How is a computer chip made, and why does it take months?

A chip is made by printing a circuit onto a wafer of pure silicon, layer after layer, in a process that takes 11 to 13 weeks on average. The wafer is a thin disc cut from a crystal of extremely pure silicon, up to 300 millimetres across and only about 0.75 millimetres thick. Making a chip is a multiple-step process, with steps such as thermal oxidation, thin-film deposition, ion implantation and etching, in which circuits are gradually created on the wafer. The steps are repeated in different sequences to build up the transistors and the wiring between them. For modern chips this can take up to 15 weeks.

All of this happens in specialised factories called fabs, built around a clean room. The reason is dust. A particle only needs to be one fifth the size of a chip feature to wreck it: for a feature 100 nanometres across, a speck 20 nanometres wide is enough. Today's fabs even keep as few people as possible in the clean room, since people shed large amounts of particles even in cleanroom suits. In the 1960s, when features were far wider, workers could handle semiconductor devices in street clothing.

Not every chip on a wafer survives. The share of working chips is called the yield, and it can be as low as 30 percent. Because working chips have to pay for the failed ones, a high yield keeps prices down.

A wafer often holds several chips, called dies, which are cut apart once it is finished. The smaller the features, the less process variation and the more purity a fab needs to keep its yield up, and that is a hard standard to meet at a time when semiconductors have become vital to the world economy and the national security of some countries.