Latest / Business Strategy Talks with Fexingo: Competitive Advantage, Market Positioning, and Growth Planning / How ASML Built a Monopoly on Lithography Machines
Transcript
- Lucas: Every advanced chip in your phone, laptop, or car — from Apple's A18 to NVIDIA's Blackwell to AMD's latest — is made with a machine that only one company on earth builds. That company is ASML, based in Veldhoven, Netherlands, and their extreme ultraviolet lithography machines are arguably the most complex products ever manufactured. Luna: And when you say 'extreme ultraviolet' — we're talking about light that's so short wavelength it gets absorbed by air, right? The whole process has to happen in a vacuum. Lucas: Exactly. EUV light has a wavelength of 13.5 nanometres — that's about a dozen atoms wide. To generate it, ASML's machines fire a high-power CO2 laser at tiny droplets of molten tin, vaporising them into plasma, which then emits EUV light. That light has to be collected by mirrors coated with thousands of alternating layers of molybdenum and silicon — each layer just a few atoms thick — and then reflected through a series of optics made by Zeiss. Luna: And no one else can make those mirrors. Zeiss has a multi-year backlog just for ASML. Lucas: Right. That's the first layer of the moat. ASML didn't build this alone — they orchestrated a consortium that included Zeiss for optics, Cymer for the laser source, and eventually more than 5,000 suppliers. But they own the integration, the IP, and the decades of learning that no competitor can replicate quickly. Luna: So how did ASML end up as the only player? Were they always dominant? Lucas: Actually, no. In the 1990s, the lithography market had three major players: ASML, Nikon, and Canon. All three made deep ultraviolet — DUV — machines using 193-nanometre light. ASML was a distant third. But in the early 2000s, the industry realised that to keep shrinking transistors, they needed to move to a much shorter wavelength. That's when ASML made a bet that almost killed them. Luna: The EUV bet. Lucas: Yes. In 2012, ASML essentially put the entire company at risk. They told customers — TSMC, Intel, Samsung — that they would invest 4.5 billion Euros and take a decade to develop EUV. And they asked those customers to invest alongside them. TSMC bought a 5 percent stake. Intel bought 15 percent. Samsung bought 3 percent. That gave ASML the capital and the committed customer base to survive the R&D phase. Luna: So the moat started with a co-investment model. The customers literally were helping create the single supplier. Lucas: Exactly. And that alignment is itself a barrier. No chipmaker would fund a competitor to ASML now — they've already paid for this technology. By the time EUV was ready for mass production, around 2019, ASML had shipped roughly 60 machines. Today, each EUV machine costs about 200 million dollars. They have a backlog of over 100 machines. And the learning curve is brutal — each generation, like the new 'high na' EUV, costs more and takes longer. Luna: And then there's the supply chain lock-in. ASML doesn't just buy mirrors from Zeiss — they've invested in Zeiss's semiconductor optics division. They own part of the supply chain. Lucas: Right. They bought a 24 percent stake in Carl Zeiss SMT in 2016. They also bought Cymer, the laser source company, in 2013. So key components are either internal or tied up in exclusive agreements. Even if a competitor wanted to enter, they'd need to rebuild a global optics and laser supply chain from scratch — assuming they could even get the talent. Luna: And talent is another moat. ASML is in a small Dutch town. They have scientists and engineers who've worked on EUV for 20 years. That knowledge is tacit, not written down. Lucas: Absolutely. There's a famous story that when ASML was developing early EUV prototypes, they would literally have to bake the mirrors in a special oven to reduce thermal expansion. The recipe for that coating is one of the most guarded secrets in the industry. And even if you had the recipe, the deposition equipment itself is custom-built. Luna: So it's not just a technology moat — it's a system of interrelated moats: customer lock-in, supply chain control, proprietary manufacturing, and deep tacit knowledge. But there must be a vulnerability here. What about the geopolitical angle? Lucas: That's the big one. Because ASML's machines are critical for advanced chips, the U.S. government has pressured the Dutch government to restrict exports to China. Since 2019, ASML has been banned