A chip designer can announce a faster AI processor, but it cannot ship that processor without a factory capable of making it reliably. Taiwan Semiconductor Manufacturing Company (TSM), usually called TSMC, provides that manufacturing capacity for many of the world’s leading chip designers. For investors, the central question is not simply whether AI spending will grow. It is whether TSMC can build the right capacity, fill it at attractive prices, and earn enough on those enormous investments to justify its valuation.
What TSMC Does and Who Pays It
TSMC is a semiconductor foundry: it manufactures chips that other companies design. Customers supply designs, work with TSMC to make them suitable for a particular manufacturing process, and reserve production capacity. TSMC then processes silicon wafers through hundreds of manufacturing steps before chips are tested and prepared for assembly.
Its customers include businesses with very different end markets. Apple (AAPL) designs processors for consumer devices. Nvidia (NVDA) designs GPUs and other chips used in AI infrastructure. Advanced Micro Devices (AMD) designs processors and accelerators for computers and data centers. Their products compete in different ways, but all require reliable manufacturing.
TSMC’s dedicated foundry model matters because it generally does not compete with customers by selling its own branded processors. Customers can share sensitive designs with a manufacturer whose core business is producing chips rather than taking their end customers.
The basic revenue mechanism is manufacturing volume multiplied by the price of the manufacturing services provided. Process technology, contract terms, product complexity, and related services affect that price. Advanced packaging and testing contribute additional revenue.
| Revenue driver | How it works | What investors should examine |
|---|---|---|
| Wafer production | Customers pay for manufacturing under agreed commercial terms | Shipment volume, pricing, and utilization |
| Advanced process mix | More demanding manufacturing processes generally command higher wafer prices | Revenue mix and the cost of ramping production |
| Specialty processes | Technologies serve applications such as connectivity, power management, and automotive electronics | Demand durability and competitive pricing |
| Packaging and testing | Services connect, assemble, and verify chips after wafer fabrication | Available capacity and customer adoption |
Higher wafer prices do not automatically mean higher profits. A new process can require more expensive equipment, additional production steps, and significant engineering work before output becomes efficient.
A useful historical reference is TSMC’s reported 2024 performance: revenue was about $90.1 billion, gross margin was 56.1%, and operating margin was 45.7%. These are dated figures, not a statement of current performance. They illustrate that a manufacturing business can generate substantial margins when its technology, pricing, and factory loading work together.
Why Manufacturing Yield Matters
Yield is the proportion of usable output from production. A wafer contains many individual chips, and defects can make some unusable. Better yields improve customer economics and help TSMC offer dependable production at scale.
Commercial arrangements vary, so investors should not assume every contract pays TSMC only for working chips. Yield still matters because customers ultimately care about the cost and availability of usable processors, not just the number of wafers entering a factory.
Why Capacity Is the Bottleneck
A semiconductor factory is not interchangeable floor space. Useful capacity requires the right equipment, process technology, trained workers, supporting utilities, and customer qualification. A building can be complete while commercially valuable production remains months away.
TSMC must therefore make spending decisions well before demand becomes certain. Factories take years to develop, while customers can revise chip orders much more quickly. This mismatch between slow supply expansion and changing demand is central to foundry economics.
There are several distinct capacity constraints:
- Wafer fabrication: The factory must have enough equipment for the required process, including lithography, deposition, and etching.
- Manufacturing yield: Installed equipment does not guarantee enough usable chips if a new process is still being refined.
- Advanced packaging: Processors and memory must be connected in configurations suitable for demanding workloads.
- Supporting components: Memory, substrates, and other inputs can limit shipments even when TSMC has available wafer capacity.
For AI accelerators, packaging is particularly important. TSMC’s CoWoS technology connects logic chips and high-bandwidth memory through advanced packaging structures. Making the processor wafer is only part of delivering a functioning accelerator package.
Expanding CoWoS capacity can relieve one constraint, but it does not guarantee that every other input will arrive on schedule. High-bandwidth memory availability, customer design readiness, and downstream assembly can still restrict completed systems.
This is why announced capacity growth needs interpretation. Ask what is expanding, which products it supports, and when customers can actually use it. More capacity for an older process does not directly solve a shortage on the newest process. More packaging equipment does not automatically create additional advanced logic wafers.
Utilization is the financial link between capacity and earnings. Factories carry large fixed costs, including depreciation, maintenance, and staffing. When more production runs through an existing facility, those costs are spread across more output. When orders weaken, expenses do not fall as quickly as revenue.
Scarce capacity can strengthen pricing, but TSMC cannot simply maximize short-term prices without considering customer relationships. Its customers make long product-development commitments. Predictable supply and coordinated technology roadmaps can be more valuable than extracting the highest possible price in one quarter.
The best outcome is not permanent shortage. It is disciplined expansion that keeps customers supplied while preserving attractive returns on the capital committed. Excessive scarcity can encourage alternative suppliers; excessive expansion can leave expensive equipment underused.
TSMC’s Competitive Position and the Bull Case
TSMC’s advantage combines process technology, manufacturing scale, customer relationships, and an extensive design ecosystem. Customers use process-specific design rules, software tools, and reusable intellectual property when developing chips. Moving a complex design to another manufacturer can require substantial engineering work and fresh qualification.
That creates switching friction, not a permanent guarantee of customer loyalty. A rival with better economics or dependable supply could still win future designs. The strongest evidence of an advantage is repeated customer adoption across successive manufacturing generations.
The main alternatives at the most advanced end of manufacturing include Samsung Electronics (005930.KS) and Intel (INTC). Their broader business structures differ from TSMC’s dedicated foundry model. Samsung also sells memory and consumer products, while Intel combines processor products with efforts to serve external foundry customers.
