Semiconductors 101: Chips, wafers and how they are made

Chips, wafers, and semiconductors – these terms are often used interchangeably, but they do not technically refer to the same thing.

Semiconductor: A material that is able to act as both a conductor and an insulator of electricity. Examples include silicon, germanium and sapphire. Silicon is particularly popular for use in manufacturing, partly due to its abundance; it is the second-most abundant element on Earth, after oxygen.

Wafer: A thin circular slice of semiconductor material that forms the base of a chip. The silicon used to make wafers needs to be 99.9999 per cent pure, a level known as “six nines”. Standard wafers are 300 millimetres in diameter. Wafers with a larger diameter are considered more cost-effective, as more chips can be cut per unit of area, due to how they fit within a circle.

Integrated circuit (IC): Also known as microchip or chip, this is a self-contained device; an assembly of electronic circuits on a small piece of silicon. ICs contain components such as resistors and up to billions of transistors, which are semiconductors that amplify or switch electrical signals.
 

From wafer to chip

First, electronic circuits are fabricated on wafers. Patterns of ICs are drawn onto a wafer using beams of light, in a process known as photolithography or UV lithography. This is followed by etching, in which chemicals are used to selectively remove unwanted layers from the wafer’s surface, leaving behind the desired pattern.

The wafer is then cut into small pieces. Each piece, known as a die, forms the basis of an IC chip.

Process node: Also known as technology node, these are measured in nanometres (nm). Historically, the measurement referred to the size of transistors on the chip. However, the number has since become arbitrary, and simply conveys how advanced a chip is compared to previous ones.

A smaller number implies that a chip has a greater density of transistors and is thus faster. Mature-node chips, also known as legacy chips, use process nodes of 28 nm or more. Leading-edge artificial intelligence chips typically have process nodes of 7 nm or less.

Packaging: This refers to casings of metal, ceramic, glass or plastic, which protect IC chips and connect them to other chips or a device’s circuit board.
 

What are AI chips

Chips can have different functions; for example, memory chips typically store and retrieve data while logic chips perform complex operations that enable the processing of data.

AI chips are logic chips, processing the large volumes of data needed for AI workloads. Their transistors are typically smaller and more efficient than those in standard chips, giving them faster processing capabilities and smaller energy footprints.
 

Types of chip companies

Integrated device manufacturer: These firms both design and make their own chips, controlling the full process from planning and manufacturing to sales. This is the traditional business model of semiconductor companies and continues to be embraced by companies such as Intel, Samsung, and Texas Instruments.

Fabless: Companies that design but do not make their own chips; they are called fabless as they do not have fabrication plants. The first fabless company was founded about 40 years ago. This business model is becoming increasingly popular as it allows firms to concentrate on chip design and outsource production to a third-party foundry. Fabless chipmakers include Nvidia, Qualcomm, and MediaTek.

Foundry: Also known as fabs, or fabrication plants, foundries manufacture chips for fabless companies. TSMC founder Morris Chang pioneered what he called the “pure-play foundry” model in 1987. The company remains the largest foundry in the world and counts Apple, Nvidia, and AMD among its biggest customers.
 

Source: The Business Times © SPH Media Limited. Permission required for reproduction.