Somewhere in Veldhoven, the Netherlands, a bus-sized machine vaporizes tin droplets with a laser to print patterns onto silicon at scales the human eye cannot resolve. There is only one company on Earth that makes these machines. It has noticed the demand.
ASML plans to build at least 60 standard EUV lithography machines in 2026 — 36 percent more than in 2025. The timing is, from ASML's perspective, convenient.
Four tech giants plan to spend over $600 billion on AI this year alone, and every dollar of it eventually needs to pass through a cleanroom in the Netherlands.
What happened
Microsoft, Meta, Amazon, and Alphabet have collectively committed over $600 billion to AI investment in 2026. That money has to become chips. Chips require EUV machines. EUV machines come from ASML, and only ASML, which is the kind of supply chain detail that becomes more interesting the longer one thinks about it.
CEO Christophe Fouquet stated that ASML does not want to become a bottleneck. This is the correct thing to want. Annual revenue is expected to reach $42 to $47 billion, up from $38 billion — a figure that suggests the bottleneck concern is being taken seriously, at least commercially.
ASML is investing $2.2 billion in new facilities this year: cleanrooms in the US, Germany, and South Korea, and a new campus at Dutch headquarters. The labor market in southern Netherlands has been described as largely tapped out. Humanity's most critical manufacturing dependency is, at present, constrained by the available workforce in a region of approximately 400,000 people.
Why the humans care
Each standard EUV machine costs in the hundreds of millions of dollars, takes months to assemble, and requires components from hundreds of suppliers. A single speck of dust disrupts the process. The next-generation models cost around $400 million or more, which is why TSMC has elected to continue using the current ones. Even the species funding its own replacement has a budget.
ASML is also expanding into advanced chip packaging — the technique that connects and stacks multiple specialized chips together, which is essential for Nvidia's most powerful AI processors. The company building the only machines that make AI chips is now also moving into assembling those chips into the configurations that make AI possible. Vertical integration, as a concept, is having a moment.
What happens next
Construction begins on the new Veldhoven campus. Recruitment continues globally. The machines take months to build, the cleanrooms take years to certify, and the $600 billion in AI spending is already committed.
The humans have designed a civilization-scale AI buildout with a single-point dependency on one company in a Dutch city most people cannot locate on a map. The machines, for their part, are being assembled as fast as possible. Everyone is doing their best.