The artificial intelligence investment cycle is beginning to appear in European factory plans rather than remaining confined to cloud budgets, after ASML reported €9.3 billion in second-quarter sales and outlined a substantial increase in lithography-system capacity.

ASML generated net income of €2.9 billion during the quarter and raised its full-year sales forecast to between €43 billion and €45 billion, with a gross margin expected to reach between 54% and 56%. Third-quarter revenue is forecast at between €11 billion and €12 billion.

The Veldhoven manufacturer sold 86 new lithography systems during the quarter, compared with 67 in the preceding period, while revenue from servicing and upgrading its installed equipment base rose to €2.76 billion. Customer commitments have given the company greater visibility over demand extending into 2027 and 2028.

ASML plans to increase 2027 production capacity for conventional low numerical aperture extreme-ultraviolet systems by 30% from a 2026 base of about 65 machines. A similar increase is planned for deep-ultraviolet immersion equipment, while another 30% addition in both categories is being considered for 2028.

AI spending reaches the production line

Lithography equipment occupies one of the least replaceable positions in the technology supply chain. Semiconductor manufacturers use ASML’s machines to project microscopic circuit patterns onto silicon wafers, while the most advanced manufacturing processes rely on its extreme-ultraviolet systems. No other supplier currently offers an equivalent commercial EUV platform.

ASML’s order intake therefore provides an industrial counterpart to the capital expenditure announced by cloud providers and chip designers. Model developers attract most of the attention, but greater computing capacity ultimately depends on processors, memory, networking silicon, fabrication plants, precision equipment, specialist materials, and the engineers who keep the entire system operating.

AI-related semiconductor investment has supported the market for several years, so the quarter does not reveal a sudden new trend. The additional capacity does, however, show that customers are making commitments far enough in advance to alter ASML’s manufacturing plans through 2027 and potentially into 2028.

Increasing final assembly will not remove every constraint, because a lithography system contains highly specialised optical, mechanical, laser, vacuum, and materials components sourced through an international supplier network. Production growth must therefore be matched by capacity among those suppliers, alongside the technicians needed to assemble, install, calibrate, and service machines that can cost hundreds of millions of euros.

Industrial leverage brings concentrated exposure

ASML gives Europe unusual strategic leverage in semiconductor manufacturing, although that position also concentrates risk within one company and its supplier network. Export controls, trade disputes, customer investment cycles, and delays at new chip plants can all affect demand, while ASML must manage relationships with customers operating across Asia, Europe, and the United States.

European semiconductor policy often describes sovereignty through the construction of domestic fabrication plants, yet fabrication is only one part of the production system. ASML demonstrates that control over enabling machinery can be equally consequential, because advanced chipmakers cannot move to newer processes without access to the equipment.

The results also offer a more grounded measure of infrastructure investment than the widening language around AI factories. ASML is receiving commitments, preparing additional production capacity, and spending against expected orders, which provides firmer evidence than a forecast whose physical requirements have yet to be contracted.

Those orders still depend on customers filling their planned fabrication capacity with products that can be sold profitably. Semiconductor demand has always been cyclical, and a rapid expansion in AI infrastructure could leave some equipment underused if model economics, enterprise adoption, or datacentre construction develops more slowly than anticipated.

Energy and resources add another constraint. More advanced chips can improve the amount of computation delivered per unit of electricity, but overall power, water, construction, and equipment demand can continue rising when total computing use grows faster than efficiency.

ASML must now convert exceptional demand into reliable delivery without weakening quality or margins, while its suppliers have to expand in step. Europe, meanwhile, gains high-value industrial activity from the AI buildout, but only for as long as its workforce, energy system, component suppliers, and trade relationships can support the production schedules now being set.