The Fabless + TSMC ModelNarrow moat

A decisive structural choice behind AMD's comeback was its fabless model: AMD designs its chips and outsources their manufacture to TSMC, the world's leading foundry, rather than running its own fabrication plants. This proved to be exactly the right bet at exactly the right time. Through the late 2010s and into the 2020s, Intel — which manufactures its own chips — stumbled badly on its process technology, falling years behind, while TSMC surged to an undisputed manufacturing lead. Because AMD builds at TSMC, its chips rode that leading-edge process, giving AMD's designs a manufacturing foundation equal or superior to Intel's — without AMD having to spend the tens of billions a year that leading-edge fabs require.

The fabless model let AMD concentrate its capital and talent on design, stay lean through its recovery, and convert TSMC's manufacturing excellence into its own competitive advantage. It was, in the comeback years, a genuine edge over an Intel weighed down by its own faltering factories. The essential honesty, though, is that this advantage is rented and shared: AMD's manufacturing edge belongs to TSMC, and every other fabless company — including Nvidia — uses the same foundry. Against Intel the model was a differentiator; against Nvidia it is a level playing field, since both design chips and both build them at TSMC. The fabless model was the right strategic choice and remains a sensible one, but it confers no exclusivity — it is a capability available to any well-funded designer, and it makes AMD dependent on a single, geographically concentrated supplier — the foundry holding 90%-plus of the world's leading edge¹ — it does not control.