Latest / The Operations Podcast with Fexingo: Process, People, and Profit in Modern Business / How ASML Built the Chipmaking Operations That Power AI
Transcript
- Lucas: You know how every conversation about AI chips eventually lands on NVIDIA? The hardware, the software moat, the data-centre buildout. But there is a company that sits one level deeper in the stack, and without it, none of those chips exist. I'm talking about ASML. Luna: The Dutch lithography company. They make the machines that print the circuits. But I think most people — myself included — don't really grasp how insane their operations actually are. Lucas: Right. And that's the episode. Because ASML is not just a chip-equipment company. It is a masterclass in what you might call 'network orchestration' — a business model where you design and integrate a hyper-complex product, but you barely manufacture any of it yourself. Luna: So it's almost like the Apple model, but for billion-dollar machines? Lucas: In spirit, yes. But the numbers are on another planet. A single extreme ultraviolet lithography machine — the EUV machine, which is the only way to make the most advanced chips today — costs about $350 million. It weighs 180 tons. It takes a full year to build. And ASML ships only about 50 of them per year. Luna: Fifty machines a year, at $350 million each. That's roughly $17.5 billion in revenue from just one product line. And they have a monopoly on it. Lucas: Complete monopoly. Every single advanced chip — from an Apple A-series processor to an NVIDIA H100 — is made using an ASML EUV machine. There is no alternative. And how ASML built that monopoly is an operations story about a 30-year bet, a network of 800 suppliers across 40 countries, and a supply chain that is simultaneously incredibly resilient and terrifyingly fragile. Luna: Let's start with the fragility part, because that's what keeps me up at night. What is the single most critical thing in that supply chain? Lucas: Two things. First, the optics. The EUV machine uses mirrors — not lenses, because EUV light is absorbed by glass. Those mirrors are made by a single German company called Carl Zeiss. They are the only supplier in the world that can polish a mirror to the required precision. We're talking about surfaces so smooth that if you scaled them up to the size of Germany, the largest imperfection would be one millimetre. Luna: One company in Germany. Single point of failure. Lucas: Exactly. And then there is the light source. The EUV machine generates light by firing a carbon dioxide laser at tiny droplets of molten tin — 50,000 times per second. That laser system is built by a US company called Cymer, which ASML acquired in 2013, but the critical subcomponents come from a handful of specialised suppliers. Luna: So two single points of failure — the mirrors and the laser. And yet they ship 50 machines a year. How do you manage that level of concentration risk? Lucas: You manage it by treating your suppliers almost as extensions of your own R&D. ASML doesn't just place orders and wait. They embed their own engineers at Zeiss and Cymer. They co-invest in new fabrication capacity. They give multi-year volume commitments that let those suppliers invest in their own operations with confidence. In many cases, ASML actually owns the intellectual property for how a component is made, even though they never touch it. Luna: So it's less 'buyer-supplier' and more of a distributed manufacturing ecosystem. But that requires a level of trust and coordination that most companies can't pull off. Lucas: It does. And it required a specific leadership decision in the early 2000s. ASML had a choice: they could try to build everything in-house, like Canon or Nikon tried to do with their lithography systems, or they could partner deeply. They chose partnership. The result was that they leapfrogged everyone in EUV, but they also made themselves dependent on a network that cannot be replicated quickly. Luna: Which brings us to logistics. You don't just ship a 180-ton machine on a standard freight truck. How do they even move these things? Lucas: It's a whole operation in itself. The EUV machine is shipped in 40 sub-assemblies, each in a specialised container that costs about $100,000. They use chartered Antonov cargo planes — the world's largest — to fly the biggest pieces from Veldhoven in the Netherlands to the customer's facility. Then a team of 30 ASML engineers takes three to six months to reassemble and calibrate the machine on site. Luna: Three to six months of installation. That means the revenue recognition is stretched out, and any delay in the field hits the income statement. Lucas: Correct. And installation is not just technical — it's also environmental. The machine operates in a hard vacuum. It has to be isolated from any vibration. The cleanroom where it sits must be class 1, meaning fewer than one particle of dust per cubic foot. If a single speck lands on a mirror during installation, the whole mirror has to be removed and sent back to Zeiss for repolishing. That can take six months. Luna: Six months of downtime on a $350 million machine. The cost of a mistake is astronomical. Lucas: And that brings us to the risk management side of ASML's operations. They essentially run a real-time monitoring system for every machine in the field. Each EUV machine generates about 50 gigabytes of data per day — temperature, vibration, laser power, mirror alignment. ASML's central operations