ASML Doesn't Make Its Own Machine. That's the Whole Trick.

ASML's machine is the most complex object regularly manufactured by humans, and ASML does not build most of it. It buys it. Roughly four-fifths of what goes into a system that prints the circuitry of the entire modern world arrives from outside the company - from a network of about 5,150 suppliers that ASML coordinates rather than owns.⁴⁷ The company calls itself, in its own filings, a 'system integrator' that focuses on final assembly and systems engineering.⁷ That is a polite way of saying it is an assembler with a very good address book.

The official story is that ASML won by vertically integrating - by bringing the hard physics in-house and refusing to depend on anyone. The real story is almost the reverse. ASML was born an outsourcer, stayed an outsourcer, and turned outsourcing into the deepest moat in technology. The trick was not making the parts itself. It was making sure no one else could buy them.

It started with no money to make anything

ASML did not choose to be lean; it had no choice. Founded on 1 April 1984 as a joint venture between Philips and ASM International - each partner putting in 2.1 million guilders, a little over four million between them - it was a thinly capitalized spin-out tasked with commercializing a Philips lithography project.⁶ There was no money to build optics, no money to build motors, no money to build the light sources. So it bought all of it. The outsourcing model wasn't a stage the company grew out of; it was the foundation it was poured on. The famous image - a leaky shed in Eindhoven - is the founders' own telling. The point of it is the same in either version: this was a company that began with nothing to integrate.

Which raises the obvious question. If ASML never owned its supply chain, how does it hold a position no competitor has dented in decades - sole production of EUV lithography systems, and the dominant share of the advanced DUV immersion machines that ran the world before EUV arrived?

Don't build the supplier. Capture it.

ASML's answer is the thing almost everyone misreads as vertical integration. It is something stranger and harder to copy: a proprietary ecosystem of captive, co-invested, sole-source partners. When the EUV light source became the single hardest problem standing between ASML and the next generation of chips, ASML did not try to grow that capability in-house. It bought the company that already had it. In October 2012 it announced the acquisition of Cymer, the light-source specialist, in a deal valued at EUR 1.95 billion, explicitly to accelerate EUV; the deal closed in May 2013.¹² Cymer wasn't a supplier anymore. It was inside the tent.

Zeiss is the more revealing move, because ASML deliberately stopped short of owning it. The optics inside an EUV scanner - mirrors polished to tolerances measured against the size of an atom - come from Carl Zeiss SMT, which ASML calls its most important strategic partner.⁸ ASML could have tried to acquire it. Instead, in late 2016 it took a 24.9% minority stake for €1 billion, closing in 2017, and left Carl Zeiss AG as the majority owner.³⁸ That number is not an accident. A quarter is enough to co-invest, co-develop the High-NA optics, and bind Zeiss's roadmap to ASML's; it is small enough to avoid owning Zeiss's factories, its balance sheet, and the antitrust headaches of swallowing your only optics supplier whole. ASML bought the relationship, not the company.

Why the assembler keeps the profit

If ASML buys most of its machine, you'd expect the value to leak out to the suppliers. It doesn't. In FY2024 ASML reported €28.3 billion in net sales at a 51.3% gross margin, with €4.3 billion poured back into R&D and 44,027 employees on the books.⁴ In 2023 it turned €27.6 billion of sales into €7.8 billion of net income.⁵ That is not the margin profile of a middleman. It is the margin of a company that owns the one thing in the system that cannot be sourced: the architecture itself - the systems engineering that decides how a light source from one captive partner, optics from another, and parts from five thousand more become a single machine that prints at the atomic scale. The components are bought. The integration is the product.

Isn't this just vertical integration with extra steps?

The fair objection: ASML owns Cymer outright and a chunk of Zeiss, so what's the difference from integrating - other than semantics? The difference is structural, and it is the whole point. True vertical integration would mean owning all of it - every supplier, every factory, every balance sheet - and carrying that capital weight forever. ASML owns almost none of it. It owns the two suppliers whose secrets are the bottleneck, and it stays an arm's-length customer to the thousands whose work is hard but substitutable. Co-investment buys roadmap control without buying the roadmap's cost. That is harder to replicate than integration, not easier: a rival can in principle build its own factories, but it cannot un-acquire Cymer or un-bind Zeiss. The honest caveat is that this design carries a risk a true integrator avoids - dependence. ASML's machine has single points of failure it does not fully own, and the relationships have to hold. They hold because ASML made itself the indispensable customer at the exact moment each partner needed one. Captured, not coerced.

ASML is praised for making the impossible machine. It would be closer to the truth to say it made the impossible supply chain - and then quietly closed it. It never needed to own the optics or the light source. It needed to be the only company those suppliers could afford to say yes to, and the only one anyone else could not. The genius was never in the manufacturing. It was in deciding which 20% to keep, which 80% to coordinate, and which two suppliers to make sure no competitor would ever get a turn with.