TSM

The company was founded in 1987 by Morris Chang, a Texas Instruments veteran who had spent decades inside the integrated device manufacturing model that defined the early chip industry, in which a single company designed, fabricated, and sold its own semiconductors. Chang's insight was that manufacturing was becoming too capital-intensive and too complex for most design houses to sustain, and that a company doing nothing but contract fabrication, never competing with its own customers, could attract the entire industry's outsourced volume. TSMC opened with around 220 million dollars in capital, roughly half from the Taiwanese government's National Development Fund and a large minority stake held by Philips, which contributed process technology and intellectual property. The early years relied on trailing-edge work that established device manufacturers were happy to hand off, but the model's real prize was the fabless companies it made possible. Firms such as Nvidia, Broadcom, Marvell, and Qualcomm could design chips without ever building a factory, and they grew up alongside TSMC. That symbiosis, the foundry and the fabless ecosystem rising together, is the structural story of the modern semiconductor industry and the foundation of everything the company became.

What TSMC sells is manufacturing capacity at the frontier of physics. It operates fabrication plants, known as fabs, that turn silicon wafers into finished logic chips through hundreds of precise steps. Customers bring a design, and TSMC produces it on a chosen process technology, or node, named historically by feature size in nanometers. The business is organized around offering a ladder of nodes, from mature processes that still serve automotive, industrial, and analog needs to the most advanced nodes reserved for high-performance computing and flagship mobile chips. Pricing rises sharply with sophistication, and the newest nodes command the highest margins because TSMC is often the only supplier able to make them at volume with acceptable yield. Alongside wafer fabrication, the company has built a large and increasingly critical advanced packaging business. Technologies marketed under names such as CoWoS stitch multiple chips and high-bandwidth memory into a single integrated package, which is essential for modern AI accelerators. Packaging has become a bottleneck of its own, and TSMC has been expanding that capacity aggressively to keep pace with demand from the AI chip designers.

The economic engine rests on a self-reinforcing advantage that competitors have found nearly impossible to break. Building a leading-edge fab costs many billions of dollars and requires the most advanced lithography equipment in existence, and the only way to justify that spending is to fill the fab with high volume. TSMC has the volume because it has the customers, and it has the customers because it has the best yield and the broadest capacity, which it can afford because of the volume. Each new node consumes enormous research and capital investment, and spreading those costs across the combined demand of Apple, Nvidia, and dozens of other designers gives TSMC a unit-cost position no smaller rival can match. The model also avoids a structural conflict that burdens its competitors. Because TSMC designs no products of its own, customers can share their most sensitive roadmaps without fear of being copied or deprioritized, a trust that a foundry owned by a chip-designing parent cannot fully replicate. The result is a moat built from capital intensity, manufacturing learning curves, an ecosystem of design tools and intellectual property tuned to its processes, and customer trust accumulated over decades.

In process leadership the company sits at the front of the industry. Through the first half of the 2020s it moved from a leading 3nm generation into its 2nm family, branded N2, which entered volume production in the second half of 2025 with Apple, AMD, and others as early adopters and demand quickly exceeding initial capacity. An enhanced version, N2P, and a related node called A16 that adds a backside power delivery scheme were positioned to follow, with the company describing the 2nm class as a long-lived major node and pointing toward a future A14 generation. Maintaining this cadence is what separates TSMC from the field. Two competitors operate at or near the leading edge. Samsung runs its own foundry but holds a far smaller share, in the high single digits, and has struggled to match TSMC's yields on the newest nodes. Intel, historically an integrated manufacturer, has been attempting to build a contract foundry business and reclaim process leadership, but it remains behind on volume and external customer adoption. Across the broader market TSMC's dominance at the leading edge is close to total, and its mature-node business competes with a wider set of foundries on price rather than capability.

