The foundry model is the division of labour that organises the semiconductor industry: fabless companies design chips, and foundries manufacture them for a fee. Nvidia, Apple, AMD and Qualcomm own no leading-edge factories; TSMC, which built the model, makes their silicon. The alternative is the integrated device manufacturer (IDM), which designs and manufactures its own chips, the model Intel is attempting to straddle by opening its fabs to outside customers.
Why the industry split
A leading-edge fabrication plant now costs tens of billions of dollars and must run near capacity to pay for itself. Foundries solve the economics by aggregating demand from many designers over one production base, spreading the world’s most expensive fixed costs. The price is concentration: the leading edge has consolidated to essentially one dominant manufacturer, TSMC, with Samsung and Intel chasing, and much of it located in Taiwan.
What it means for investors
The model shapes how value flows through every AI headline. Foundries earn manufacturing margins on volume and utilisation; designers capture product margins but carry allocation risk when capacity is scarce. Foundry capex is the clearest window into real chip demand, because it is spent years ahead of revenue, and “foundry customer wins” are the scoreboard for whether a challenger’s manufacturing is credible.
The strategic layer
Because the leading edge is concentrated in one place, the foundry model is also a geopolitical fact: export controls, subsidy programmes like the CHIPS Act, and the push to build fabs in the US, Europe and Japan are all attempts to manage its concentration. When governments subsidise semiconductor plants, they are buying insurance against the model’s single point of failure.
Where you’ll meet this in our coverage
Intel Q2 2026 Earnings: The Fastest Growth in 15 Years Lands in a Bear Market
TSMC Q2 2026 Earnings: The Bill for Proving the AI Build-Out Is Real