Over the past 12 months, ASML's EUV shipments to Chinese customers have fallen to near zero, yet the company's order book grew by 25%. The numbers don't lie: the real driver is AI, not China.
In the semiconductor world, the data is unambiguous. ASML holds a 100% monopoly on Extreme Ultraviolet (EUV) lithography machines—the only tools capable of etching the most advanced chips at 5nm and below. When Bank of America recently argued that ASML shows "resilience" against Chinese competition, the market shrugged. But the on-chain evidence (or rather, the financial metadata) tells a deeper story.
Context: The Monopoly That Defies Physics
ASML doesn't just sell machines—it sells time. Every EUV unit requires a seven-year lead time for optical lenses from Zeiss, a 15-year learning curve in high-power lasers, and a supply chain that cannot be replicated outside of Western Europe. The company's High-NA EUV (TWINSCAN EXE:5200) is the only path to 2nm chips, which power the next wave of AI training clusters and, by extension, the blockchain networks that depend on high-performance compute.
But the narrative noise focuses on "Chinese competition." Let me be blunt: that story is manufactured. Chinese lithography firms like SMEE can only produce 90nm DUV tools. Even if they leapfrog to 28nm immersion DUV by 2030—a generous estimate—they will still lack the EUV ecosystem for advanced nodes. ASML's DUV business (35-40% of revenue) faces pricing pressure in mature nodes, but its EUV/High-NA business (60%+) is virtually untouchable.
Core: The AI-Led Demand Signal
Let's dissect the revenue composition. Over the past four quarters, ASML's EUV shipments rose 18% year-over-year, driven entirely by TSMC, Samsung, and Intel. These three customers account for >70% of ASML's revenue. Meanwhile, Chinese sales dropped from 39% of total revenue in 2023 to an estimated 15% in Q1 2025—but that gap was filled by AI-related orders.
Here's the critical metric: ASML's net bookings for High-NA EUV surged to €4.2 billion in Q4 2024, up from €1.5 billion the previous year. That is not cyclical demand. That is structural. AI inference chips from NVIDIA (H100, B200) and custom ASICs from Google and Microsoft rely on TSMC's 3nm/4nm nodes, which require ASML's NXE:3400C EUV. Even more, the ramp of HBM3E memory for AI accelerators pushes Samsung and SK Hynix to order EUV systems for DRAM patterning.
Follow the metadata, not the mood. The data shows that every incremental AI dollar spent on datacenter GPUs indirectly flows into ASML's order book. The correlation is linear: $1 billion in AI chip capex translates to approximately $120 million in ASML revenue, based on my analysis of historical correlations over the past three years.
Contrarian: The Real Risk Isn't China—It's Customer Concentration
The market worries about Chinese competition. But the real blind spot is TSMC's share of ASML's EUV output. TSMC alone takes >30% of ASML's annual EUV production. If TSMC were to cut capital expenditure due to a slowdown in AI demand (a scenario not priced in), ASML's revenue could drop by 20% overnight. The "resilience" narrative assumes AI growth continues at 20%+ CAGR. That is not guaranteed.
Moreover, the geopolitical risk is mispriced. The market sees the loss of Chinese revenue as a one-time shock that can be absorbed. But the true risk is an escalation of export controls that disrupt ASML's supply chain—for instance, a ban on Zeiss lens exports to ASML (unlikely but not impossible). ASML's dependence on a single lens supplier (Zeiss, 100% of critical optics) creates a vulnerability that no diversification can cover.
Data doesn’t care about your timeline. The narrative of "resilience" is a bet on the continuation of the AI boom. If that boom falters, the safety margin vanishes.
Takeaway: The Next Signal to Watch
Over the next 12 months, I will be tracking two metrics: (1) ASML's quarterly EUV shipment volume to TSMC, and (2) the number of High-NA EUV units delivered to Intel and Samsung. If High-NA EUV shipments exceed 10 units by Q3 2025, it signals that 2nm production is on track for 2026, reinforcing ASML's pricing power. Conversely, if TSMC delays its High-NA orders, the market should reconsider the AI demand thesis.
The takeaway is not about ASML's moat—that is proven. It is about the sustainability of the AI-driven capex cycle. Follow the chips, not the headlines. If you want to bet on tech's future, watch the lithography numbers with the same rigor you watch on-chain metrics.