Every time Taiwan Semiconductor Manufacturing reports another remarkable quarter, I ask myself the same question: How can one company be simultaneously one of the strongest businesses in the world and one of the most uncomfortable stocks to own?
The answer is simple. TSMC manufactures the most important chips on Earth from an island that sits at the center of one of the most dangerous geopolitical rivalries on Earth.
That creates an investment capable of inspiring admiration and indigestion at the same time.
TSMC is the manufacturing engine behind much of modern computing. It produces advanced semiconductors for artificial intelligence systems, smartphones, data centers, automobiles, communications equipment, and countless devices most of us use without thinking about the microscopic engineering inside them.
When a technology company announces a revolutionary processor, investors usually celebrate the company whose name appears on the presentation. TSMC is often the business actually responsible for turning the design into a physical product.
It is the quiet factory behind the loudest technology trends.
The company has also become one of the clearest beneficiaries of the artificial intelligence boom. Demand for advanced computing power continues to grow, leading customers to compete for TSMC’s most sophisticated manufacturing capacity.
Unfortunately, the stock market has noticed.
As of July 22, 2026, TSMC’s American depositary receipts traded around $421. The shares had risen dramatically over the preceding year and were no longer valued like a misunderstood industrial company. Investors were paying a substantial price for TSMC’s technological leadership, extraordinary margins, and expected growth.
That brings me back to the central question: Is TSMC still a buy for long-term investors?
My answer is yes—but with a major qualification.
I still consider TSMC a compelling long-term business. I do not consider it a stock that should be purchased recklessly at any price. At current levels, I would build a position gradually, preserve cash for volatility, and accept that geopolitical fear could produce severe declines even if the company continues reporting excellent results.
The investment case remains powerful. The margin of safety has become less generous.
TSMC Is Not Just Another Semiconductor Company
To understand TSMC, I have to begin with its business model.
TSMC is a pure-play semiconductor foundry. It manufactures chips designed by other companies. Its customers can focus on architecture, software, and product development without building fabrication plants that cost tens of billions of dollars and require almost absurd levels of technical precision.
This separation created the modern “fabless” semiconductor industry.
Companies can design advanced chips without owning the factories that manufacture them. TSMC provides the fabrication expertise, production capacity, process technology, packaging services, and manufacturing reliability needed to transform digital designs into physical silicon.
That sounds straightforward until I consider what advanced semiconductor manufacturing actually involves.
A leading-edge fabrication plant must manipulate materials at dimensions measured in nanometers. It relies on extraordinarily complex equipment, chemical processes, ultrapure water, stable electricity, sophisticated software, specialized engineers, and a supply chain distributed across several continents.
The process requires enormous capital, years of development, and relentless quality control. A tiny manufacturing defect can destroy the usefulness of a chip containing billions of transistors.
I sometimes struggle to install a screen protector without trapping a piece of dust beneath it. TSMC manufactures processors at scales where a particle too small for me to see can become a significant operational problem.
The company’s advantage is not simply that it owns expensive factories. Competitors can raise capital and buy equipment. TSMC’s strength comes from decades of accumulated manufacturing knowledge, process integration, customer relationships, engineering talent, production data, intellectual property, and operational discipline.
That combination is extraordinarily difficult to reproduce.
The Latest Results Were Exceptional
TSMC’s second-quarter 2026 results were not merely good. They were the sort of numbers that make financial spreadsheets look as though they have developed a caffeine dependency.
The company reported quarterly revenue of NT$1.27 trillion, equivalent to approximately $40.2 billion. Revenue increased 36% from the previous year and 12% from the first quarter.
Net income reached NT$706.56 billion. Diluted earnings per share rose to NT$27.25, equivalent to $4.31 per American depositary receipt. Net income and earnings per share increased 77.4% year over year.
Gross margin reached 67.7%. Operating margin was 60.3%, and net profit margin was 55.6%.
Those figures deserve a pause.
