Sand. That is where this starts.

Not metaphorical sand. Actual silicon dioxide, mined from beaches and riverbeds, purified to 99.9999999 percent purity, sliced into wafers thinner than a human hair, and then etched with patterns so small that the wavelength of visible light is too coarse to draw them. You need extreme ultraviolet light for that.

One company in the Netherlands makes the machine that produces that light. One company. ASML. There is no backup. There is no substitute. If that factory in Veldhoven burns down tomorrow, the global chip supply goes into a multi-year freeze.

That is not a metaphor for fragility. That is the actual situation.

I think about this every time someone says the word infrastructure in a speech and means roads. Roads are infrastructure. Chips are something else. Chips are the substrate on which modernity runs. Your roads do not function without traffic management systems.

Your hospitals do not function without imaging equipment. Your banks do not function without compute. Your military does not function without guidance systems. Everything that makes a modern state operate runs on processed sand.

I. What a Chip Actually Does

Most people have a vague understanding that chips are important. Fewer understand what they actually are.

A semiconductor chip is, at its core, a switch. A transistor switches current on or off. One transistor is useless. A billion transistors on a fingernail-sized die, switching billions of times per second, arranged in logic gates that can add numbers, compare values, store state, and execute instructions, that is a processor. The latest Apple M4 chip has around 28 billion transistors. Apple does not manufacture it. TSMC in Taiwan does.

Someone at a dinner party: But surely America could just build those fabs if it wanted to?

Me: It is trying. The CHIPS Act in 2022 put 52 billion dollars toward that goal. Intel is building fabs in Ohio. TSMC is building fabs in Arizona. The Arizona fab is expected to reach volume production sometime around 2026. The gap between intention and execution is roughly a decade and several hundred billion dollars.

Someone at a dinner party: So it will be fine eventually.

Me: Eventually. Ask yourself what happens between now and eventually if Taiwan becomes a conflict zone.

That is the conversation most people are not having loudly enough.

Taiwan produces around 90 percent of the world's most advanced chips. Not 30 percent. Not 50 percent. Ninety. The entire global tech ecosystem, from consumer electronics to cloud infrastructure to weapons systems, runs through a 36,000 square kilometer island that China considers a breakaway province and the United States has treaty commitments to defend without those commitments being entirely explicit.

This is not a tech story. This is a geopolitical story that happens to involve technology.

II. The Supply Chain Nobody Thought About Until It Broke

In 2021, a drought in Taiwan reduced water availability for chip fabs. Chip manufacturing uses an enormous amount of ultrapure water for wafer washing. TSMC uses around 156,000 tonnes per day. The drought was a warning. Nobody planned for it because nobody had to plan for it before.

Then COVID happened and the auto industry, which had confidently cancelled chip orders in early 2020 expecting a demand collapse, watched demand recover faster than expected and found itself unable to source the basic microcontrollers that regulate fuel injection, window motors, and airbag deployment. Ford lost 2.5 billion dollars in profit in 2021 because it could not get chips that cost less than a dollar each. General Motors idled plants. Vehicles sat without dashboards.

The chip shortage did not expose a flaw in the chip industry. It exposed a flaw in how the entire manufacturing economy thought about risk. Everyone had optimised for cost and speed. Nobody had optimised for resilience. Those are different things and they have different price tags.

I have a specific memory of reading about this in early 2021 and thinking: this is what happens when decades of just-in-time supply chain logic meets a single-point-of-failure manufacturing geography. Toyota had buffer stock of chips because Toyota's manufacturing philosophy includes buffer stock for critical components. Most Western automakers did not. Toyota lost less.

The lesson is not complicated. The execution is.

III. The Conversation Between Governments

Here is what the semiconductor trade war actually looks like from the inside of a supply chain conversation.

In October 2022, the Biden administration issued export controls preventing American companies from selling advanced chips or chip-making equipment to China without a license. Not just chips. The tools to make chips. The engineers with American citizenship who worked at Chinese fabs were told they had to choose between their job and their passport. Hundreds left.

The logic is simple. The most advanced chips are around 5 years ahead of what China can currently manufacture domestically. If you stop China from getting those chips and stop it from getting the equipment that makes them, you slow its ability to develop next-generation AI systems and military applications that depend on compute.

