HOW A CPU IS MADE: FROM SAND TO BILLIONS OF TRANSISTORS

If we trace the path all the way back to the beginning, we arrive at something you might not expect: sand. The silicon used in the semiconductor industry is derived from materials like silica — the substance that makes up a large portion of sand and quartz.

But the distance between a grain of sand and the processor inside your computer is not just a simple industrial process. Along this path, chemistry, physics, optics, materials engineering, software, robotics, and machines of extraordinary precision all come together.

A CPU, or Central Processing Unit, is the component that executes program instructions. When you run a program, the processor is constantly receiving, interpreting, and executing instructions. In the simplest view:

But the CPU doesn't do this with a few large, visible components. At the lowest level, everything comes down to transistors. A transistor can be thought of as a tiny electronic switch — a switch that can control the flow of current.

Early computers didn't have anything like today's processors. One of the most famous examples was ENIAC, built in the 1940s, which used a massive number of vacuum tubes. What you see today in a tiny chip could once fill an entire room of electronic equipment.

In 1947, Walter Brattain, John Bardeen, and William Shockley at Bell Labs developed the transistor. But there was still a problem: if you wanted to connect thousands of transistors with wires, the circuit would quickly become unmanageable. The solution was the Integrated Circuit (IC).

In 1969, the Japanese company Busicom asked Intel to design a set of integrated circuits for their 141-PF calculator. But Intel engineers pursued a different idea: instead of a large set of custom circuits, they moved some of the functionality to a programmable processor.

The result was the Intel 4004, introduced in November 1971:

Today, 2,300 transistors seems like a tiny number. But in 1971, placing that many transistors on a single chip was a historic milestone. The 4004 proved that what previously required a massive collection of circuits could fit on a single small chip.

After the 4004, processors evolved at a rapid pace:

This growth didn't just mean faster CPUs. More transistors meant the ability to build:

The primary material for most semiconductor chips is silicon. Silicon isn't available in nature in the pure form needed for chip manufacturing. First, silicon must be extracted and purified to an extreme degree. Why is purity so important? Because we're building structures at the nanometer scale, and even a tiny amount of impurity can change the electrical properties of the material.

Pure silicon is melted, and using methods like the Czochralski process, a large single-crystal ingot of silicon is grown. This cylinder is called a Silicon Ingot. The ingot is then sliced into very thin discs called Wafers.

You can think of the wafer as a canvas. But the painting that will be done on it cannot be seen with the naked eye, and the paintbrush is not an ordinary brush — it's light.

Chip manufacturing takes place in an environment called a Cleanroom. In this environment, air, temperature, humidity, and contamination levels are strictly controlled. Wafers are moved automatically by robotic systems during many stages.

This is one of the most important points about chip manufacturing. No one picks up a wafer and prints billions of transistors on it in one go. A chip is built layer by layer:

In reality, a CPU is not carved — it's shaped layer by layer.

One of the key steps in chip manufacturing is Photolithography. The wafer surface is coated with a light-sensitive material called Photoresist. Then, a very precise pattern is created on it using light.

The smaller the features we want to create, the harder it becomes to control the light. For advanced chip generations, one of the most important technologies is EUV — Extreme Ultraviolet Lithography:

In EUV machines, generating the light itself is an incredibly complex process. Tiny droplets of tin are hit by laser pulses, creating a plasma that emits EUV light. This light cannot be guided with glass lenses like ordinary light — instead, highly specialized mirror systems are used.

In the simplest view, a transistor is a switch:

By controlling the electrical properties of the semiconductor material, current can be controlled. One of the important processes here is Doping — a carefully controlled amount of specific atoms is introduced into the silicon structure to change its electrical properties.

Suppose we've built billions of transistors. Is the CPU ready? No. Transistors need to communicate with each other. This is where another part of the chip enters the story: Interconnects.

Transistors are the buildings of this city. Metal interconnects are the streets, and electrical signals travel through this network. If this network isn't designed correctly, even billions of powerful transistors can't make a good CPU.

The factory doesn't know on its own where to place transistors. Before a single wafer enters the production line, the CPU must be designed:

In other words, before a CPU is a piece of silicon, it is first a massive computational design.

On a large wafer, many Dies are placed. Then it must be determined which dies are functional. The percentage of usable dies relative to all dies produced is called Yield.

The package is not just a protective frame — it must also handle electrical connections, signal transfer, heat dissipation, and mechanical protection. Today, the Chiplet concept allows different parts of the processor to be built on separate dies and connected within a single package.

In 1965, Gordon Moore made an observation that later became known as Moore's Law: the number of transistors that can be placed on a chip would increase rapidly over time. This wasn't a precise physical law — it was more of an observation and prediction about industry trends.

But as structures got smaller, new problems emerged:

The strangest part of all is that a CPU isn't built from sand — it's built from information. The raw material is silicon. But what turns it into a CPU is the design pattern, the manufacturing instructions, the masks, the transistor positions, and billions of engineering decisions.

We haven't just made matter smaller. We've written information into matter. It started with sand; it arrived at computation.