The ASML lithography monopoly comes down to one fact: every advanced chip in the world, in every phone and data center and AI accelerator, is printed on a machine made by one company in one Dutch town. No second supplier. No domestic alternative any government has managed to build. Twenty years ago I watched this company pass Nikon on the strength of a bet on water, and now it sits at a point every roadmap in the industry has to route through. The only questions left are about how it uses a position nobody has ever held.

TSMC’s A16 brings backside power delivery into the angstrom era, Intel’s 18A-P sharpens the company’s most advanced process, and both lean on EUV layers printed by ASML scanners. Pick any leading-edge node from any manufacturer and the same machine sits in the middle of it. The foundry fight is real and brutal, and it is being waged entirely on ASML’s equipment, between customers who all buy from the same supplier. Whoever wins the node race, ASML sells the scanners. It never has to back a winner because it supplies all of them, and that is the safest seat in the industry right up until it becomes the reason every government on earth wants a say in what ASML does.

High-NA has moved from a single demonstration tool to real installations in real fabs. Intel went first and has bet its whole comeback on the technology, with the 14A node built around it. TSMC went the other way and decided High-NA is simply too expensive for now, staying on its proven EUV flow for the layers it cares about and leaving High-NA for some later node. That split tells you the economics I questioned two years ago are still not settled, even with the machines sitting on fab floors, and the first production chips off them are still only months away rather than here. What is settled is the direction. The leading edge is moving to High-NA when ASML says it is ready and at the price ASML sets, and the only argument left is how fast each customer jumps. A tool that costs the better part of half a billion dollars has a captive set of buyers, because the alternative is falling off the roadmap altogether.

Hyper-NA is already on the roadmap, a further widening of the aperture toward roughly 0.75, aimed at the nodes of the 2030s. It is years out, harder and more expensive than High-NA, and Zeiss will once again have to invent optics that do not yet exist. I have learned not to bet against ASML on the engineering, because the company has spent twenty years making impossible machines work through sheer refusal to quit. What I will say is that each generation now serves a smaller pool of customers who can afford it. The cutting edge is narrowing to a handful of companies buying machines that cost as much as a wide-body jet, and that concentration carries consequences nobody has fully reckoned with.

EUV has never reached a Chinese fab and will not under the current order, and the high end of DUV is restricted on top of it. That cap on China now looks permanent. SMIC has pushed impressively far on older immersion tools and clever multi-patterning, but I wrote earlier this year about the wall it hits without EUV, and that wall has not moved an inch since. China is pouring staggering resources into building its own lithography, and it may eventually turn out a domestic DUV tool worth using. An indigenous EUV scanner is a different mountain, the one that took ASML and its partners decades and tens of billions to climb. Beijing will keep trying because it has no choice, and the gap between trying and shipping a production EUV machine is measured in years no amount of money obviously buys back.

The entire modern economy, from the phone in your pocket to the data centers feeding the AI boom, runs on advanced chips, every one of them printed by a single company’s machines, and that company leans on a supply chain centered on one campus in the Netherlands and one optics maker in Germany. That is concentration risk on a scale the industry rarely says out loud. A fire, an earthquake, a war, one bad enough political miscalculation, and the chokepoint becomes a stoppage. Governments have spent years treating ASML as a weapon to point at China. Far fewer of them seem to be thinking about what happens if the chokepoint they all lean on fails for reasons that have nothing to do with policy.

I have followed ASML for twenty years, from the immersion bet that put it ahead to the EUV gamble that should have sunk it and somehow did not. The engineering story is one of the most impressive in modern industry, a company that solved problems everyone else wrote off as impossible and earned its position fairly. The position it earned is the problem. ASML built the one tool the future of computing cannot do without, and in doing it became something no equipment maker was ever supposed to be: a fault line running under the entire industry, sitting in plain sight of every government that matters while none of them can build their own. The hardware has never been the part that worries me. It is the single campus in the Netherlands, the lone optics maker in Germany, and a digital economy that now balances on a supply chain one bad week could break, with almost nobody outside the industry able to name the company holding it all up.