The obvious question: Does this actually work?

Me: It buys time. It does not solve the problem. China has allocated around 143 billion dollars toward domestic semiconductor development. It has hundreds of thousands of engineers in the field. In 10 to 15 years, assuming competent execution, the gap closes. The controls are a delay tactic, not a permanent solution.

The follow-up question: So why do it?

Me: Because 10 years is a long time in technology. If you are the United States and you believe the next era of geopolitical power is determined by who leads in AI, quantum computing, and advanced manufacturing, you want those 10 years. You use them to build domestic capacity, to build alliances, to widen the lead while the controls hold.

The Netherlands matters here too. ASML, the only company that makes extreme ultraviolet lithography machines, is Dutch. The United States pressured the Netherlands government to restrict ASML from selling its most advanced machines to China. The Dutch government complied. This is what a chip embargo looks like in practice: quiet conversations between foreign ministries about export licenses for machines that most citizens have never heard of.

IV. What I Saw From Inside the Supply Chain

I have spent time close to this world, not as a chip designer but as someone who has watched how industrial deployments think about component sourcing. The conversation I keep seeing repeat itself goes roughly like this.

Engineering team: We need to redesign the board because the microcontroller we have been using for six years is end-of-life and the replacement has a 52-week lead time.

Operations: 52 weeks is more than a year. We have customer commitments in Q3.

Engineering team: I know. The alternative is the part from a manufacturer in Shenzhen that has 8-week lead time but no long-term supply commitment and we have no visibility into their process reliability.

Operations: So our options are wait a year or take a risk on an unknown supplier.

Engineering team: Essentially yes. Welcome to 2024.

This is not an edge case. This is a normal procurement conversation happening across thousands of companies. The chip shortage technically ended in 2023, but the underlying geography did not change. The concentration did not change. The geopolitical exposure did not change.

What changed is that companies are now keeping more inventory, qualifying second sources, and thinking harder about which components are genuinely critical versus nice to have. That is progress. It is not enough.

V. Literacy as a Political Act

I want to say something about why any civilian should care about this.

When you hear a politician talk about decoupling from China, you are hearing a conversation about fab geography. When you hear about the CHIPS Act, you are hearing about the time lag between policy and productive capacity.

When you hear about AI export controls, you are hearing about compute as a weapon. None of this is abstract. All of it eventually shows up in the price of your phone, the wait time for your car, the reliability of your bank's infrastructure, and the foreign policy commitments your government makes.

Semiconductor literacy is not technical education. It is civic education. If you cannot understand why a single fab in Taiwan is a reason for both diplomatic caution and military planning, you cannot fully understand the foreign policy choices your government is making in your name.

The breath hacker principle here is simple. Most people learn about power by watching it be exercised visibly. Wars, elections, sanctions, speeches. The real levers are invisible. They are underground cables, undersea fibre, satellite orbits, fab yields, and the purity of silicon wafers. The people who understand those levers have asymmetric influence over those who do not.

You do not need to become a semiconductor engineer. You need to know enough to ask the right questions when your government tells you it is making decisions in your interest.

Oil was the 20th century's most strategic resource because industrial economies ran on combustion. The 21st century runs on computation. The strategic resource is not crude from the ground. It is purified silicon, exotic gases, and the nanometer-scale precision to etch logic into matter.

The concentration of that capacity in Taiwan, with dependencies running through the Netherlands, Japan, South Korea, and the United States, is not an accident. It is the product of decades of investment, specialisation, and deferred political risk. Deferred risk does not disappear. It accumulates.

We are now in the period of paying down that debt. The CHIPS Act, the export controls, the reshoring conversations, the diplomatic pressure on allies about fab access: all of it is a belated recognition that the world built a critical system on a single point of failure and called it efficiency.

It was efficient. It is now fragile. Those are related.

Efficiency and resilience are not the same value. Every decade spent optimising for cost and speed is a debt you eventually pay in fragility. The companies and countries that survive disruption are not the ones that cut the most fat. They are the ones that kept enough slack to absorb a shock they did not predict. Resilience is expensive until the moment you need it. Then it is priceless.

The sand was always there. We just forgot what it took to turn it into something that thinks.