Foundry competition cannot be judged from process names alone. Labels such as 3 nanometers or 2 nanometers identify manufacturing generations, not a single directly comparable physical measurement. Customers assess performance, power consumption, chip density, yield, design support, and delivery reliability together.
The bull case rests on several mechanisms. First, demanding computing workloads can increase the value of advanced manufacturing. A processor that delivers more useful work within a power budget may justify higher manufacturing costs.
Second, TSMC can serve competing chip designers. If AI spending shifts between customers, it may retain manufacturing demand rather than depending on one branded product winning every contest. This does not eliminate customer concentration, but it broadens the routes through which demand can reach its factories.
Third, advanced packaging expands TSMC’s role beyond producing individual chips. As customers combine multiple chip components and memory, coordination between manufacturing and packaging becomes more important.
Finally, scale can reinforce execution. Large production volumes provide manufacturing experience and help support research spending across successive process generations. That relationship can sustain an advantage when investment produces dependable customer output.
None of these arguments settles valuation. A capable business can still deliver disappointing shareholder results if the share price already assumes exceptional growth. Our Morningstar Investor review explains how independent valuation research can support, rather than replace, an investor’s own assumptions.
The Bear Case: Heavy Spending and Concentrated Risk
TSMC’s factories are both its advantage and its largest financial obligation. Capital expenditure consumes cash before the associated capacity produces revenue. Depreciation then affects reported earnings over subsequent years as equipment is used.
In 2024, TSMC reported capital expenditure of approximately $29.8 billion. That historical figure shows the scale involved. It should not be treated as a forecast for later years or as evidence that future spending will earn the same return.
Investors should distinguish operating cash flow from free cash flow. Operating cash flow reflects cash generated through business operations. Free cash flow commonly subtracts capital expenditure. A period of rapid expansion can therefore show strong earnings growth without equally strong growth in cash remaining after investment.
The interpretation requires care. Lower free cash flow can reflect productive expansion rather than a weakening business. Conversely, calling all factory spending an investment in growth does not establish that it will earn an adequate return. Utilization, pricing, and customer demand must eventually validate the spending.
Overseas expansion introduces another trade-off. Factories outside Taiwan can improve geographic resilience and satisfy customer or government requirements. But construction costs, staffing, supply-chain depth, and operating conditions differ across locations. Subsidies can offset some costs without making every site equally profitable.
Taiwan remains central to TSMC’s manufacturing network. Military conflict, a blockade, or other severe disruption could affect production, transport, customer access, and financial markets simultaneously. Overseas facilities reduce some exposure but do not quickly reproduce the full manufacturing ecosystem.
Customer concentration is another risk. Large customers can account for substantial revenue and influence technology roadmaps. A major customer’s weaker product cycle, inventory correction, or internal manufacturing decision can affect factory loading.
AI demand also needs to be separated from the broader semiconductor cycle. Strong accelerator orders can coexist with weak demand for smartphones, industrial chips, or other products. Aggregate revenue growth may hide uneven utilization across processes.
Export controls and trade restrictions can narrow the set of products TSMC is allowed to manufacture for particular customers. These rules can change independently of consumer demand or manufacturing performance.
Finally, owning several semiconductor businesses does not necessarily spread these risks. A chip designer, equipment supplier, and TSMC may all depend on the same AI capital-spending cycle. The principles in our portfolio diversification guide help distinguish a long holdings list from genuinely different economic exposures.
What to Watch Next in a TSMC Stock Analysis
A useful TSMC stock analysis connects operating developments to cash generation and valuation. Monthly revenue releases can reveal changes in sales momentum, but quarterly reports and management commentary provide more context about margins, manufacturing mix, and spending.
Use a consistent checklist rather than treating each production announcement as an isolated event:
- Advanced-process adoption: Are customers moving important products onto newer processes, and is that adoption generating meaningful revenue?
- Capacity and utilization: Is new equipment producing qualified output, and are existing factories well loaded?
- Packaging availability: Is additional capacity easing shipment constraints, or has the limiting factor moved elsewhere?
- Margin progression: Are better pricing and product mix offsetting new-process ramp costs, overseas expenses, and depreciation?
- Capital expenditure and cash flow: Is investment translating into customer demand and cash generation over time?
- Customer and end-market mix: Is growth broadening, or becoming more dependent on a narrow set of AI programs?
Read these indicators together. Rising revenue alongside falling margins might reflect the temporary cost of introducing a new process. It could also indicate weaker pricing or an unfavorable mix. Management’s explanation should be checked against subsequent results.
For valuation, compare more than a single forward price-to-earnings ratio. Build a base case for revenue, operating margin, taxes, and capital expenditure. Then test what happens if utilization weakens, factory ramps take longer, or packaging growth slows. The key question is how much future success the current valuation already requires.
Keep security units and currencies consistent. TSMC’s US-listed American depositary shares represent five Taiwan-listed common shares each. Comparing prices, earnings per share, or dividends across the two listings requires the correct share ratio and exchange rate. Otherwise, an apparent valuation difference may simply be a calculation error.
Screening can help organize comparisons, but it cannot capture every manufacturing constraint. Our guide to using stock screeners for technology-company research explains why comparable financial metrics are a starting point rather than a finished investment case.
TSMC makes money by converting difficult manufacturing processes and expensive capacity into dependable chip supply. Its strongest attributes are technical execution, scale, and customer trust. Its central risks are the cost of maintaining those strengths, the cyclicality of demand, and geographic concentration. The practical test is whether each new wave of spending creates productive capacity at attractive economics—not simply whether demand for more computing continues.