TSMC's 51% Net Margin Looks Like Software—Its $64 Billion Capital Budget Does Not
Taiwan Semiconductor Is Printing Money From AI. Now Comes the Expensive Part: Proving Today's Chip Shortage Won't Become Tomorrow's Overcapacity.
October 17, 2026 | Stock Analysis • Artificial Intelligence • Semiconductor Investing
I've followed enough earnings reports to know that occasionally a company delivers numbers so extraordinary that I find myself checking whether someone accidentally added an extra digit.
Taiwan Semiconductor Manufacturing Company (NYSE: TSM) just delivered one of those reports.
Third-quarter revenue surged 50.9% in New Taiwan dollar terms. Net income jumped 68.7%. Gross margins reached 66.4%, operating margins climbed to 58.7%, and net profit margins hit an almost ridiculous 51.1%.
Let me put that last number into perspective. For every $100 in revenue, TSMC retained roughly $51 in accounting profit after expenses and taxes.
That's the kind of profitability I normally associate with dominant software businesses, companies that can sell another subscription without constructing a multibillion-dollar industrial facility first.
TSMC, unfortunately, doesn't have that luxury.
Its customers cannot download additional semiconductor manufacturing capacity from the cloud.
When demand increases, TSMC needs more factories, more equipment, more engineers, more electricity, more sophisticated manufacturing processes, and the sort of capital expenditures that make ordinary corporate expansion plans look like a trip to the office supply store.
Management is planning to spend between $60 billion and $64 billion on capital expenditures in 2026 alone.
And the spending isn't expected to stop there.
TSMC anticipates that its cumulative capital expenditures from 2026 through 2028 will be significantly higher than those of the previous three years.
So here's the question I keep coming back to: What happens when a company earning software-like margins has to keep making industrial-scale investments to defend them?
Because that's the investment debate surrounding TSMC today.
The company has demonstrated that artificial intelligence is creating enormous demand for its manufacturing capabilities. What it hasn't demonstrated yet—because nobody can—is exactly how much of today's extraordinary profitability will survive after the next enormous wave of manufacturing capacity comes online.
And that distinction matters much more to long-term shareholders than one spectacular quarter.
Let's Start With the Numbers, Because They're Almost Absurd
TSMC's third-quarter results, released October 15, 2026, were exceptional by almost any reasonable financial standard.
| Financial metric | Q3 2026 result |
|---|---|
| Revenue | NT$1.494 trillion |
| Revenue growth (NT$) | +50.9% year over year |
| Revenue in U.S. dollars | $46.72 billion |
| Net income | NT$763.1 billion |
| Net income growth | +68.7% year over year |
| Earnings per ADR | $4.60 |
| Gross margin | 66.4% |
| Operating margin | 58.7% |
| Net profit margin | 51.1% |
| 2026 capital spending guidance | $60–$64 billion |
Source: TSMC third-quarter earnings release, October 15, 2026.
One detail worth clarifying is that the 50.9% revenue growth was reported in New Taiwan dollars. Measured in U.S. dollars, revenue increased 41.2%, reflecting exchange-rate differences.
Either way, those are remarkable growth rates for a business of this size.
We aren't talking about a tiny technology startup growing from $10 million to $15 million in annual revenue.
We're talking about a company generating nearly $47 billion in a single quarter while expanding at rates that many much smaller companies would struggle to achieve.
What impresses me even more is the profitability.
A 58.7% operating margin in semiconductor manufacturing is extraordinary. Fabricating advanced chips is one of the most technically demanding and expensive industrial activities on the planet.
Yet TSMC has transformed that complexity into an economic advantage.
Its customers need manufacturing capabilities that are extraordinarily difficult to reproduce, and TSMC is collecting exceptional profits for providing them.
That's a wonderful position to occupy.
But investors need to distinguish between exceptional current economics and permanently sustainable economics.
Those aren't necessarily the same thing.
TSMC Isn't Just Riding the AI Boom. It's Collecting the Manufacturing Toll.
When most investors think about artificial intelligence, they immediately think about Nvidia.
And I understand why.
Nvidia's processors have become central to the AI infrastructure buildout. Its technology and ecosystem have positioned the company as one of the most important beneficiaries of accelerated computing.
