From sand to silicon: what a chip actually is

Almost everything with a battery or a plug now contains chips. A modern car carries well over a thousand of them. The servers that train and run AI models hold some of the most complex objects humans have ever made. Yet the starting material is about as ordinary as it gets: sand.

This first part of the series covers what a semiconductor is, how sand becomes a wafer, what a transistor does, and the main families of chips. The rest of the series builds on these ideas.

What "semiconductor" means

Materials fall roughly into three groups by how easily electricity flows through them. Metals such as copper are conductors. Glass and most plastics are insulators. A semiconductor sits in between, and that in-between position is what makes it useful: its conductivity can be controlled very precisely.

Silicon is the most widely used semiconductor. Pure silicon conducts poorly. Adding tiny, controlled amounts of other elements, a process called doping, changes that. Phosphorus adds spare electrons and makes n-type silicon. Boron leaves gaps, called holes, and makes p-type silicon. Where n-type and p-type regions meet, electricity can be steered, switched on and switched off. Every transistor is built from arrangements of these regions.

Other semiconductors matter too. Silicon carbide and gallium nitride handle high voltages and temperatures, so they are used in electric vehicle inverters and fast chargers. Gallium arsenide and indium phosphide are used in radio-frequency and optical parts. But silicon carries the vast majority of the world's chips.

From sand to a wafer

Silicon is the second most abundant element in the Earth's crust, after oxygen. It occurs mostly as silicon dioxide in quartz and sand. Turning it into something a chip can be built on takes several steps, each more demanding on purity than the last.

1. Metallurgical-grade silicon. High-purity quartz is reduced with carbon in an electric arc furnace. The result is about 98–99% pure, good enough for aluminium alloys and solar feedstock, but far from chip grade.
2. Electronic-grade polysilicon. The silicon is converted into a gas, distilled, and deposited back as solid polysilicon, most often in the Siemens process. Purity reaches nine to eleven nines: 99.9999999% or better, roughly one stray atom per billion.
3. Single crystal. Polysilicon is melted in a quartz crucible. A small seed crystal is dipped in and slowly pulled upward while rotating. This is the Czochralski method, and it grows a single cylindrical crystal, called an ingot, in which every atom sits in the same lattice.
4. Wafers. The ingot is ground to an exact diameter, sliced with diamond wire, and lapped, etched and polished until the surface is flat to within nanometres. A standard 300 mm wafer is about 0.775 mm thick.

Most leading fabs today use 300 mm wafers. Older and specialty fabs use 200 mm or 150 mm. A larger wafer holds more chips per pass through each tool, which is why the industry moved up in size over the decades.

The transistor: a switch with no moving parts

The basic building block of a chip is the transistor, and the most common kind is the MOSFET (metal-oxide-semiconductor field-effect transistor). It has three terminals: a source, a drain, and a gate between them. Applying a small voltage to the gate opens or closes a channel in the silicon underneath, letting current flow from source to drain or stopping it.

That on–off behaviour represents the ones and zeros of digital logic. Connect transistors in the right patterns and they form logic gates. Connect gates and you get adders, memory cells, processors and, eventually, an AI accelerator. NVIDIA's H100, a widely used AI chip, contains about 80 billion transistors on a single piece of silicon.

Transistors are not only switches. In analog and power chips they amplify signals or control large currents. The same physics applies; the design goals differ.

The main families of chips

Each family has its own manufacturing needs. Leading-edge logic chases ever smaller features. Memory chases density and cost per bit. Power and analog chips often run on older, larger geometries where reliability and voltage handling matter more than size. India's first fab, being built by Tata Electronics with Taiwan's PSMC at Dholera in Gujarat, is planned around nodes from 28 nm to 110 nm for power management, display drivers, microcontrollers and computing logic.

Why this matters now

Global semiconductor sales passed $600 billion in 2024, and AI is now the fastest-growing source of demand. Chips have become a question of industrial strategy for most large economies, India included. Understanding the basics makes it much easier to follow the decisions being made about where fabs go, which technologies they use and what has to be built around them.

In Part 2, we go inside a fab and follow a wafer through the steps that turn it into chips.