TSMC converted more than half of its quarterly revenue into net profit. Many companies hold meetings to discuss how they might someday achieve a respectable operating margin. TSMC produced a net margin above 55% while investing tens of billions of dollars in some of the most complicated factories ever built.
Management guided for third-quarter revenue between $44.6 billion and $45.8 billion. The company expected a gross margin of 65% to 67% and an operating margin of 56% to 58%. Continued demand for advanced process technologies and the rapid ramp of its 2-nanometer technology were expected to support the quarter. TSMC’s second-quarter earnings release provides the complete results and guidance.
This does not look like a company struggling to find demand.
It looks like a company attempting to build capacity quickly enough to satisfy customers who increasingly depend on it.
Advanced Manufacturing Is the Center of the Story
During the second quarter, 2-nanometer products accounted for 3% of wafer revenue, despite being in the early stages of production. Three-nanometer technology contributed 30%, 5-nanometer contributed 33%, and 7-nanometer contributed another 11%.
Combined, technologies at 7 nanometers and below accounted for 77% of total wafer revenue.
That number tells me several things.
First, TSMC is not relying primarily on old production lines serving slow-growth markets. A large majority of its wafer revenue comes from advanced technologies.
Second, customers are moving rapidly toward newer manufacturing processes. Advanced nodes can offer improvements in performance, power efficiency, and transistor density, which are essential for artificial intelligence accelerators, smartphones, and high-performance computing.
Third, TSMC has demonstrated that it can turn technological leadership into commercial volume.
Laboratory breakthroughs are impressive, but shareholders cannot deposit a research paper into a brokerage account. TSMC’s real advantage is its ability to develop advanced processes and then manufacture enormous quantities of chips with yields customers can depend on.
The company’s 2-nanometer ramp is especially important. If it proceeds successfully, it should reinforce TSMC’s lead and create another cycle of customers migrating toward more advanced, more valuable manufacturing.
Technology leadership is never permanent, but TSMC has repeatedly shown an ability to move the finish line while competitors are still running toward it.
Artificial Intelligence Is Fueling a Powerful Expansion
The artificial intelligence buildout has changed TSMC’s growth profile.
Training and running advanced AI models require enormous computing power. That demand supports sales of accelerators, networking equipment, custom silicon, memory systems, servers, and related infrastructure.
Many of those products require advanced manufacturing and packaging.
TSMC does not have to predict which individual AI application will dominate. It manufactures chips for many of the companies competing to build the infrastructure. This gives it a position resembling a supplier of essential equipment during a gold rush—except the shovels contain billions of transistors and require a factory clean enough to make an operating room look casual.
This does not make TSMC immune to an AI downturn.
If technology companies reduce capital spending, delay data-center projects, or discover that expected AI revenue cannot justify infrastructure costs, chip demand could slow. Semiconductor customers may also overorder when capacity is tight, creating the risk of an inventory correction later.
But I do not believe the AI opportunity depends on one product cycle.
Artificial intelligence is moving into data centers, personal computers, smartphones, industrial systems, vehicles, robotics, healthcare devices, communications equipment, and consumer products. Not every proposed use will become profitable. Some will disappear after investors recover from their excitement.
The broader demand for computing, however, is likely to continue growing.
TSMC benefits not only from the number of chips sold but also from their increasing complexity. More advanced processors require leading-edge manufacturing, sophisticated packaging, and deeper collaboration between designers and manufacturers.
TSMC sits directly in that path.
The Customer Relationship Creates a Powerful Moat
TSMC’s business depends on trust.
A semiconductor designer may spend years and billions of dollars developing a new processor. Choosing a manufacturing partner is not the same as switching suppliers for office paper.
The chip design must be adapted to the foundry’s manufacturing process. Engineers use specialized design tools, intellectual-property libraries, process rules, packaging technology, and testing procedures. The customer and foundry work together long before mass production begins.
Once that relationship is established, changing manufacturers can be expensive, time-consuming, and technically risky.
