ASML’s $400M EUV lithography machine central to AI chip race and geopolitical chip controls

ASML, a Dutch company, is the sole supplier of extreme-ultraviolet (EUV) lithography machines, which are used to pattern the tiny circuitry on silicon wafers that become the chips powering AI accelerators. The newest generation of these machines, priced at $400 million each, can pattern features as small as eight nanometers—about the width of 40 silicon atoms—enabling denser, faster, and more energy-efficient chips.

The company’s high-NA EUV machines use lasers to vaporize molten tin droplets tens of thousands of times per second, generating extreme-ultraviolet light that is then focused through a series of precision mirrors to etch patterns on silicon wafers. This process is central to keeping pace with Moore’s Law, which has historically driven exponential gains in computing power.

ASML produces roughly 90% of all chip-lithography tools worldwide, making it an unavoidable supplier for any firm seeking to manufacture advanced semiconductors. The AI industry’s voracious demand for high-performance chips has intensified reliance on ASML’s technology, as companies like OpenAI and Anthropic build server farms to train and deploy increasingly large models.

Geopolitically, ASML’s dominance has made it a focal point of export controls. Since 2019, the U.S. government has pressured the Dutch government to impose an embargo preventing ASML from selling high-end lithography machines to Chinese firms, framing advanced chips as a strategic resource akin to oil. Analysts describe the situation as a choke point where ASML’s role resembles that of the Strait of Hormuz in global energy markets.

Competitive pressure is growing. China is investing heavily to replicate ASML’s technology, while startups such as Substrate aim to enter the lithography market with cheaper, smaller, and potentially more capable machines. Despite these efforts, ASML’s engineers and executives emphasize that the company’s near-term future remains unchallenged, with its latest machines promising to extend the AI hardware cycle for at least another decade.