ASML Slows AI Growth with Its EUV Machines

⚡

Key Takeaways

1ASML, the leader in EUV lithography, is crucial for the production of advanced semiconductors, essential for AI.

2Demand exceeds supply: ASML plans to deliver 60 EUV systems in 2026, but this remains insufficient.

3Production capacities are already sold out until 2026, increasing pressure on the industry.

💡Why it matters — Dependence on ASML limits the rapid expansion of AI, impacting major players in the tech sector.

ASML: An Essential Player for AI

Artificial intelligence, often seen as a rapidly expanding force, relies on a physical infrastructure of which ASML is a crucial pillar. The recent results from the Dutch company highlight a reality often overlooked: the growth of AI is limited by hardware constraints.

ASML is at the heart of the semiconductor production chain, essential for the data centers and GPUs that power AI. However, this chain is currently under pressure, unable to keep up with the growing demand.

Demand Outstripping Supply

ASML's financial results for the last quarter are excellent, but they reveal tension in the market. Christophe Fouquet, CEO of ASML, and Roger Dassen, CFO, have warned that supply will not be able to meet demand in the near future. ASML's customers have already sold their capacities through 2026, and constraints are expected to persist beyond that.

This situation marks a shift in the semiconductor industry, traditionally subject to cycles of growth and recession. Now, demand sustainably exceeds supply, affecting all levels of the chain, from advanced components to memory.

The EUV Lithography Bottleneck

While attention often focuses on chip and GPU manufacturers, the real bottleneck lies upstream, at ASML. The company is the only one capable of producing EUV lithography machines, essential for manufacturing advanced semiconductors. Without this equipment, it is impossible to increase production capacity.

ASML plans to deliver 60 EUV systems in 2026 and 80 in 2027. Although these numbers are on the rise, they are insufficient to meet current demand.

Production Limited by Physical Constraints

Unlike software, semiconductors cannot be mass-produced instantly. Each EUV machine is the result of thousands of components and years of development. Accelerating their production could compromise their reliability. Thus, the expansion of production capacity is slow and dependent on a global supply chain.

ASML does not seek to slow down AI, but its central position makes it an inevitable bottleneck.

Towards Increased Production

In the face of these constraints, the industry is adapting. Rather than simply increasing volumes, it seeks to optimize each unit produced. Christophe Fouquet speaks of "higher lithographic intensity," meaning increased reliance on lithography technologies for each chip.

This strategy reinforces dependence on ASML, as the more complex the chips, the more manufacturing steps they require, thereby increasing demand for equipment.

Growth Already Planned

ASML's future capacities are already largely allocated. Customers, such as foundries and memory manufacturers, are securing their equipment over several years through contractual commitments. The growth of AI is thus not only constrained but also already organized.

This means that access to computing resources no longer solely depends on innovation or demand, but on the ability to invest upstream. Only players capable of bearing these constraints, such as hyperscalers, large industrials, and states, can fully benefit from the available resources.