ASML's 2030 Order Book Flips the AI Capex Peak Narrative

The Market Got It Wrong — Again

ASML reported a third-quarter earnings beat on Wednesday and its stock immediately dipped. The numbers were fine. They were better than fine on the profit line. But the market wanted something else, and it didn’t get it — at least not in the first minutes of trading.

Sales of €13.05 billion narrowly missed the FactSet consensus of €13.14 billion. That was the headline-grabbing miss. Earnings per share, reported at the equivalent of $14.50, smashed the estimate of $12.17. Gross margin expanded to 57.5 percent from 54 percent in the prior quarter and 51.6 percent a year ago. Year-over-year, earnings surged 127 percent and sales climbed 49 percent. The fourth-quarter guide called for €14.7 billion in sales, well above the analyst model of €14.02 billion.

The stock fell 0.6 percent in premarket trading. It was also trading below its 50-day moving average.

What the market missed is the part of the earnings call that actually matters for the next decade of semiconductor investment.

Customers Are Booking Through 2030

Christophe Fouquet, ASML’s chief executive, did not mince words. Customers are increasing their short- and medium-term equipment demand forecasts, driven by their views on the size and duration of AI-related investments. Some are expressing interest in placing orders for 2029 and even 2030.

Let me emphasize the timeline. This is not a company talking about strong demand this year and next. This is a company saying its largest customers — the ones building the most advanced AI chips on Earth — are reserving production capacity three to five years out.

That is an extraordinary signal. In capital-intensive industries, order visibility that far out is rare. It means buyers have already run their own models, secured their own financing, and committed to a trajectory that assumes AI silicon demand will not flatten for years. It means the capex cycle that Wall Street has been debating — is AI chip spending cyclical or structural? — has already been answered inside these customers’ boardrooms.

Fouquet went further. ASML sold 113 lithography systems in Q3, up from 86 in Q2 and 66 a year earlier. The pace is accelerating. And the current supply plan for 2027 and 2028 will not fully meet customer demand. ASML is now investigating ways to increase capacity for 2028 beyond the 30 percent expansion it had previously communicated.

The Margin Story Behind the Beat

Bernstein analyst David Dai called Q3 revenue “broadly in line” but flagged margins as the key highlight. He is right to do so. A gross margin jump from 51.6 percent to 57.5 percent in a single year is not incremental improvement — it is a structural shift in the company’s cost position as it scales EUV and high-NA EUV production.

Higher margins at ASML mean two things. First, the company can reinvest more into capacity expansion without diluting returns. Second, it signals that the mix of products being sold is shifting toward the most expensive, most complex systems — the ones used for the most advanced AI chips. That is consistent with the order-book narrative. Customers are not just buying more tools. They are buying the best tools.

Dai reiterated his outperform rating with a price target of €2,859. The stock sits in a 15-week consolidation pattern with a buy point at $1,999.96 — its all-time high reached on June 30. It has since fallen below its 50-day line. The technical setup is ambiguous. The fundamental story is not.

What This Means for TSMC and Samsung

The implications for Taiwan and South Korea are direct and significant.

TSMC is ASML’s single largest customer by revenue. The company’s advanced-node expansion — its new fabs in Kumamoto, Japan, and the planned facility in Sakata — are predicated on precisely the kind of order visibility ASML is describing. If TSMC’s customers are booking capacity through 2030, then TSMC’s capital expenditure plan, which has run well above $40 billion annually in recent years, is not a speculative bet. It is a response to contracted demand.

Samsung is the second major player in this equation, and its position is more complicated. Samsung’s memory business has faced its own cyclical headwinds, but its foundry division has been struggling to win design wins at the most advanced nodes. The question for Samsung is whether ASML’s 2030 order book creates a spillover effect — whether TSMC’s capacity constraints and pricing power will push some AI chip demand toward Samsung’s alternative nodes, or whether the gap will simply widen.

The more likely path is the latter. ASML’s data suggests that the customers booking through 2030 are the ones who already have access to cutting-edge EUV capacity — which means TSMC, not Samsung. Samsung’s foundry revival, if it happens, will depend on a different set of bets: process improvements, customer relationships, and a willingness to accept lower margins in the short term.

The AI Infrastructure Timeline Shifts

Perhaps the most underappreciated implication of ASML’s guidance is what it means for the broader AI infrastructure buildout.

Every EUV tool shipped to a fab accelerates the timeline for new AI chip designs to reach volume production. If ASML cannot meet demand for 2027 and 2028 — and is now looking at 2028 specifically — then the bottleneck shifts from design to fabrication. The companies with the most advanced AI models will be constrained not by their algorithms but by their access to lithography capacity.

This is a bottleneck that has no quick fix. ASML is the world’s only manufacturer of EUV lithography equipment. There is no substitute. No competitor. No alternate supplier that can scale in a fraction of the time. The company’s decision to explore capacity increases beyond 30 percent for 2028 is a signal that even ASML’s own supply chain is under pressure — from suppliers of lenses, lasers, and precision mechanics.

For the AI industry, this means the buildout of training and inference infrastructure will continue to outpace the expansion of manufacturing capacity for at least the next two to three years. That favors companies with deep relationships with TSMC and ASML. It disadvantages everyone else.

The Bottom Line

ASML’s Q3 results were not a disappointment. The earnings beat, the margin expansion, the accelerating unit shipments, and the 2030-order visibility all point to a company at the center of a demand cycle that shows no sign of topping.

The market’s reaction was a reminder that even when the data is clear, the narrative can lag. The “AI capex peak” thesis has been circulating among some analysts and investors. ASML’s numbers suggest that thesis is wrong — or at least, premature.

Who wins from this? TSMC, naturally, and the AI companies that have secured advanced-node capacity with them. Who loses? Anyone betting that the chip-equipment demand curve is flattening. And who gets squeezed? Everyone waiting for afab capacity that is already booked through the end of the decade.