TSMC has also built its identity around not competing directly with its customers. Because it focuses on contract manufacturing, chip designers can share sensitive product information without worrying that their foundry will use that knowledge to strengthen a competing chip business.
That neutrality has helped create a broad ecosystem.
In 2025, TSMC manufactured 12,682 products for 534 customers using 305 different process technologies. That scale gives the company enormous manufacturing experience and spreads development costs across a wide customer base.
This is where size becomes self-reinforcing.
More customers generate more production volume and data. Greater volume supports larger research and capital budgets. Better technology attracts more customers. The resulting cash flow funds the next generation of factories and process development.
Competitors are not trying to catch a stationary company.
They are trying to catch a company investing more than many corporations earn in total revenue.
Capital Spending Is Both a Strength and a Risk
TSMC expects its 2026 capital budget to reach between $52 billion and $56 billion. Earlier in the year, management indicated spending would likely be near the high end of that range.
Approximately 70% to 80% of the budget is allocated to advanced process technologies. Other spending supports specialty technologies and advanced packaging, testing, and related infrastructure.
This enormous investment tells me management sees durable demand. Semiconductor fabrication capacity cannot be built instantly. Management must commit capital years before the final revenue arrives.
The spending also strengthens TSMC’s competitive position. Few companies can fund leading-edge research and fabrication at this scale while maintaining an exceptionally strong balance sheet.
TSMC describes its balance sheet as a fortress, supported by high credit ratings. Its investor materials state that since its public listing, the company has produced revenue and earnings compound annual growth rates of approximately 18.6% and 18.7%, respectively. Its strategic targets for 2024 through 2029 include approaching 25% annual revenue growth in U.S. dollar terms, maintaining gross margin at 56% or higher, and generating return on equity in the high-20% range through the cycle. TSMC’s investor overview outlines those long-term objectives.
Still, capital spending creates risk.
Factories can be underutilized if demand slows. New plants may cost more than expected. Overseas operations may generate lower margins than the company’s facilities in Taiwan. Currency movements, labor expenses, utilities, regulatory requirements, and construction delays can all reduce returns.
The company must make enormous decisions based on forecasts about technologies and demand several years into the future.
Even the world’s best manufacturing company does not own a crystal ball. If it did, the crystal ball would presumably require an advanced TSMC processor.
The Global Expansion Is Necessary but Expensive
TSMC is expanding outside Taiwan, particularly in the United States.
The company’s planned American investment has grown to approximately $165 billion. Its Arizona plans include six semiconductor fabrication facilities, two advanced packaging facilities, and a research-and-development center. TSMC’s Arizona overview describes the scale of that expansion.
The strategy makes sense.
Customers and governments want geographically diversified semiconductor production. The pandemic, global chip shortage, trade restrictions, and rising tension between China and the United States exposed the vulnerability of concentrating advanced manufacturing in one location.
Manufacturing more chips in the United States can strengthen customer relationships, improve supply-chain resilience, reduce political pressure, and provide access to government incentives.
But manufacturing in Arizona is unlikely to match the cost structure of TSMC’s Taiwanese operations immediately.
Labor, construction, supply-chain maturity, and operating expenses may be higher. The company must recreate an ecosystem that developed around Taiwan’s semiconductor industry over decades.
The overseas expansion may therefore pressure margins.
I view that cost as a form of strategic insurance. Insurance rarely feels cheap when purchased, but it seems brilliant after the roof disappears.
Investors should not expect diversification to eliminate geopolitical risk. Taiwan will remain central to TSMC’s advanced manufacturing for years. The international expansion reduces concentration gradually; it does not teleport the company’s entire production network to safety.
The Taiwan Risk Cannot Be Treated Like a Footnote
No serious TSMC analysis is complete without discussing China and Taiwan.
China claims Taiwan as part of its territory and has not renounced the use of force. Military exercises, political tension, cyber activity, trade restrictions, and competition between China and the United States create a risk that cannot be modeled neatly in a spreadsheet.
