ASML Posted a Record €11.7 Billion Quarter -- and Said It Still Can't Build Enough Chip Machines Through 2028

ASML reported record third-quarter revenue of 11.66 billion euros on Wednesday, beat its own prior guidance, and raised its full-year 2026 outlook to 44-45 billion euros -- and then told investors the good quarter doesn't change the underlying problem. Chief executive Christophe Fouquet said plainly in the earnings release: "our supply as currently planned for both 2027 and 2028 will not fully meet the demand of our customers." Some customers, he added, are now placing orders for 2029 and even 2030 -- delivery windows that are normally years away from being discussed, let alone booked.

The numbers behind the record quarter are straightforward. Net income reached 4.96 billion euros on a 57.5% gross margin, up from 54.0% on 9.33 billion euros of net sales the quarter before; ASML sold 113 new lithography systems and 8 used ones, closed the quarter with 12.74 billion euros in cash, and guided Q4 revenue to a range of 14.2 to 15.2 billion euros. It also repurchased 4.1 billion euros of its own shares in the quarter under its 2026-2028 buyback program and will pay an interim dividend of 1.88 euros per share on November 4. None of that is the actual news. A company beating guidance and raising its outlook is routine; a monopoly supplier telling its own customers it cannot build enough of its product for two years running is not.

The quarter-on-quarter jump is the plainest evidence that demand hasn't eased.

This is not a new complaint from ASML -- it is the same bottleneck getting visibly worse. In July, after its Q2 results, ASML raised its 2026 guidance for the first time to this 43-45 billion euro range and began pressing its largest customer, TSMC, for higher prices on both its most advanced EUV tools and its older DUV systems, on the logic that a sole supplier that is effectively sold out through 2027 has no reason to keep prices where they were. TSMC resisted the increase at the time; some Chinese customers agreed to pay it. Wednesday's results confirm that the "sold out" framing from July wasn't a negotiating position -- it was accurate, and it has since extended a full year further, into 2028.

ASML's own response has been to speed up, not just to expand. Chief financial officer Roger Dassen has described cutting the time from production start to shipping on its EUV machines from roughly 22 weeks a few quarters ago toward a target of 15 to 16 weeks, by trimming testing protocols while holding quality, shifting cleanroom space from research to production, and working with suppliers to clear component bottlenecks. That is a company trying to manufacture its way out of a shortage it cannot simply build new factories to fix in time -- EUV machines are custom-engineered, multi-story installations years in the making, not a product line that scales by adding a shift. The company expects to produce roughly 65 of its low-NA EUV systems this year, and has already committed to a 30% capacity increase on top of that for 2027 -- which, by that arithmetic alone, is still, per Fouquet's own account, short of what customers want.

It is worth being explicit about why one Dutch company's production schedule is a story about artificial intelligence at all. Every frontier AI model, from the largest labs to the smaller challengers, ultimately runs on chips fabricated at a small number of leading-edge foundries -- and every one of those foundries' most advanced nodes depends on EUV tools that only ASML makes. A GPU shortage usually reads, in the press, as Nvidia's problem or a hyperscaler's procurement headache. This earnings call is the reminder that the actual physical constraint sits one layer further upstream, at a single company whose 2026 output, record quarter included, still isn't enough -- which means the pace of the AI buildout through 2028 is, in a very literal sense, already decided by a production schedule ASML published this week, not by how much capital any AI lab or cloud provider is willing to spend.

“Our supply as currently planned for both 2027 and 2028 will not fully meet the demand of our customers.” — Christophe Fouquet, CEO of ASML, Q3 2026 earnings release

Fouquet's own hedge is worth reading precisely: ASML is "investigating ways to increase capacity for 2028 beyond the 30% increase we previously communicated," not committing to one. That is a company studying a bigger expansion, not announcing it -- the distinction between the two is exactly where a reader should watch next. Capacity expansion at this scale means new assembly lines for a machine with roughly 100,000 parts, built by a workforce ASML has to train as fast as it hires; it is not a decision that converts into shipped tools inside a single quarter, however strong this one was.