centre in Veldhoven can see anomalies before the customer does. They'll often pre-ship a replacement part before the machine even breaks. Luna: Predictive maintenance at scale. But is that enough to offset the concentration risk? Lucas: It mitigates operational risk, but not geopolitical risk. And that's the open question for ASML. The Dutch government, under US pressure, has already restricted exports of EUV machines to China. That's a market ASML can't serve. And if tensions escalate further, the supply chain itself could be weaponised. What if Germany restricts Zeiss exports? What if the US restricts Cymer? Luna: And ASML can't just dual-source those components. There is no second Zeiss. No second Cymer. So the monopoly cuts both ways. Lucas: Exactly. And that's where the operations model becomes a strategic vulnerability. But it's also what makes the company so hard to compete with. Anyone trying to build a rival EUV machine would need to replicate not just ASML's design, but the entire supplier network and the decades of embedded co-engineering. It would take at least a decade and tens of billions of dollars. Luna: Speaking of building things over decades — we spend a lot of time on this podcast breaking down how companies operate, and it's only possible because we don't have to interrupt the conversation for ads. If these deep dives have given you a new way to think about your own work or your own business, it's worth knowing that listener support is what keeps this show completely ad-free. You can help at buy me a coffee dot com slash fexingo. No pressure, just a simple way to keep the model going. Lucas: Yeah, and it genuinely makes a difference. Alright, back to ASML. One more layer I think is worth unpacking: how ASML actually manages its supplier network operationally. They use what they call a 'supplier matrix' — each key component is mapped with a risk score based on delivery time, quality, and geopolitical exposure. They meet with their top 50 suppliers every quarter to review capacity plans. Luna: That sounds like a lot of coordination overhead. How many people does ASML have in supply chain management? Lucas: About 2,500 people in their global supply chain organisation. For a company of roughly 40,000 employees, that's about 6 percent of the workforce. And they have a dedicated team that does nothing but map the sub-tier suppliers — the suppliers to their suppliers — to identify hidden single points of failure. Luna: So they're managing risk three levels deep. That's a level of operational sophistication that most companies don't even attempt. Lucas: But the flip side is that it creates massive barriers to entry. A new entrant couldn't just hire 2,500 supply chain experts — they'd also need the relationships, the trust, and the data-sharing infrastructure that ASML has built over 30 years. Luna: So the operations model itself is the moat. Lucas: Exactly. The technology is patented, but the operations network is the real competitive advantage. And it's not something you can copy by reading a case study. Luna: I want to go back to something you said earlier — ASML's choice to partner rather than build in-house. That feels like a decision that runs counter to how many tech companies think. Why did they make that bet? Lucas: Partly necessity. In the 1990s, ASML was a small player compared to Nikon and Canon. They didn't have the capital to develop everything in-house. So they partnered with Zeiss for optics, with Philips for certain components, with the Belgian research centre IMEC for process development. Over time, that network became a self-reinforcing ecosystem: suppliers invested in asml specific capacity because they saw the orders coming, and ASML invested in supplier R&D because they needed the innovation. Luna: Almost like a venture capital model for supply chains. They funded the development of their own suppliers. Lucas: Exactly. And that's a very different approach from a company like Intel, which historically built its own manufacturing and owned its own fabs. Intel's model was vertical integration. ASML's model is network orchestration. And in the capital equipment business, the orchestration model has proven more scalable, because you can tap into the best expertise across the world without having to build it yourself. Luna: But it also means you're only as strong as your weakest link. And when that link is a single factory in Germany, the whole system is exposed. Lucas: Right. And that's the tension at the heart of ASML. The same network that makes them dominant also makes them vulnerable. The question for the next decade is whether they can maintain that network in a world where trade barriers are rising and national security concerns are pulling supply chains apart. Luna: And that's where the story is still being written. ASML's operations have been a triumph of coordination and long-term thinking. But the geopolitical environment is changing faster than any supply chain can adapt. Lucas: It's the ultimate stress test for a network-orchestration model. If ASML can navigate the next five years while keeping its supply chain intact, it will be one of the great operations stories of the century. If not, we may look back at this moment as the peak of globalised high-tech manufacturing. Luna: Well, that's a sobering thought. But it also makes ASML the most interesting company to watch in the entire semiconductor ecosystem. Lucas: Absolutely. And that's what we'll be watching.