chip manufacturing
Chip manufacturing is the industrial process in which tiny circuits are built layer by layer on silicon wafers – these become the processors and memories of phones, cars and AI systems.
In plain terms
A chip is a fingernail-sized silicon die on which up to several billion tiny switches (transistors) work. Chip manufacturing is the production of this: extremely pure silicon obtained from sand becomes a finished processor or memory chip over weeks or even months.
Imagine having to print a map of an entire large city, with streets, houses, and wiring, onto the surface of a stamp, across dozens of layers, flawlessly. The "printing press" here is a facility whose air is many times cleaner than in operating rooms, because a single grain of dust can destroy a chip.
From an artificial intelligence perspective, this is the physical foundation: large language models and AI agents all run on special chips produced in such facilities. If chips are scarce, AI services become more expensive and expand more slowly.
The industry's peculiarity is that design and manufacturing are often split: many well-known companies only design the chip and entrust its production to a few giant contract manufacturers. The most advanced chips in the world can be produced by only three companies (Taiwan's TSMC, South Korea's Samsung, and the US Intel), and among them TSMC is by far the largest.
How it works
- Raw material: a cylindrical crystal is grown from high-purity silicon, then cut into thin, mirror-finished discs, known as wafers.
- Patterning: the wafer is coated with a light-sensitive layer, and the circuit pattern is projected onto it with light through a mask – this is lithography. For the finest patterns, extreme ultraviolet (EUV) light is used.
- Layering: the exposed pattern is developed and etched, new material layers are applied, and the electrical properties of the silicon are selectively modified. This cycle repeats dozens of times, and the entire manufacturing process consists of several hundred steps, or even more than a thousand for the most advanced chips, before the transistors and the wires connecting them are built.
- Inspection and dicing: every chip on the finished wafer is measured, then the wafer is diced, and defective pieces are discarded.
- Packaging: good chips are mounted in protective casings and provided with leads. For AI chips, multiple dies – for example, a processor and memory – are often integrated into a single unit.
An everyday example
You order a new car, and the dealer says you will have to wait months because some control chips are missing – this happened on a large scale during the chip shortage that began at the end of 2020 and peaked in 2021–2022. A partly similar bottleneck lies behind it when an AI service limits usage because it does not have enough server capacity.
Who is affected
A common misconception
Limits and risks
- Building a modern chip factory costs tens of billions of dollars and takes years, so capacity cannot quickly follow a sudden surge in demand.
- Manufacturing is geographically highly concentrated, so a natural disaster, export restriction, or political conflict can disrupt global supply.
- The process requires vast amounts of water, energy, and special chemicals, posing a significant environmental burden.
- Size reduction is approaching physical limits, so each new generation is more expensive and harder to manufacture than the previous one.
Similar, but not the same
- chip design – Design involves creating the circuit's 'blueprint' with software, while manufacturing is its physical realization in the factory – the two are often done by different companies.
- data center – The data center uses ready-made chips for computation, while the chip factory produces them.
- electronics assembly – The assembly plant mounts pre-made chips onto circuit boards and installs them into devices; it does not manufacture the chips itself.
Related concepts
semiconductor industry, AI infrastructure, supply chain
lithography, Contract manufacturing, packaging, silicon wafer
GPU, MI-chip, data center, computing capacity, export restriction
A brief history
The first integrated circuit was presented by Jack Kilby in 1958 at Texas Instruments; the mass-producible, silicon-based version was developed a few months later, in 1959, by Robert Noyce at Fairchild. The foundation of today's division of labor was laid by the 1987 founding of Taiwan's TSMC: it was the first company to specialize exclusively in contract manufacturing of others' designs.
Where things stand
Chip manufacturing is a mature, global industry, driven at the end of 2026 by demand for AI chips. The bottleneck is the most advanced manufacturing and packaging capacity, as well as the supply of memory chips needed for AI servers. The United States, the EU, and several Asian countries are using state support to build their own production.
In the news
- AI chip demand lifts third-quarter growth at European chemical and industrial gas firmsNov 5, 2026 · Ipar
- Semiconductor equipment makers secure supply contracts through 2028 on AI demandNov 4, 2026 · Ipar
- Semiconductor equipment market could reach $155.13 billion in 2026 and $244 billion by 2032Oct 27, 2026 · Ipar
- Spintronics AI weighs ₹2,800 crore GaN chip fab in Madhya PradeshOct 27, 2026 · Ipar
- GlobalFoundries opens AI-enabled facilities control centre in Singapore with SiemensOct 26, 2026 · Ipar
- AI hallucinations in court filings are slowing, but judges still sound alarmsOct 19, 2026 · Jog
- Hyundai Steel launches H ONE brand for AI data centers and chip fabsOct 12, 2026 · Ipar
- VIS considers second Singapore semiconductor fab as AI demand growsOct 5, 2026 · Ipar
Also known as
chip manufacturing · chip manufacturing · semiconductor manufacturing · microchip manufacturing · chip manufacturing · semiconductor manufacturing · semiconductor fabrication · fab
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.