Semiconductors: Everything Depends on a Handful of Buildings

Almost every object with electricity running through it contains semiconductors — phones, cars, washing machines, medical equipment, weapons, the servers behind everything online.

They are, in one sense, an ordinary manufactured good. In another, they are unlike anything else in the modern economy, because of a single structural fact: designing a chip is competitive and widely distributed, while manufacturing the most advanced ones is concentrated to an extraordinary degree.

That asymmetry explains most of what happens in the industry, including why an unremarkable shortage of cheap chips stopped car factories worldwide, and why a component category became a central subject of foreign policy.

Why manufacturing concentrated and design did not

Chip design requires talented engineers and expensive software. It is difficult, and hundreds of organisations do it.

Manufacturing at the leading edge requires a fabrication plant, and a modern one is among the most expensive objects human beings construct. It must be built in conditions cleaner than an operating theatre, using machines that manipulate matter at scales close to the size of individual atoms. And because the technology advances continuously, the plant must be substantially rebuilt every few years just to remain current.

Follow that economics and consolidation is inevitable. Only a few organisations in the world can sustain that level of capital expenditure repeatedly, and the ones that fall behind cannot catch up, because catching up means outspending the leader while earning less than them. The result is not a market failure — it is what these particular costs do.

The concentration goes further than most people realise. The machines used for the finest patterning come from a very small number of suppliers. Certain specialist chemicals and materials have similarly narrow sources. The chain has several points where the number of viable suppliers is countable on one hand.

This is a textbook theory of constraints situation applied to an entire industry: the whole world's electronics output is governed by the throughput of a small number of specific facilities, and no amount of investment elsewhere in the chain changes that.

Why a small shortage caused a large crisis

The chip shortage that halted vehicle production is worth understanding because the mechanism was not what most coverage suggested.

Demand for cars had not surged. What happened was closer to a coordination failure amplified by the structure of the supply chain — a textbook bullwhip effect. Orders were cut early in a period of uncertainty, capacity was reallocated to other customers, and when orders returned, the queue for capacity was long.

Two features of semiconductors made this worse than it would be in most industries.

Lead times are long and capacity is rigid. You cannot add a production line in response to a spike. The equipment takes years to install and commission. Supply is essentially fixed in the short run, so all adjustment happens through queueing and price.

The chips that caused the problem were the cheap ones. Vehicles need many simple, mature components, not cutting-edge processors. Those older-generation parts are produced on older equipment that nobody is eager to expand, because the margins are thin. A car cannot ship without a part costing a few dollars, so a low-value component gated an extremely high-value product.

That is the general shape of supply chain fragility: the binding constraint is rarely the expensive, visible input. It is usually something cheap that everyone assumed would always be available.

How chips became a matter of state

A component becomes strategic when two things are true at once: everything depends on it, and it cannot be sourced elsewhere within a reasonable timeframe. Leading-edge semiconductors satisfy both.

This turns commercial decisions into political ones. Where a plant is located stops being purely a question of cost and tax. Who may buy which class of chip becomes a policy instrument. Export controls on manufacturing equipment become a way of shaping another country's capabilities years into the future, because a fab that cannot buy the machines cannot advance.

Several governments have responded by subsidising domestic manufacturing. Whether this works is genuinely uncertain, and worth being honest about rather than confident. Building a plant is achievable with enough money. Building the surrounding ecosystem — the specialist suppliers, the experienced engineers, the accumulated process knowledge that exists mainly as institutional habit — takes far longer and cannot simply be purchased. Some national efforts will likely produce capable domestic industries; others will likely produce expensive facilities that remain a generation behind.

The deeper tension is that the concentration everyone now worries about was not an accident or a policy failure. It was the efficient outcome of letting the most capable producer serve the world. Efficiency and resilience were traded against each other, mostly without anyone deciding to make the trade — and unwinding it means accepting higher costs permanently.

Which is the honest summary of the industry's position: the thing that made chips cheap and abundant for decades is the same thing that makes the supply fragile, and there is no arrangement available that delivers both.