisfied while charging enough to fund an increasingly expensive road map. Management has emphasized that it prices based on the value it provides while avoiding sudden increases that could damage customers’ competitiveness.
That balance is important.
TSMC’s customers need its technology, but TSMC also needs successful customers capable of selling profitable products. Extracting every available dollar today could weaken the ecosystem tomorrow.
A strong foundry does not merely rent manufacturing space. It becomes a long-term partner in customer product development.
Risks I Will Not Ignore
The AI opportunity is enormous, but TSM is not a risk-free way to purchase a ticket.
The first risk is demand normalization. AI infrastructure spending cannot grow at extraordinary rates forever. Even if the long-term opportunity remains intact, customers may experience digestion periods after large investment waves.
The second is customer concentration. A small number of buyers increasingly influence leading-edge demand.
The third is execution. N2, advanced packaging, Arizona, and other international expansions must all ramp effectively.
The fourth is margin pressure. Overseas manufacturing, rising depreciation, new-node costs, inflation, and foreign exchange can offset the benefits of strong revenue.
The fifth is geopolitics. TSMC’s importance and its location in Taiwan make geopolitical risk impossible to exclude from the investment case. Anyone valuing the company purely from an income statement is studying a three-dimensional object with one eye closed.
The sixth is valuation. A great company can become a disappointing investment when purchased at a price assuming flawless performance until the sun burns out.
I am not claiming any of these risks must derail the story. I am saying they belong in it.
Enthusiasm becomes speculation when it stops acknowledging what can go wrong.
What I Will Watch When TSMC Reports
When the next earnings report arrives, my checklist will be fairly simple.
First, I want revenue near or above the $44.6 billion to $45.8 billion guidance range. Stronger results would confirm that advanced-chip demand remains robust.
Second, I want gross margin within the 65% to 67% range. A result near the high end would be especially encouraging given the N2 and overseas-fab pressure.
Third, I will examine the revenue mix. HPC growth, advanced-node contributions, and N2 adoption should reveal where demand is strongest.
Fourth, I want an update on advanced packaging. Additional CoWoS capacity must translate into greater output.
Fifth, I will listen closely to capital-spending commentary. The $60 billion to $64 billion budget needs to be supported by specific, durable customer demand.
Sixth, I want evidence that Arizona and other overseas projects are advancing without major delays or cost surprises.
Finally, I will focus on management’s tone regarding 2027. Quarterly results matter, but the stock’s larger value depends on whether AI-related semiconductor demand remains a multiyear infrastructure build.
My Bottom Line
I approach TSMC’s next earnings report with optimism tempered by respect for how much optimism is already embedded in the story.
The company is delivering extraordinary growth. It sits at the center of advanced semiconductor manufacturing. Its customers are demanding more capacity, its leading-edge technologies dominate the revenue mix, and management is investing aggressively to support what it sees as a structural expansion in AI and high-performance computing.
At the same time, the spending is enormous, margin headwinds are real, customer concentration deserves attention, and global expansion introduces costs that cannot be removed through an enthusiastic PowerPoint transition.
The next quarter does not need to prove that TSMC is an important company. That debate ended long ago.
It needs to demonstrate that demand remains strong enough to justify the expansion, that margins can absorb the cost of technology leadership, and that management is adding capacity with discipline rather than chasing every forecast wearing an AI badge.
I expect the headline numbers to be strong. The more important information will live underneath them.
Are customers still pressing for more capacity?
Is N2 ramping smoothly?
Is advanced packaging becoming less constrained?
Are overseas factories progressing toward efficient scale?
Can TSMC continue producing margins that would make most manufacturers stare silently into the distance?
Those answers will tell me more than whether earnings beat an estimate by a few percentage points.
TSMC is building the factories, packaging systems, and manufacturing processes that could define the next era of computing. The opportunity is extraordinary because the task is extraordinarily difficult.
That is the investment case in one sentence.
Everyone wants more advanced chips. TSMC is one of the few companies capable of producing them at scale. Now it must expand that scale without allowing costs, complexity, or investor expectations to become more advanced than the chips themselves.
This article reflects my analysis and is intended for informational purposes, not personalized investment advice.
Comments
Post a Comment