Every AI query you fire off traces back to a single Dutch company that prints the patterns on the world’s most advanced chips. ASML CEO Christophe Fouquet sat down with TechCrunch AI in Beverly Hills this week to address the question hanging over a $530 billion monopoly: can anyone actually break it? His answer, as TechCrunch AI reports, was a calm and pointed no.
The context matters. Microsoft, Meta, Amazon, and Google are committing more than $600 billion to AI infrastructure this year alone. ASML’s extreme ultraviolet lithography machines, which sell for $200 million to $400 million each and take months to assemble, are the only tools on the planet capable of producing the leading-edge silicon those buildouts depend on. Demand has outrun supply so badly that Fouquet expects hyperscalers to be chip-starved for the next two to five years.
The challengers, and why Fouquet shrugs
Three threats have surfaced around ASML’s moat. Fouquet addressed each one directly with TechCrunch AI.
- Substrate, a Peter Thiel-backed San Francisco startup, has raised over $100 million at a unicorn valuation on the promise of building a rival machine. Fouquet’s read: “Wanting to have it and having it, that’s still a huge difference.” He pointed out that ASML only pulled off EUV because 80% of the system was already built from prior generations. Even then, generating EUV light alone took 20 years.
- xLight, a U.S. government-backed laser startup, is targeting just the light source inside the machine. ASML is cooperating with them, but Fouquet says the jury is out on whether it beats ASML’s own roadmap on cost or performance.
- Chinese reverse-engineering reports: Fouquet noted the obvious wall, that you need an actual EUV machine to reverse-engineer one, and there are none in China.
What stands out here is how unbothered Fouquet sounds. He’s not posturing. He’s describing a technical moat measured in decades, not quarters.
Why this matters for the AI build cycle
ASML is the chokepoint nobody outside the semiconductor world talks about, but every AI roadmap quietly assumes it. A few realities follow from that.
- Chip scarcity is structural, not cyclical. When the supplier of the supplier says shortages last five years, capacity planning at every layer above gets harder. Compute prices have a floor under them.
- High-NA EUV is the next pricing fight. TSMC publicly called ASML’s newest $350 million-plus machines too expensive. Fouquet’s counter is that cost-per-wafer drops 20 to 30% on advanced layers, and he’s betting the same arc that played out with low-NA EUV repeats. History favors him.
- Geopolitics is now a chip-equipment story. Export controls on EUV to China, U.S. funding for xLight, Thiel money chasing Substrate. Lithography has become national-security infrastructure, and ASML sits at the center of it.
Practical takeaways
For AI practitioners and businesses planning compute-heavy roadmaps:
- Don’t assume GPU supply normalizes soon. The bottleneck isn’t Nvidia’s design cycle, it’s wafer capacity gated by ASML throughput.
- Build long contracts now. Hyperscalers are locking in multi-year capacity deals. Smaller players who wait will pay the spot premium.
- Watch alternative architectures. When the leading-edge node is permanently supply-constrained, efficient model design, smaller specialist models, and inference-time optimization stop being academic and start being margin.
- Track the challengers, but discount the timelines. Fouquet’s track record on lithography is unmatched. Substrate and xLight are real, but neither produces a manufacturing-grade alternative inside this AI build cycle.
The broader signal from Fouquet’s interview: the AI infrastructure boom is bumping into a 30-year-old physics problem owned by one company. That dynamic shapes pricing, geography, and competitive positioning across the entire stack. Full interview is at the original source.