Dutch semiconductor-equipment giant ASML says artificial intelligence demand is now so strong that its planned manufacturing capacity will not be enough to satisfy customers through 2028, pushing chipmakers to discuss orders extending as far as 2030.
ASML reported record third-quarter 2026 net sales of €11.7 billion and net income of €5.0 billion, with a gross margin of 57.5%. The company expects full-year sales between €44 billion and €45 billion.
Reuters reported that customers including TSMC, Samsung and Intel are expanding semiconductor capacity as demand generated by AI models and data centers continues to rise.
Customers are already thinking about 2030
The strongest signal in ASML’s results was not its quarterly revenue.
CEO Christophe Fouquet said customers continue to raise their short- and medium-term forecasts because of their expectations for the size and duration of AI-related investments. Some customers are already expressing interest in placing equipment orders for 2029 and 2030.
ASML now expects that the supply it currently plans for both 2027 and 2028 will not fully meet customer demand. The company is studying ways to raise 2028 capacity beyond the previously announced 30% increase.
The Financial Times separately reported that ASML expects AI-driven demand for its chipmaking equipment to exceed supply through at least 2028.
The AI bottleneck extends upstream from GPUs
Much of the AI infrastructure discussion focuses on processors from Nvidia, AMD and other chip designers.
But those chips cannot exist without semiconductor manufacturing equipment.
ASML holds an especially important position because it is the sole supplier of extreme ultraviolet lithography, or EUV, machines used to manufacture the most advanced semiconductor nodes.
That means capacity constraints can propagate through the entire supply chain.
If foundries cannot obtain enough lithography tools, simply placing more orders for AI accelerators does not immediately create more chips.
The Financial Times reported that ASML is increasing production of both its existing EUV systems and next-generation High NA machines, while customers including TSMC, Samsung and Intel expand manufacturing plans.
AI infrastructure spending is becoming a multi-year industrial cycle
ASML’s numbers also suggest that chip manufacturers no longer view the AI buildout as a short-lived demand spike.
The company sold 113 new lithography systems during the third quarter, up from 86 in the previous quarter, while installed-base management sales reached €3.3 billion.
Reuters reported that ASML plans to increase output while upgrading machines already installed at customer facilities to create additional wafer capacity.
For founders and infrastructure investors, the important signal is how far upstream the AI boom has travelled.
Demand is no longer constrained only by access to the latest GPUs, data-center space or electricity. The companies building those GPUs increasingly need more manufacturing equipment simply to keep supply moving.
That turns AI infrastructure into a chain of interlocking bottlenecks.
The next limitation on compute might not be whether a cloud provider can buy enough accelerators. It could be whether the semiconductor industry can install enough of the machines needed to manufacture them in the first place.