But every advanced chip needs to be manufactured somewhere.
That's where TSMC enters the picture.
TSMC manufactures semiconductors for major technology companies, including Nvidia, Apple, AMD, and others.
It provides the extraordinarily sophisticated fabrication capabilities that allow chip designers to turn their designs into physical products.
And the more powerful those chips become, the more difficult manufacturing them gets.
Advanced semiconductor manufacturing requires enormous precision, years of process development, sophisticated lithography equipment, and expertise that cannot simply be purchased overnight.
This creates a powerful competitive advantage.
A customer cannot casually switch to another manufacturer because someone offers a slightly lower price.
Moving an advanced chip design between manufacturing processes can require redesign, validation, engineering resources, and substantial time.
The manufacturing relationship itself becomes strategically important.
That gives TSMC something I love seeing in a business: pricing power supported by genuine technical differentiation.
Not pricing power created by marketing slogans.
Not artificial scarcity manufactured through some elaborate advertising strategy.
Actual scarcity arising from the difficulty of producing something customers desperately need.
And right now, AI companies need enormous quantities of advanced semiconductor capacity.
That demand is helping TSMC generate some of the strongest margins in its history.
But scarcity creates another problem.
It encourages expansion.
And expansion eventually changes the economics of scarcity.
Two-Nanometer Manufacturing Is Where Things Get Especially Interesting
One of the most revealing details in TSMC's earnings report was the growing contribution of its newest manufacturing technology.
Two-nanometer production accounted for 8% of total wafer revenue during the third quarter.
Three-nanometer technology contributed another 33%, while five-nanometer represented 28%.
Together, those three leading-edge process generations accounted for 69% of wafer revenue.
That's significant.
It demonstrates how quickly TSMC's revenue mix is moving toward its most advanced manufacturing capabilities.
And those capabilities are precisely where its competitive position is strongest.
Developing a new manufacturing node involves enormous research expenses, sophisticated equipment, years of engineering work, and difficult production ramp-ups.
The initial economics can be challenging because yields, manufacturing efficiency, and utilization improve over time.
But once a process matures, a manufacturer with strong demand and technical leadership can potentially earn extraordinary returns.
TSMC is demonstrating the attractive side of that equation.
Its customers want advanced manufacturing capacity, and many have limited alternatives for their most demanding products.
That demand supports premium pricing and substantial profitability.
But I don't want to confuse technological leadership with immunity from economic cycles.
Two-nanometer manufacturing will eventually mature. New technologies will emerge. Existing facilities will age, and additional capital will be required.
TSMC's competitive advantage is enormous.
It just happens to be an advantage that requires continual reinvestment.
And continual reinvestment is the part of this story I think deserves more attention.
A 51% Net Margin Is Wonderful. Spending $64 Billion to Maintain Growth Is a Different Conversation.
If TSMC were a software company, I might evaluate its remarkable profitability very differently.
Software businesses can sometimes expand revenue without increasing physical infrastructure spending proportionally.
Once a product has been developed, distributing it to additional customers may require relatively little incremental capital.
Semiconductor manufacturing doesn't work that way.
TSMC needs physical factories, specialized machinery, equipment installation, production qualification, and expensive process upgrades.
A fabrication plant isn't a temporary expense.
It's a massive financial commitment extending across many years.
And the company's manufacturing equipment doesn't magically retain its economic usefulness forever.
Technologies evolve. Customers migrate toward newer processes. Equipment depreciates, facilities require maintenance, and capital must be allocated to future generations.
This is why net income and free cash flow can tell very different stories for capital-intensive businesses.
Net income reflects revenue less accounting expenses, including depreciation.
Capital expenditures, however, represent actual cash spent acquiring or improving long-lived assets.
Those expenditures generally aren't deducted from net income immediately. Instead, their costs are recognized through depreciation over time.
So a company can report extraordinary profits while simultaneously committing enormous amounts of cash to new factories.
That isn't necessarily a problem.
In fact, it can be an excellent use of capital when future returns exceed the cost of the investment.
But investors have to examine the relationship between the money being invested today and the cash those investments are expected to produce tomorrow.
