With its current 10,500 developers, ASML has a formidable engineering squad. In its first ten years, however, the company had to make do with one-seventieth of that number. The consequence: it had to focus on only one machine.
ASML is the most effective and, at the same time, the least efficient organization. It’s a character sketch that’s doing well in the high-tech corridors these days. ASML employees themselves and people around them are saying it with a mix of sarcasm and admiration. Sarcasm about the visible waste, admiration for the seemingly inexhaustible power of the R&D machine. It’s “a luxury that only ASML can afford,” suppliers usually add jealously.
Everyone understands that the development of highly complex systems requires a lot of people. Hundreds of specialists and thousands of designers. In addition, you need a lot of system thinkers, experienced people who can oversee the stuff, with a good eye for what customers really need.
To get the best out of such a team is a matter of balance. After all, development strength doesn’t grow linearly with manpower. On the contrary. At a certain point, adding more technicians is counterproductive. In high-tech, and also at ASML, many projects ran better when fewer people were assigned to them.
In the early days at ASML, there was a scarcity that’s now unimaginable. There was hardly any turnover, but the development of steppers required many technicians. This period spanned a bizarre ten years, the first four of which Philips and ASM invested 132 million guilders in their joint venture. After that, ASM was forced to drop out and Philips kept the litho club on a drip until things finally started to move in 1994.
ASML was constantly operating in crisis mode during that decade. It could only afford to develop one machine flavor. Customers with special needs were talked down, a task entrusted to current CTO Martin van den Brink. “We were always selling the machine we had,” he told me in 2015. “I didn’t have ten different development efforts to make ten different machines. It was one size fits all. It was hard enough to make one good machine. All kinds of special requests from customers we couldn’t fulfill. So I was always busy with the customer, with negotiations, with specs.”
That intensive consultation with chip manufacturers resulted in a machine that allowed ASML to meet the specifications of as many customers as possible. Van den Brink: “That was always the objective. We made the standard machine so good that we could always meet the specifications. And where we couldn’t, we just had to convince the customer our machine was still better.”
Looking at the bare figures, ASML grew from over a hundred employees in 1984 to about eight hundred in 1994. In the lean years from 1989 to 1991, the development team continued to increase steadily from 121 to 182, while revenue continued to decline, from 91 to 65 million guilders. The financial man at the time, Gerard Verdonschot, said repeatedly: stopping development wasn’t an option; that would have meant certain death.
It wasn’t until 1994 that ASML broke through a magical barrier. It started to make a profit and also more or less achieved a goal stated in its first year: one million guilders in sales per employee.
Comparing 1990 to 2020, the number of developers has increased from 150 to 10,500. The percentage of R&D employees has grown from 31 to 37 – the absolute number has multiplied by seventy. There will most certainly be some inefficiency, but those 10,000+ people are now working on over a dozen product lines that are 10-100x more expensive than the one machine was thirty years ago. We also find many developers in the companies that have since been acquired, such as Cymer and HMI. The latest acquisition, Berliner Glass, was a major expansion as well. It added another 1,600 employees, bringing the total to 28,000.