from selling EUV machines to Chinese customers. And more recently, even mid-range DUV machines are restricted. That limits ASML's addressable market — China was about 15 percent of revenue in 2023, and it's dropping. Luna: But China is trying to build its own lithography capability. Could that eventually erode ASML's moat? Lucas: It's possible in the long run, but the gap is enormous. China's best domestic lithography effort, from Shanghai Micro Electronics Equipment, is still at the 90-nanometre node — that's 2004-era technology. To get to EUV, they'd need to solve the same physics problems ASML spent 20 years and $9 billion on, while being cut off from Western suppliers and talent. It's not impossible, but it would take a decade or more, and even then, they'd likely be multiple generations behind. Luna: So ASML's moat is partly protected by geopolitics — the very restrictions that limit their market also protect them from new entrants. Lucas: Exactly. And the installed base is another huge advantage. ASML has shipped about 200 EUV machines as of mid-2026. Each machine requires regular maintenance, upgrades, and consumables — the tin droplets alone are a recurring revenue stream. TSMC, for instance, has over 70 EUV machines. The cost to switch to a different supplier — if one existed — would be astronomical, because all their process recipes are optimised for ASML's specific tool parameters. Luna: That's what makes it a true moat — the switching costs are effectively infinite. You can't just plug in a different machine and get the same yield. Lucas: Right. And this moat extends beyond EUV. ASML also dominates the DUV market for older nodes, with about 90 percent market share. They have a service and upgrade business that generates high-margin recurring revenue. In 2025, their installed base service revenue was over 6 billion Euros, with gross margins above 45 percent. Luna: If today's conversation gave you a clearer picture of why chip manufacturing is so concentrated — genuinely, that is the kind of insight these episodes aim for. And the reason we can keep doing deep dives like this without ads every eight minutes is listener support. If you found this useful, you can buy us a coffee at buy me a coffee dot com slash fexingo. It's a simple way to keep the show independent. Lucas: And it really does make a difference. Even a small contribution helps cover research, editing, and hosting. We appreciate everyone who's supported so far. Luna: Alright — back to ASML. Lucas, you mentioned the high na EUV machines. Is that the next chapter of the moat? Lucas: Yes. high na — numerical aperture — is essentially a bigger lens system that allows even finer patterning. ASML's first high na machine, the EXE:5000, shipped to Intel in early 2024. It costs around 400 million dollars. And they're already working on the next generation. The R&D cost for high na was about 3 billion Euros. Again, no competitor can match that spend without a guaranteed customer base. Luna: So ASML's moat is self-reinforcing. Each generation makes the barriers to entry higher, and the customers are locked into the upgrade cycle. Lucas: Exactly. But there are two potential threats worth watching. One is technological disruption — if some completely different lithography method, like nanoimprint or directed self-assembly, leapfrogs EUV. The other is geopolitical fragmentation — if export controls push China to create a parallel supply chain, or if the U.S. and Europe try to onshore chip manufacturing and end up nationalising some of ASML's technology. Luna: Nanoimprint — that's where you stamp the pattern like a stamp on a coin, right? Canon has been working on it for years, but it hasn't achieved the resolution or throughput for advanced nodes. Lucas: Correct. Canon's nanoimprint tool can do about 80 wafers per hour for some layers, but EUV does over 200. And the defect rate is higher. So disruption is still a distant possibility. But ASML's R&D spending — about 4 billion Euros annually — is a bet that they'll keep improving EUV fast enough to stay ahead. Luna: So the lesson for business strategy is: a deep-tech moat is built not just on one invention, but on a system of interdependent advantages — supply chain, talent, customer co-investment, and continuous reinvention. Lucas: Exactly. And it's a reminder that moats can also be fragile if they depend on a single technology or a geopolitical regime. For now, ASML's position looks unassailable. But in the long history of technology, nothing stays unassailable forever.