And when the annual capital budget approaches $64 billion, even relatively small forecasting mistakes can become extraordinarily expensive.
The Most Dangerous Word in Semiconductor Investing Is "Forever"
I've watched enough technology cycles to become suspicious whenever investors begin treating extraordinary demand as though it will continue indefinitely.
That doesn't mean I believe artificial intelligence is a temporary fad.
I don't.
I think AI represents a major technological development with substantial long-term implications for computing, enterprise software, automation, scientific research, and digital infrastructure.
But believing in the technology isn't the same thing as believing every dollar currently being invested in AI infrastructure will earn an attractive return.
History offers plenty of examples of transformative technologies accompanied by investment cycles that temporarily became excessive.
Railroads transformed transportation.
Telecommunications transformed communication.
The internet transformed practically everything.
None of those developments prevented investors from occasionally paying too much for infrastructure built around perfectly legitimate long-term opportunities.
That's the distinction I want to keep in mind with TSMC.
The future of artificial intelligence can be enormously successful while individual investments in AI infrastructure still produce disappointing returns.
Cloud providers and other large technology customers are spending aggressively because they anticipate substantial future demand for AI computing.
But eventually those expenditures need to be supported by actual economic activity.
AI applications need to generate revenue, reduce costs, increase productivity, or create enough value to justify the infrastructure supporting them.
If that happens, TSMC could enjoy years of exceptional demand.
If it doesn't happen as quickly as expected, customers may delay purchases, extend equipment usage, or moderate future spending.
And that could affect manufacturing utilization.
Which brings us to the number I consider one of the most important variables in this investment.
Utilization: The Number That Could Matter More Than the Next Earnings Beat
Semiconductor fabrication plants have substantial fixed costs.
The factories must be built. Equipment must be purchased. Engineers must be employed. Manufacturing systems need maintenance.
Many of those costs continue whether production is operating near capacity or significantly below it.
That creates operating leverage.
When demand is strong and factories operate at high utilization, fixed costs can be spread across more production.
Profitability improves.
But the reverse also works.
If demand weakens while manufacturing capacity remains available, revenue can decline faster than costs.
Margins compress.
And a business that looked spectacularly profitable during a period of scarcity can suddenly appear far less attractive.
Consider an illustrative manufacturing project.
Suppose TSMC invests $15 billion in a new manufacturing-capacity program.
Assume that capacity could support $12 billion in annual revenue at full economic utilization.
Also assume a 50% incremental cash contribution margin before a fixed $1 billion annual allowance for continuing capital and maintenance requirements.
These assumptions are hypothetical. They are not TSMC's disclosed economics for any specific factory or manufacturing process.
But they help illustrate the relationship between demand, utilization, and investment returns.
Three Possible Demand Scenarios
| Scenario | Strong AI demand | Moderate demand | Weaker demand |
|---|---|---|---|
| Economic utilization | 95% | 75% | 50% |
| Annual revenue | $11.4B | $9.0B | $6.0B |
| Cash contribution at 50% | $5.7B | $4.5B | $3.0B |
| Annual maintenance/reinvestment allowance | $1.0B | $1.0B | $1.0B |
| Annual cash available for capital recovery | $4.7B | $3.5B | $2.0B |
The differences are substantial.
Under the strong-demand scenario, the project generates approximately $4.7 billion annually after the assumed continuing reinvestment allowance.
At 75% utilization, that figure falls to $3.5 billion.
At 50% utilization, it drops to $2 billion.
These are simplified steady-state estimates, not forecasts of TSMC's companywide free cash flow. They exclude construction timing, production ramp losses, working-capital changes, taxes, financing effects, and potential residual value.
But they demonstrate why utilization matters so much.
Now suppose the project must recover its initial $15 billion investment over 12 years while satisfying a 10% required annual return, with no residual value.
That requires roughly $2.20 billion in annual cash available for capital recovery.
Under these simplified assumptions, the break-even utilization level for meeting that return requirement is approximately 53%.
That makes the 95% and 75% scenarios attractive.
The 50% scenario falls short.
And that's before we consider the possibility of construction delays, weaker pricing, higher operating costs, or faster-than-expected technological obsolescence.
