The fabless-foundry model is the structure that organizes much of the modern semiconductor industry: it separates the design of an integrated circuit from its manufacture. A “fabless” company designs chips but owns no fabrication plant; a “foundry” owns the fabs and manufactures chips to others’ designs but sells no products of its own. This division of labor lets a small team with a good design bring a chip to market without the multibillion-dollar cost of building and running a leading-edge fab.
The model was pioneered by the Taiwan Semiconductor Manufacturing Company, founded in 1987 by Morris Chang as the world’s first dedicated, pure-play foundry. TSMC’s own description of the business explains that the dedicated-foundry idea “greatly lowered the threshold needed to establish an IC design company” by absorbing the rising complexity of process technology and the enormous capital required to build wafer fabs. Because TSMC committed to never competing with its customers’ products, fabless designers could trust it with their designs.
This separation built directly on the design discipline established by the Mead-Conway VLSI revolution, which had introduced scalable, geometry-based design rules and the multi-project-chip idea — the notion that design could be cleanly decoupled from fabrication. The foundry model turned that decoupling into an industry: design houses send a standardized layout to a foundry, which fabricates it on a shared process.
The consequences have been profound. Fabless giants such as Nvidia, Qualcomm, AMD, Apple’s silicon group, and countless startups exist because they can rent world-class manufacturing rather than own it, while foundries like TSMC concentrate the staggering capital and expertise needed to stay on the leading edge. The model has also become geopolitically central, since a handful of foundries now manufacture the most advanced chips the entire computing industry depends on.