ASML CEO: No Rivals in Sight for EUV Lithography

Manufacturing the chips that run models like Claude Opus 4.7 or train the next generation requires machines that only one company in the world knows how to build. ASML, the Dutch EUV lithography firm, generated over 28 billion euros in revenue in 2025 and remains the most strategic bottleneck in the entire semiconductor supply chain. Its CEO, Christophe Fouquet, doesn't seem to lose sleep over it.

In an interview with TechCrunch published on May 5, Fouquet was remarkably relaxed when the conversation turned to potential competitors. The quote that gave the original article its title sums it all up: "No one is coming for us". It's not gratuitous bragging; it's a technical and economic reality that deserves careful analysis.

Why ASML has no real rival today

EUV lithography (Extreme Ultraviolet Lithography) is the technique that enables etching transistors at 3 nm and below onto silicon. Developing a functional EUV machine took ASML more than two decades and cumulative investment that analysts estimate in the tens of billions of euros. Canon and Nikon, the only historical competitors in conventional optical lithography, never completed the transition to commercial-scale EUV.

China has attempted to accelerate its own advanced lithography program through SMEE (Shanghai Micro Electronics Equipment), but export restrictions imposed from 2023 and renewed in 2025 have cut off access to critical components. Fouquet, without naming it directly, alluded to this context when noting that building the supplier ecosystem, extreme-precision optics and control software that surrounds an EUV machine isn't something that gets replicated in a single budget cycle.

What this has to do with AI

The connection to the AI ecosystem, and thus to the Claude universe, is direct. Every time Anthropic, Google or any other lab expands its inference infrastructure or trains a new model, it needs chips manufactured on advanced nodes: H100, B200, TSMC's future 2 nm accelerators. All pass through ASML machines.

Demand for AI compute has been one of the engines sustaining the semiconductor investment cycle even when the consumer market faltered. Fouquet acknowledged at Milken that orders tied to data centers and AI now represent a significant fraction of ASML's order book, though he didn't provide specific figures in the interview.

This has practical implications for anyone working with Claude's API or planning large-scale deployments: the availability and cost of advanced compute remain dependent on a supply chain with a single critical provider at its most delicate link.

The geopolitical risk Fouquet cannot ignore

The CEO's confidence has a logical limit: geopolitics. ASML's export licenses are subject to decisions by governments, primarily the Netherlands and the United States, and the regulatory environment has shifted several times in the past three years. Fouquet was prudent on this point, simply noting that the company operates within the current legal framework and that its business plans account for different scenarios.

That's not an evasive answer; it's the only honest one in a context where a decision in The Hague or Washington can alter which customers can buy which machines. TSMC, Samsung and Intel—ASML's three largest clients—operate in different jurisdictions with sometimes divergent interests.

Editor's Take

ASML's position is solid in the medium term, but "no one is coming for us" describes the present, not a structural guarantee. For those building on AI infrastructure, it's worth remembering that all that compute power remains anchored in a physical chain with very little redundancy.