Suddenly, utilization isn't merely a manufacturing statistic.
It's a major determinant of whether billions in capital expenditures create or destroy shareholder value.
The Bull Case: AI Demand Remains Strong Enough to Absorb Everything TSMC Builds
The strongest investment argument for TSMC is straightforward.
AI computing demand continues expanding. Advanced chip designs require sophisticated manufacturing processes, and TSMC remains exceptionally well positioned to supply those capabilities.
New manufacturing capacity enters production into an environment of sustained demand.
Utilization remains high.
Pricing stays firm.
Manufacturing yields improve.
And TSMC earns attractive returns on its enormous capital investments.
Under that scenario, the company's current spending program could reinforce its competitive advantage rather than dilute profitability.
It could make TSMC even more essential to its customers.
There's also an important strategic consideration.
Capacity constraints can push customers to explore alternative manufacturing relationships.
By expanding aggressively, TSMC may prevent competitors from gaining opportunities simply because it cannot satisfy demand.
Capital expenditures therefore serve two purposes.
They provide additional production capability, and they defend TSMC's competitive position.
I think that second purpose is easy to underestimate.
A company possessing such a powerful manufacturing advantage has good reason to protect it.
And management's October earnings commentary indicates confidence in continued multiyear demand.
The bull case isn't difficult to understand.
TSMC is building infrastructure for a computing transition that could continue for many years.
If demand remains sufficiently strong, today's enormous capital spending may look remarkably sensible in hindsight.
The Base Case: Strong Growth, but Gradually More Normal Economics
My moderate scenario is more nuanced.
I expect AI demand can remain substantial without continuing at its current extraordinary growth rate indefinitely.
As new manufacturing capacity becomes available, scarcity premiums may decline.
Customers may gain somewhat greater flexibility in production planning.
Overseas factories could introduce higher operating costs.
New manufacturing technologies may require expensive ramp-ups before achieving mature efficiency.
And the overall mix of products and customers may change.
None of these developments would necessarily destroy TSMC's competitive advantage.
They could simply move the company from extraordinary profitability toward profitability that is still exceptional by semiconductor industry standards.
Consider gross margin.
TSMC reported 66.4% in the third quarter, down 130 basis points sequentially from 67.7%.
That's hardly a financial disaster.
Most industrial businesses would be delighted with margins anywhere near those levels.
But for investors assigning premium valuations based on exceptional profitability, the direction of margins matters.
TSMC's fourth-quarter guidance calls for gross margin between 65% and 67%, and operating margin between 56.5% and 58.5%.
That suggests management anticipates very strong profitability, although not uninterrupted margin expansion.
I wouldn't interpret one quarter of margin movement as proof of structural deterioration.
But I would monitor whether overseas manufacturing costs, new-node ramp expenses, exchange rates, and other factors begin producing more persistent pressure.
A company can remain an outstanding business while becoming a less attractive investment at an overly optimistic valuation.
That distinction matters.
The Bear Case: The AI Boom Continues, but Spending Gets Ahead of Demand
The bearish scenario doesn't require artificial intelligence to fail.
It simply requires expectations to outrun reality.
Imagine the major AI infrastructure buyers expanding aggressively for several years.
Chip manufacturers respond by adding capacity.
Equipment suppliers expand production.
Advanced packaging facilities increase output.
Then AI infrastructure customers discover that near-term revenue growth, project returns, power availability, or financial constraints don't support another round of equally aggressive purchases.
Demand doesn't disappear.
It slows.
Orders get delayed.
Capacity utilization declines.
And some of the factories planned during the period of extraordinary scarcity reach production just as the market becomes less constrained.
That's the risk.
TSMC isn't spending $64 billion to support last quarter's revenue.
It's investing because management believes future demand will justify the capacity being built.
Those are forward-looking commitments.
And forecasting demand several years into the future is difficult even for excellent management teams.
I also worry about the interaction between capital intensity and technological progress.
A factory designed around a particular manufacturing generation must earn sufficient returns before customer demand migrates to newer technologies.
If product transitions accelerate, capital recovery becomes more challenging.
If demand slows, manufacturers may face pressure to use price concessions to maintain utilization.
