Semiconductor manufacturing
The Machine That Only One Company on Earth Knows How to Build
Every advanced chip in every smartphone, data centre, and AI system alive today was made using a machine so complex that its supply chain spans hundreds of companies across a dozen countries — and if the single Dutch factory that makes it burned down tonight, the global tech industry would grind to a halt within two years.
The Idea
At the heart of modern chip manufacturing is a process called extreme ultraviolet lithography — EUV for short. The basic idea of lithography is ancient: you use light to project a pattern onto a surface. In chipmaking, that pattern becomes the circuitry etched into silicon. For decades, engineers pushed lithography further by using shorter and shorter wavelengths of light, because shorter wavelengths can resolve finer details — and finer details mean more transistors, more power, less energy. EUV takes this to a surreal extreme. The light used has a wavelength of just 13.5 nanometres, which is not visible light at all but soft X-ray territory. To generate it, a machine fires 50,000 tiny droplets of molten tin per second, hits each one twice with a high-powered laser, and the resulting plasma emits EUV light. That light is then bounced off a series of mirrors so precise that if they were scaled to the size of Germany, the largest imperfection would be smaller than a marble. A single EUV machine contains more than 100,000 components, weighs as much as two fully loaded commercial aircraft, and costs roughly the equivalent of a small nation's annual science budget. Only one company — ASML, headquartered in Eindhoven — has ever successfully built one. The physics is so demanding, and the accumulated engineering knowledge so deep, that no rival has come close.
In the World
In 2020, the United States government added ASML's EUV machines to its export control list, preventing them from being shipped to China. It was a single policy decision — but the geopolitical tremor it caused revealed just how much of the modern world rests on this one chokepoint. TSMC in Taiwan, Samsung in South Korea, and Intel in the US are the only fabs currently operating EUV machines. Every chip powering an iPhone, a cloud server, or a generative AI model traces its existence back through that lineage. China, despite pouring tens of billions into its domestic semiconductor industry, has been unable to replicate EUV — not for lack of money or engineers, but because the machine depends on optical components from Germany, laser technology from California, and control software built over three decades of incremental refinement. ASML itself was nearly bankrupt in the 1990s and survived partly through a quiet investment from Intel, which wanted the technology to exist. The machine's story is less a tale of lone genius than of an extraordinarily fragile, extraordinarily intricate web of global cooperation — one that geopolitics is now actively trying to cut. Taiwan produces over 90% of the world's most advanced chips. That number, in light of everything above, is either a marvel of industrial concentration or one of the great systemic risks of the modern era, depending on the morning you're having.
Why It Matters
Most of us interact with the outputs of semiconductor manufacturing dozens of times a day without any sense of how physically extraordinary — or precarious — the process is. Understanding that a single company in a mid-sized Dutch city sits at the apex of the entire global technology stack changes how you read the news. Export controls, Taiwan tensions, AI investment races — these aren't abstract geopolitical chess moves; they are, in large part, fights over access to machines and the knowledge baked into them. It also reframes how we think about technological progress. The relentless march of Moore's Law — the doubling of transistor density roughly every two years — has never been inevitable. It has always been the result of specific people, in specific places, solving problems that seemed physically impossible, funded by fragile coalitions of interest that could easily have collapsed. That's not a reason for anxiety. It's a reason to find the whole enterprise genuinely astonishing.
A Question to Ponder
When something as consequential as the world's most advanced chip-making technology depends on a single supplier, a single island, and decades of irreplaceable tacit knowledge — how should societies think about the difference between efficiency and resilience?
Get a new one of these every morning.
Start learning with Thinkable