A blockade, invasion, or sustained disruption could have catastrophic consequences for TSMC, its customers, the semiconductor industry, and the global economy.
This is not an ordinary business risk.
If a restaurant opens too many locations, I can estimate the effect on margins. If a retailer misjudges inventory, I can examine the markdowns. If military conflict interrupts the world’s most important concentration of advanced semiconductor manufacturing, the usual valuation model becomes decorative.
Some investors assume TSMC’s importance protects it because every major economy depends on its output. That dependence may discourage conflict, but I would never treat it as a guarantee.
Economic self-interest has failed to prevent wars before. Human history did not become rational merely because supply chains grew complicated.
Diversification into Arizona, Japan, and Europe helps. So do efforts by governments to preserve stability. But the geopolitical discount on TSMC exists for a legitimate reason.
I would not invest money in TSMC that I could not tolerate seeing decline sharply. I would also avoid allowing it to become an irresponsibly large percentage of my portfolio, regardless of how much I admire the company.
A wonderful business can still inhabit a dangerous neighborhood.
Competition Remains Real
TSMC leads the global foundry industry, but it does not operate alone.
Samsung continues investing in advanced manufacturing. Intel is attempting to establish a significant foundry business. Governments are providing subsidies to expand domestic semiconductor production. Other foundries compete in mature and specialty processes.
Competition could pressure pricing, attract customers seeking alternative suppliers, or narrow TSMC’s technological advantage.
I am not dismissive of those efforts. Semiconductor manufacturing is strategically important, and governments are willing to spend heavily to reduce dependence on Taiwan.
However, money alone does not guarantee manufacturing success.
A leading-edge foundry must deliver performance, power efficiency, density, yield, volume, reliability, design support, packaging, and predictable schedules. Customers cannot sell excuses inside their products. A processor that arrives late or fails to meet specifications can damage an entire product cycle.
TSMC’s record gives customers confidence.
Competitors may improve, and some will win meaningful business. The total semiconductor market may grow enough to support several major manufacturers. But replacing TSMC at the leading edge would require more than building factories. It would require replicating a deeply integrated ecosystem and decades of accumulated expertise.
That is possible over time. It is not easy.
Customer Concentration Deserves Attention
TSMC serves hundreds of customers, but a limited number of enormous technology companies account for a meaningful share of demand.
That concentration has advantages. The largest customers possess the resources to develop advanced processors and order enormous volumes. Their products help fund TSMC’s next generation of technology.
It also creates dependence.
If a major customer loses market share, shifts some manufacturing elsewhere, changes product strategy, or develops an alternative supply arrangement, TSMC could feel the impact.
Export restrictions add another complication. Government rules may limit which advanced chips can be sold to specific countries or customers. TSMC must navigate restrictions imposed by multiple jurisdictions while protecting commercial relationships.
Artificial intelligence has made semiconductors a national-security issue. Once politicians discover that a product is strategically important, the rulebook tends to grow rapidly and develop several appendices.
I expect regulatory complexity to remain a permanent part of TSMC’s business.
The Valuation Is the Main Reason I Would Buy Gradually
At approximately $421 per ADR, TSMC is no longer obviously cheap.
Valuation estimates differ depending on earnings forecasts, exchange-rate assumptions, and data providers, but the stock recently traded around the mid-20s multiple of expected earnings. Its trailing valuation also sat well above many of its historical levels.
I do not consider that valuation absurd for a company growing revenue above 30%, expanding earnings much faster, generating exceptional margins, and occupying a central position in artificial intelligence infrastructure.
But the valuation leaves less room for disappointment.
A stock can decline even when the business performs well. If investors currently expect extraordinary growth and the company merely delivers strong growth, the earnings multiple can contract.
That is the strange cruelty of high expectations. A company can bring home excellent results and still be asked why they were not miraculous.
At the current price, I would not make one large purchase. I would divide my intended investment into several pieces.
I might begin with 25% to 35% of the desired position, add on meaningful market weakness, and continue buying over six to twelve months if the long-term thesis remains intact.