And if new capacity enters production during weaker conditions, depreciation and other fixed costs can weigh on reported margins.
None of this is inevitable.
But it would be a mistake to assume that today's extraordinary profitability automatically protects TSMC from future semiconductor cycles.
Geopolitical Risk Isn't Something I Can Ignore
No serious investment analysis of TSMC is complete without discussing Taiwan.
TSMC's manufacturing capabilities are strategically important to the global economy.
Its facilities are central to the supply chains supporting advanced computing, smartphones, networking equipment, and other technologies.
That importance creates substantial geopolitical exposure.
Tensions involving Taiwan and China remain a serious concern for investors.
Disruptions to transportation, trade, manufacturing operations, or international security could affect TSMC and its customers.
Even without an outright military conflict, export restrictions and changing international trade policies can influence which customers can purchase advanced technologies.
TSMC's overseas expansion provides some geographic diversification.
But building manufacturing capacity outside Taiwan introduces costs, complexity, and execution challenges.
The company has to balance resiliency and customer requirements against the economics that made its existing manufacturing network so effective.
And that balance isn't free.
Manufacturing resilience costs money.
Geographic diversification costs money.
Maintaining technological leadership costs money.
All of these expenditures may be strategically justified.
But shareholders still need to consider their financial consequences.
A necessary investment doesn't automatically become an investment with an attractive return.
What Would Make Me More Bullish?
The next several quarters will provide important evidence about whether TSMC can sustain its extraordinary economics while expanding capacity.
I would become more confident if new manufacturing technologies ramp efficiently, overall utilization remains high, and pricing stays firm despite capacity additions.
I'd also want to see evidence that free cash flow grows meaningfully after accounting for the capital spending necessary to support future production.
The distinction between growth spending and maintenance spending matters here.
A portion of TSMC's investments supports technological transitions, upgrades, and the continued operation of its manufacturing network.
Another portion expands production capacity.
But even capacity expansion that generates future growth may be necessary to defend the company's market position.
That makes it difficult to treat the entire capital budget as purely discretionary.
I'd examine the returns on invested capital across multiple years rather than relying on a single exceptional quarter.
If TSMC can sustain strong returns while its invested capital base increases substantially, that would reinforce the argument that the company possesses a durable economic advantage.
I'd also monitor the evolution of advanced-node revenue, management's capacity commentary, and the financial health of the large customers driving AI infrastructure spending.
Those indicators would help distinguish a healthy multiyear growth cycle from a period when investment commitments are beginning to exceed realistic demand.
What Would Make Me More Cautious?
Several developments would make me reconsider the attractiveness of TSMC's long-term financial outlook.
Persistent declines in gross margin despite increasing production volumes would concern me.
So would repeated capital-spending increases unsupported by improving future cash-flow expectations.
I'd become especially cautious if major AI customers started slowing infrastructure purchases at the same time TSMC was bringing substantial new production capacity online.
That's precisely when operating leverage can reverse.
I'd also pay attention to whether manufacturing efficiency and yield improvements offset the higher costs associated with new technologies and overseas facilities.
A new factory operating well below expected capacity may be strategically useful, but it can still become a disappointing investment.
And I'd watch the relationship between accounting earnings and cash generation.
When earnings grow dramatically but free cash flow fails to keep pace over several years, I want to understand why.
Sometimes the explanation is excellent investment opportunities.
Sometimes it's a warning that the business requires more capital than investors originally expected.
The difference can determine whether shareholders enjoy attractive long-term returns.
Valuation: A Great Company Can Still Be a Bad Investment at the Wrong Price
Here's where I separate my admiration for TSMC from my willingness to buy its stock at any valuation.
I consider TSMC one of the most strategically important companies in the global technology industry.
Its manufacturing capabilities are extraordinarily difficult to replicate.
Its technological leadership is impressive.
And its ability to generate exceptional margins while operating such a capital-intensive business is remarkable.
But none of that means the stock is automatically attractive regardless of price.
I don't want to value TSMC solely by annualizing its strongest recent quarter.
Doing so assumes today's profitability offers a reliable representation of long-term earnings power.
Perhaps it does.
Perhaps it doesn't.
I'd rather consider a range of outcomes involving revenue growth, normalized gross margins, capital requirements, and free cash flow.
