Pick up a phone or look at a data-centre server and you are looking at one of the most international objects ever made. The processor inside was almost certainly designed in one country, manufactured in another using machines from a third, and assembled with components from two or three more.
This explainer follows one advanced chip through that chain. At each step it asks the same two questions: who does this, and how many alternatives are there?
Step one: design, led by the United States
Most leading chips begin as a design, not a factory. NVIDIA is the best-known example of what the industry calls a “fabless” company: it designs processors and sells them, but owns no fabrication plants.
American firms dominate this stage. The industry’s trade association reports that companies headquartered in the United States accounted for just over half of all chip sales worldwide in 2025. The specialised software used to design chips is also largely American, though reliable public figures for that market are hard to find.
Step two: the machines, from the Netherlands
To turn a design into a physical chip, a manufacturer has to print billions of tiny features onto silicon. The machines that do the printing are called lithography systems, and the most advanced kind uses extreme-ultraviolet light.
One company makes them: ASML, based in the Netherlands. In 2025 it sold 48 of these systems. Each is among the most complex machines in existence, built from components supplied by hundreds of other firms, including specialists in Germany and the United States.
This is the narrowest point in the whole chain. There is no second source.
Step three: fabrication, concentrated in Taiwan
The designs and the machines meet in a fabrication plant, or “fab”. Here one company towers over the rest. TSMC, based in Taiwan, manufactures chips for other companies under contract. One research firm estimates that it took 72.5% of all contract chipmaking revenue in the second quarter of 2026. Its nearest competitor, Samsung Electronics, had under 6%.
The concentration is geographic as well as corporate. In 2023 the same firm estimated that Taiwan was home to about two-thirds of the world’s capacity for advanced chips.
That is changing slowly. TSMC now produces chips in Arizona and in Japan and is building in Germany, as described in our explainer on chip subsidies. A 2024 industry study projected that the United States would go from making none of the world’s most advanced logic chips in 2022 to more than a quarter by 2032. Projections of this kind depend on plants that are still being built.
Step four: memory, from South Korea
A processor is useless without memory to hold the data it works on. Memory chips are a separate industry with its own leaders, most of them in South Korea.
Samsung is the largest producer of the standard type, known as DRAM, with an estimated 38% of revenue in mid-2026. But in the specialised high-bandwidth memory used alongside AI processors, the leader is SK hynix, with an estimated half of the market. The American firm Micron is the third major producer.
SK hynix illustrates how tightly the chain is bound together. Its prospectus for its US listing in 2026 shows that a single customer provided almost a quarter of its revenue, and that it relies in turn on TSMC for part of its most advanced products.
Step five: materials and tools, from Japan
Less visible, but just as necessary, are the materials: the ultra-pure silicon wafers on which chips are built, and the light-sensitive chemicals used in printing them. Japan is the leading supplier of both. Figures cited by the US Commerce Department put Japanese firms’ share at about half of the world market for each, and higher still for some specialised equipment.
Why it is built this way
Nobody designed this arrangement. It emerged because every step rewards scale and experience.
A leading-edge fab costs tens of billions of dollars and is obsolete within years, so only companies that can fill it continuously can afford one. A lithography system embodies decades of accumulated engineering. The purity required of materials takes years to master. In each case, the leader’s advantage tends to grow.
The result is extraordinary efficiency in normal times and limited resilience in abnormal ones. If any single step were interrupted, there would be few places to turn.
Where China fits
China is the largest market for chips and a growing producer of less advanced ones, but it is restricted from buying the most advanced manufacturing equipment and processors by the controls described in our explainer on export restrictions. It holds leverage of its own through its dominance in processing certain minerals the industry needs. Its domestic memory and manufacturing firms are gaining share from a low base; how fast is one of the industry’s open questions.
A caution about the numbers
Every market share in this article is an estimate by a private research firm. They are the best information available, and different firms broadly agree on who leads. But they measure different things, some revenue and some capacity, some all chips and some only advanced ones, and they are revised.
Some of the most frequently repeated statistics in this field turn out to have no traceable source at all. We have left those out. The pattern of concentration is not in doubt; precise percentages should be held loosely.
How this connects
Because so much value and so much risk sit in this chain, it touches several other subjects on this site: the spending on AI infrastructure that drives demand for it, the export controls that govern who can buy from it, and the subsidies intended to spread it more widely.