My rough valuation framework would look like this:
Below $360, I would consider the shares increasingly attractive, assuming no major deterioration in the business or geopolitical situation.
Between $360 and $400, I would view TSMC as a reasonable long-term accumulation candidate.
Between $400 and $450, I would buy cautiously and gradually.
Above $475, I would require stronger earnings estimates or a longer time horizon before adding aggressively.
These are not magical boundaries. A stock does not become precisely worth buying because it falls one dollar below an arbitrary number. The ranges simply help me prevent enthusiasm from replacing discipline.
The Dividend Is Helpful, but It Is Not the Main Attraction
TSMC pays a dividend, providing shareholders with income while they wait for long-term growth.
However, I would not purchase the stock primarily for yield. The company’s greatest opportunity lies in reinvesting capital into advanced manufacturing, packaging, global capacity, and research.
I want management to return excess cash, but I do not want a growing dividend to come at the expense of technological leadership.
In semiconductors, the company that stops investing does not become comfortably mature. It becomes obsolete with impressive speed.
The dividend is a welcome part of the total return. The primary investment case remains earnings growth and the company’s strategic position.
What Could Make Me Sell?
I would reconsider my thesis if TSMC lost its manufacturing lead across multiple generations, experienced persistent yield problems, or began losing major customers because competitors offered demonstrably better technology.
I would also watch margins closely. Temporary pressure from overseas expansion or new-process ramps would not automatically alarm me. A sustained decline caused by weaker pricing power, poor capital allocation, or structural cost problems would be more concerning.
A significant reduction in long-term AI and high-performance-computing demand would require reassessment, especially if customers accumulated excess inventory or reduced capital spending.
Geopolitical developments could also force a decision. I would not react to every military exercise or headline, but a clear change in the probability of blockade or conflict would alter the risk calculation.
Finally, valuation matters. If the stock rose far faster than earnings and began pricing in years of near-perfect execution, I would consider trimming.
I do not believe in falling in love with stocks. The stock will not remember my loyalty, send me a birthday card, or apologize for falling 40%.
My Long-Term Verdict
I rate Taiwan Semiconductor a cautious long-term buy.
The company possesses one of the strongest competitive positions in the global economy. Its manufacturing expertise, customer relationships, scale, advanced technology, financial strength, and role in artificial intelligence create a durable investment case.
The latest results reinforce that view. Revenue is expanding rapidly. Margins are exceptional. Two-nanometer production is ramping. Management is investing aggressively to support anticipated demand.
But the stock’s rise has reduced the margin of safety. TSMC now carries significant expectations alongside its permanent geopolitical risk.
For an investor with at least a five- to ten-year horizon, a diversified portfolio, and the emotional ability to tolerate severe volatility, I believe the shares remain worth accumulating.
For someone seeking a quick gain, a stable stock, or an investment untouched by global politics, TSMC may be a poor fit. This company can produce excellent financial results and still fall sharply because of a military headline, export rule, currency movement, or change in market sentiment.
My preferred approach is simple: buy in stages, avoid chasing rallies, keep the position appropriately sized, and use significant weakness to add if the business remains strong.
TSMC does not need every AI company to win. It needs computing demand to continue growing and customers to require increasingly advanced manufacturing. I believe both conditions are likely to persist.
The company sits at the point where artificial intelligence, smartphones, cloud computing, national security, industrial policy, and modern life converge.
That position is extraordinarily valuable.
It is also extraordinarily complicated.
I am willing to own that complexity—but only at a price and position size that allow me to sleep when the headlines inevitably become unpleasant.
Verdict: Long-term buy, preferably through gradual accumulation.
Risk level: High, primarily because of geopolitics, cyclicality, capital intensity, and valuation.
Time horizon: At least five years, preferably ten.
At roughly $421: Start small or hold existing shares; become more aggressive during meaningful pullbacks.
Stock prices and financial information are current as of July 22, 2026. This article is general commentary and not personalized financial advice.
Comments
Post a Comment