The customer base concentrates the world's most valuable technology companies. For years Apple was the single largest customer, anchoring each new node with its iPhone and Mac processors. By 2025 the surge in artificial intelligence reshaped that ranking, with Nvidia reportedly overtaking Apple to become the largest customer as data center accelerator demand exploded, each accounting for roughly a fifth and a sixth of revenue respectively, followed by MediaTek, Qualcomm, AMD, Broadcom, and Intel. High-performance computing, which includes AI chips, grew to more than half of quarterly revenue, overtaking smartphones as the company's largest end market. This concentration is a strength and an exposure at once. The same handful of customers that fill the leading-edge fabs also means a slowdown in AI capital spending or a stumble at a major account would be felt sharply, even as the breadth of designers relying on TSMC limits the damage any single defection could cause.

Leadership passed through a deliberate succession. Morris Chang retired in 2018 after building the company and briefly returning to steer it through a difficult period. C.C. Wei, a long-serving engineer and executive, became Chief Executive Officer in 2018 and added the role of Chairman in June 2024, consolidating board-level and operational authority in a single figure after years of a split arrangement. The executive bench is deep with long-tenured engineers, including co-chief operating officers overseeing the sprawling manufacturing organization. The culture is widely described as disciplined, engineering-driven, and intensely focused on yield, reliability, and the relentless schedule of node development. That operational rigor, more than any single product, is what customers buy when they commit a generation of designs to TSMC's factories.

Strategy in the current era centers on two simultaneous bets. The first is sustaining the process and packaging lead through continued heavy capital expenditure, keeping the company a generation ahead so that the most demanding chips have nowhere else to go. The second is geographic diversification in response to customer and government pressure to reduce the concentration of advanced manufacturing in Taiwan. The company has committed a very large sum, reported in the range of 165 billion dollars, to build a cluster of fabs in Arizona, with the first already producing and later phases targeting 2nm-class processes. It also operates a fab in Kumamoto, Japan, through a joint venture, with a second Japanese fab planned, and has explored European capacity. These overseas projects carry higher costs and have faced schedule and demand challenges, and they are intended to supplement rather than replace the Taiwan base, which retains the cost and cadence advantages of the company's home cluster.

The risks are specific and significant. The largest is geopolitical. The concentration of the world's most advanced chip production in Taiwan, an island that China claims and that faces sustained military pressure, creates a tail risk with consequences far beyond the company itself. Analysts have estimated that a conflict over Taiwan could inflict trillions of dollars of damage on the global economy, and the same concentration that gives Taiwan a so-called silicon shield also makes TSMC a single point of failure for the modern technology supply chain. Beyond geopolitics, the business is cyclical and capital-hungry, requiring tens of billions in annual spending whose returns depend on demand that can soften. Customer concentration ties the company's fortunes to a few buyers and, increasingly, to the durability of AI infrastructure spending. Technological risk is constant, since a missed node transition or a yield failure at the leading edge would be expensive and hard to reverse. The overseas expansion adds execution and margin risk, and export controls and trade policy involving the United States and China can reshape which customers TSMC may serve and on what terms.

Taken together, Taiwan Semiconductor Manufacturing is the indispensable manufacturing layer beneath the digital economy, a company whose pure-play foundry model turned chip design and chip fabrication into separate industries and then captured the most valuable half of that split. Its position rests on a moat of capital scale, manufacturing expertise, and customer trust that no rival has closed despite years of effort and government support, and its leadership at the most advanced nodes gives it pricing power and strategic weight that few industrial firms possess. The same factors that make it dominant also make it exposed, concentrating extraordinary value and extraordinary risk in factories clustered on a contested island, with fortunes increasingly tied to a small group of customers riding the artificial intelligence cycle. For an investor, the company represents a rare combination of structural dominance and structural fragility, where the durability of the franchise must be weighed against geopolitical and cyclical forces largely outside its control. TSM offers exposure to that profile through a US-listed depositary receipt on the business that quietly manufactures the foundation of the computing era.