In a bullish scenario, sustained AI growth and high utilization could justify substantial earnings expansion.
In a moderate scenario, revenue might continue growing while margins normalize and capital spending remains elevated.
In a bearish scenario, weaker utilization and persistent capital requirements could cause earnings and cash flow to fall short of optimistic expectations.
The appropriate valuation depends on the probability assigned to those outcomes.
And that's a much more complicated exercise than multiplying next year's projected earnings by an arbitrary price-to-earnings ratio.
A premium company deserves a premium valuation only when its future cash flows justify it.
The market doesn't owe shareholders a premium multiple simply because the underlying business is impressive.
My Investment Thesis: The Quality Is Obvious. The Capital Returns Are the Test.
After reviewing TSMC's third-quarter results, I remain impressed by the quality of the business.
It's difficult to argue with revenue growing more than 50% in local currency terms, net income increasing almost 69%, and net margins exceeding 51%.
Those numbers demonstrate remarkable operating strength.
And I believe TSMC is positioned to remain critically important to the AI ecosystem.
But the capital-expenditure guidance changes how I evaluate the opportunity.
A $60 billion to $64 billion annual spending program is a major financial undertaking, even for a company of TSMC's size.
It reflects management's confidence in future demand.
It also creates substantial expectations that future demand must satisfy.
My central investment question isn't whether TSMC can manufacture the world's most advanced semiconductors. It has already demonstrated that capability.
My question is whether the next generation of manufacturing capacity will generate attractive returns after accounting for its full cost.
That's what determines whether today's enormous investments create lasting shareholder value.
If TSMC can preserve high utilization, defend pricing, and maintain strong through-cycle returns on invested capital, its current expansion may prove to be one of the most valuable industrial investments of this era.
If demand moderates sharply or new capacity becomes underutilized, today's exceptional margins may offer less protection than investors expect.
And that's why I refuse to confuse a wonderful earnings report with the conclusion that every future investment will be equally wonderful.
Final Thoughts: TSMC Is Winning the AI Gold Rush. Now It Has to Pay for the Mine.
The most interesting thing about TSMC's third-quarter earnings isn't simply how much money the company made.
It's the contrast between how profitable the business has become and how much capital it must continue investing.
A 51.1% net margin looks like something from a software company's financial statements.
A $64 billion capital budget looks like industrial manufacturing operating on an almost unimaginable scale.
And somehow, TSMC is managing to combine both.
That's extraordinary.
But it also creates a fascinating investment challenge.
During a period of scarcity, manufacturing capacity becomes exceptionally valuable.
Customers compete for access. Prices remain firm. Utilization stays high. Profitability can become remarkable.
Eventually, however, companies respond to scarcity by investing in more capacity.
And the economics of an industry can change when supply catches up with demand.
TSMC's technological position may allow it to manage that transition better than most manufacturers.
Its competitive advantages are real, substantial, and difficult to duplicate.
But they're not a substitute for disciplined capital allocation.
I don't believe AI is going away.
I don't believe demand for advanced semiconductors is about to disappear.
And I certainly don't believe TSMC is suddenly becoming an ordinary manufacturing business.
What I believe is that the next phase of the investment story will be decided by the returns TSMC earns on the billions it is spending today.
The earnings report tells me that existing manufacturing capacity is extraordinarily profitable.
The capital budget tells me management expects enormous future opportunities.
But expectations aren't cash flows.
Factories aren't free.
And even the world's best semiconductor manufacturer has to earn an adequate return on its investments.
That's the test I'll be watching.
Because TSMC has already proven that it can make incredible amounts of money during an AI-driven chip shortage.
Now it needs to prove that spending incredible amounts of money to eliminate that shortage won't eventually undermine the economics that made the investment so attractive in the first place.
And for long-term shareholders, that may be the most important question of all.
Sources
Primary company disclosures:
Investment disclosure: This article represents my interpretation of publicly available information and is not personalized investment advice. The three manufacturing scenarios are illustrative assumptions, not management forecasts or estimates of specific TSMC factories. Investors should review updated financial disclosures and their own risk tolerance before making investment decisions.
Comments
Post a Comment