The sentence that sounded like a mistake
In 1987, Morris Chang founded Taiwan Semiconductor Manufacturing Company on a premise that, to most established chipmakers of the era, sounded less like a strategy and more like a self-inflicted handicap: TSMC would manufacture chips designed by other companies, and would design none of its own, by policy, permanently [1]. In an industry where a company’s own proprietary chip designs were widely treated as its core competitive asset, choosing never to have any looked, to most contemporaries, like giving up the most valuable part of the business on purpose.
Why that looked strange to the incumbents of the era
The dominant model of the time was the integrated device manufacturer, or IDM: a company like Intel or NEC designed its own chips and manufactured them in its own fabs, treating manufacturing capacity as a captive resource that primarily served its own product roadmap. IDMs did sometimes sell spare fabrication capacity to outside customers, but only opportunistically, as a secondary revenue stream when their own production demand left room to spare — never as the entire basis of the business [1]. A company organizing itself around manufacturing exclusively for others, with no proprietary designs of its own to fall back on, had no established precedent to point to as proof the model could work at scale.
The insight the incumbents were missing
TSMC’s bet depended on a structural reclassification that turned out to be the whole idea: instead of treating other chip-design companies as competitors to be out-designed, a pure-play foundry treats them as customers — the more design companies exist and need manufacturing, the larger TSMC’s addressable market becomes, with zero design competition against any of them [2]. This only works if enough companies exist that want to design chips without building or owning fabs themselves — a category that, in 1987, was still small, expensive to enter, and largely unproven as a standalone business model.
The precondition it created
TSMC’s pure-play bet did not just add one company to the industry; it created the structural precondition for an entire category of companies that did not meaningfully exist before it: the fabless design industry. Nvidia, AMD, Qualcomm, Apple’s silicon design teams, Broadcom, and effectively every company covered throughout this cohort’s accelerator and startup tracks that designs chips without operating its own fabs depends, directly or indirectly, on a pure-play foundry existing to manufacture what they design [4]. Without TSMC’s 1987 bet proving the model could work, the capital and expertise required to both design leading-edge chips and build leading-edge fabs would have remained a combined barrier to entry blocking most of the companies this cohort spends its pages covering.
Why the idea eventually stopped looking strange
The pure-play foundry model did not win the argument quickly. It took roughly a decade of compounding evidence — fabless design companies scaling successfully on top of TSMC’s capacity, process technology improving fast enough to keep foundry customers competitive with IDMs’ own in-house manufacturing — before the model shifted from a curiosity to the industry’s dominant growth pattern. By the 2000s, the fabless-plus-foundry split had become the default structure for new entrants, and by the 2020s, this cohort’s own foundry-track briefings document TSMC as the company essentially the entire advanced-node industry runs through, a position unimaginable to most observers of the 1987 founding.
The part of the story that resembles Fairchild’s
This cohort’s companion briefing on the Fairchild Eight documents a similar pattern a generation earlier: an idea that looked like an unnecessary risk at the moment of founding, validated only in retrospect by the scale of what it eventually enabled. TSMC’s 1987 bet and Fairchild’s 1957 exit are separated by three decades and solve entirely different problems, but both share the same underlying lesson this cohort returns to repeatedly across its history track — the industry’s most consequential structural innovations rarely look obviously correct to contemporaries at the moment they are made, and are validated only by everything that gets built on top of them afterward.
Why this history matters for reading TSMC’s present-day dominance correctly
This cohort’s present-day TSMC briefings document a company operating at a scale and technological lead that can read, in isolation, as an inevitable outcome. It was not inevitable. It was a specific, contestable bet made in 1987 against the era’s dominant business logic, sustained through a decade where the model’s success was genuinely uncertain, before compounding into the position TSMC holds today — a distinction worth keeping in mind before treating any single company’s current dominance, in this industry or any other, as though it were always the obvious outcome all along.
