The Semiconductor Stack — Fab, Fabless, Foundry, OSAT
How the chip industry actually divides labour: design, manufacturing, packaging, and the roles each company plays
Open the back of any modern smartphone, laptop, automobile dashboard, or data-center server, and you'll find one or more chips with a logo on top — Apple, Nvidia, AMD, Qualcomm, MediaTek. Almost none of those chips were physically manufactured by the company whose logo they bear. The Apple A18 Pro that powers the iPhone 16 was manufactured at one of TSMC's fabs in Taiwan on the N3E process; the Nvidia H100 that powers the AI training boom was also manufactured at TSMC on a customised 4nm-class process; the AMD Ryzen 9 in millions of laptops was manufactured at TSMC.
The companies on the front of those chips are designers — they specify the circuit, run the verification, license the customer relationship. The actual silicon-fabrication step happens at a separate company called a foundry. Approximately 90% of the world's leading-edge logic foundry capacity (≤7nm) sits in Taiwan; total foundry market share for Taiwan-based companies (TSMC plus UMC and others) is roughly 65-70% across all process nodes.
That is the opening. Finishing a lesson is where it stops being interesting and starts being useful: the full lesson runs to 7 sections and ends with 6 practice questions. A free account is what opens the rest, and the other 255 lessons in the Academy with it. No card.
What this lesson covers
- 1The fabless / foundry / OSAT split — and why it happened
- 2Where each public company sits in the stack
- 3Cost flow through the stack — what each layer captures
- 4Where the gross margin sits — chip stack layer by layer
- 5Q4 2024 — Nvidia at $3.5T market cap and what it actually built
- 6Where to see this on the platform
- 7Summary