Hidden Economics

The Semiconductor Foundry and the Economics of Extreme Capital Intensity

Work through the most capital-intensive manufacturing business on earth and come away with the general rule: when fixed cost dominates and capacity cannot be stored, utilisation stops being an operating metric and becomes the entire business model.

  • Advanced
  • 13 min total
  • 15 chapters

What decision this helps you make: Whether to own capacity or buy it, how to price when almost all of your cost is already sunk, and how to tell a temporary utilisation dip from a structural one before you cut the wrong thing.

What this topic is

A semiconductor foundry manufactures chips designed by other companies. It does not sell chips; it sells wafer capacity, and the capacity lives inside a facility costing tens of billions of dollars that becomes obsolete in under a decade. Almost every cost is committed before the first wafer is made, so the price of any individual wafer bears almost no relationship to what it costs to produce one more — and every hour the machines are not running is revenue that can never be recovered.

Why it matters

This is the extreme end of a structure that shows up everywhere: hotels, airlines, gyms, clinics, hire fleets, print shops, data centres, agencies selling billable hours. In all of them, capacity is perishable and cost is committed, which means the intuitive rules about pricing, discounting and cost control are inverted. The foundry makes those inversions impossible to miss, because the numbers are so large that a mistake shows up in the same quarter.

Who should learn it

Owners of any business with expensive capacity that cannot be stored, anyone deciding between owning an asset and renting access to one, and anyone who wants to understand why the world depends on a handful of factories and why that is an economic outcome rather than an accident.

What you will understand

  • Why cost per unit is almost entirely a function of utilisation when fixed costs dominate
  • How yield, the share of usable output on a wafer, compounds with utilisation to produce a much steeper curve than either alone
  • Why the industry consolidated to a handful of leading-edge producers, and why subsidy has not reversed it
  • How to price and how to cut when most of the cost is already committed

Prerequisites

Common misconception

"Chips are expensive because the materials and the process are expensive." The physical inputs are almost irrelevant. A wafer of silicon is a disc of refined sand, and the electricity, chemicals and labour to process one are a small fraction of the price. What is expensive is the building, the machines inside it and the years of engineering that made the process work at all — and every dollar of that was spent before the first order arrived. The cost of a chip is a share of a sunk cost divided by however many chips actually get made and actually work. Change the divisor and you change the cost of the product, without touching the product.