Foundry vs fabless
Also known as: fabless, foundry, pure-play foundry, chip foundry, fabless semiconductor company, IDM
A fabless chip company designs semiconductors but outsources manufacturing, while a foundry manufactures chips designed by other companies.
Nvidia, AMD, Apple, and Qualcomm are fabless; TSMC, Samsung Foundry, GlobalFoundries, and Intel Foundry make chips for customers. While this split looks like a simple division of labour, it concentrated the world's most advanced chipmaking in a handful of factories, mostly in Taiwan, and that's now a geopolitical issue.
Foundry vs fabless, explained at three levels
At home
Think of a fabless company as an architect and a foundry as a construction company. Apple designs the chip in your iPhone, but TSMC builds it in its factory. Most famous chip brands don't actually make their own chips.
At work
For a business, the fabless model means you can create custom chips without building a factory that costs tens of billions of dollars. You pay a foundry to manufacture your design, which lowers the barrier to entering the chip market.
At home
Chip factories, called fabs, are extraordinarily expensive and complex. A leading-edge fab can cost more than $20 billion. That's why only a few companies can afford to make the most advanced chips, and why designers share them rather than each building their own.
At work
Fabless companies compete on design and software, while foundries compete on process technology, yield, capacity, and trust (they must protect customers' designs). Capacity allocation at leading nodes is a strategic issue: during the AI boom, access to TSMC's advanced capacity and packaging shaped which companies could ship products.
At home
The model works because of standardized design tools and process design kits that let designers target a foundry's manufacturing process. In my opinion, the hidden cost is dependency: when most leading-edge chips come from factories in one region, any disruption, from earthquakes to conflict, could ripple through the entire tech economy.
At work
Governments are now subsidizing domestic manufacturing through measures like the US CHIPS and Science Act and the EU Chips Act, and TSMC, Samsung, and Intel are building fabs in the US, Japan, and Europe. Yet leading-edge production remains concentrated, overseas fabs often cost more to run, and the ecosystem of suppliers and skilled workers can't be relocated quickly. Supply chain planning should assume concentration persists for years.
Real-world examples
At home
Sophie assumed Apple made the chips in her MacBook. Reading about the M-series processors, she learned Apple designs them, but TSMC manufactures them in Taiwan. The same factory makes chips for many competing brands.
At work
Thomas runs hardware strategy at a Canadian startup designing chips for industrial sensors. Rather than raising billions for a fab, his team licenses design tools and books production with a foundry. Their biggest risk isn't design. It's securing enough manufacturing capacity at a mature process node.
How it compares
| feature | Fabless | Foundry | IDM |
|---|---|---|---|
| What they do | Design chips | Manufacture chips for others | Design and manufacture their own chips |
| Examples | Nvidia, AMD, Apple, Qualcomm | TSMC, Samsung Foundry, GlobalFoundries | Intel, Texas Instruments |
| Capital needs | Lower | Enormous | Enormous |
| Main risk | Dependent on foundry capacity | Massive investment, customer concentration | Keeping pace on both design and manufacturing |
Fabless companies design, foundries build, and IDMs do both.
For analysts
When evaluating hardware suppliers, ask where their chips are actually made. I'd add foundry concentration to the risk register for any product that depends on advanced chips, because a capacity crunch can delay shipments regardless of who designed the part.
History & origin
For decades, chip companies were 'integrated device manufacturers' (IDMs) that designed and built their own chips, as Intel still largely does. In 1987, Morris Chang founded Taiwan Semiconductor Manufacturing Company (TSMC) with Taiwanese government backing, betting that a company making chips only for others could thrive. It did. Fabless pioneers such as Xilinx and Chips and Technologies emerged in the 1980s, and Nvidia (1993) grew up fabless. AMD spun off its factories as GlobalFoundries in 2009. Intel announced its own foundry push in 2021, and in 2025 the US government took a roughly 10% stake in Intel. By the mid-2020s TSMC held around two-thirds of the global foundry market and most leading-edge production. In my view, the foundry model is one of the most successful business innovations in tech history, and also one of its biggest single points of failure.