How Are CPUs Made? Inside the Complex Journey from Sand to Silicon

This article reveals the process of how CPUs are manufactured. Most people know CPUs are made of silicon, but the detailed processes and technologies that turn sand into these remarkable chips are often unclear; let’s explore them.

On this planet, aside from oxygen, silicon is the most abundant element in the crust. All electronic components are extracted from sand, where silicon exists as silicon dioxide (SiO₂).

Once silicon is purified to the required level (99.9999999% purity), it forms a continuous 100‑kilogram silicon ingot.

The journey from a pile of sand to the final product involves many complex steps, used by Intel, AMD, ARM, and others.

The CPU—central processing unit—is the most densely integrated circuit core by volume and the only component that cannot be counterfeit; its manufacturing process represents the pinnacle of modern technology.

The processor manufacturing process can be roughly divided into raw sand selection (quartz), purification into silicon ingots, wafers, photolithography, etching, ion implantation, metal deposition, metal layers, interconnects, wafer testing and dicing, core packaging, grading tests, packaging, and market release.

Silicon melting into ingots: through multiple purification steps, electronic‑grade silicon (EGS) is obtained, with at most one impurity atom per million silicon atoms.

The single‑crystal silicon ingot is cylindrical, about 100 kg, with 99.9999% purity. It is sliced horizontally into circular silicon wafers, which are polished to near‑perfect smoothness.

Photoresist coating: a blue photoresist liquid is poured onto the rotating wafer, forming a thin, even layer.

The photoresist layer is then exposed to ultraviolet light through a mask, making the exposed areas soluble. The mask contains the pre‑designed circuit pattern, so UV exposure transfers this pattern onto the wafer.

A single wafer can yield hundreds of processors; focusing on one, we see how transistors are made. Transistors act as switches controlling current flow, and about 30 million of them can fit on the tip of a needle.

Developing the photoresist: after UV exposure, the soluble regions are washed away, leaving the pattern that matches the mask.

Ion implantation: in a vacuum, accelerated ions doped with specific atoms are shot into the silicon, forming a doped layer that changes the silicon’s conductivity. The ion beam can exceed 300,000 km/h.

After ion implantation, the photoresist is removed, leaving the implanted regions (shown in green) with different atoms.

At this point, transistor fabrication is essentially complete. Then, three holes are etched in the insulating layer (red) and filled with copper to interconnect transistors.

Copper plating: a layer of copper sulfate is electroplated onto the wafer, depositing copper ions onto the transistors, forming a thin copper layer.

Polishing: excess copper is polished away, revealing the metal interconnect layer, about 500 nm thick, forming a multi‑layered high‑speed highway of circuits.

Wafer testing: each chip is functionally tested against reference patterns, and defective cores are discarded.

The wafer is then diced into 300 mm (12‑inch) dies, each representing a processor core.

The image shows an Intel Core i7 core.

CPU packaging: the substrate (with pins), the die, and the heat spreader are stacked together, forming the visible processor package, about 20 mm (1 inch) in size.

The final grading test identifies each processor’s key characteristics—maximum frequency, power consumption, heat output—and determines its performance tier.

After manufacturing and testing, the processors are packaged and shipped to OEMs or retail markets.

Different CPUs serve various purposes, from rendering games and videos to powering complex film production pipelines.

Today, all CPUs are manufactured following similar rules—x86, ARM, DRAM, SoC, ASIC—differing mainly in wiring, instruction sets, and specific process technologies.

Author: 懒豆豆. Optimized by 21CTO community.