There is one company in the world that makes the machines used to print the circuits inside every advanced chip. That company is ASML, based in the Netherlands, and its newest machine costs $400 million per unit. Chip factories are lining up to buy them anyway, because without access to tools like this, making better chips is simply not possible.
To understand why this matters, consider what the machine actually does. It shines a precise beam of light onto a silicon wafer, burning a pattern of circuits into it. The finer the light, the tinier the circuits, and the more of them you can fit onto a chip. More circuits per chip means more computing power, less energy used, and ultimately cheaper, faster AI. ASML's new machine prints features as small as 8 nanometers, roughly the width of 40 silicon atoms placed side by side.
The numbers behind ASML's business are striking. The company posted about $37 billion in revenue in 2025, up roughly 20% from the year before, driven almost entirely by demand from AI chip manufacturers. It sold 48 of its advanced machines in 2025 alone. The company expects revenue to rise further in 2026, with its own internal target of up to $44 to $60 billion in annual revenue by 2030.
Only three companies on earth can manufacture the most advanced chips at scale: TSMC in Taiwan, Samsung in South Korea, and Intel in the United States. All of them depend on ASML. Intel received the very first of the new machines and has already processed over 30,000 test wafers. Intel's bet is that being the first to use this technology gives it a head start in competing with TSMC as a contract chip manufacturer for AI companies.
TSMC, which makes chips for Apple, Nvidia, and most of the big AI labs, is taking a more cautious approach. It has confirmed it will not use ASML's newest machines for its production lines until at least 2028 to 2030, choosing instead to squeeze more performance out of its existing equipment. The logic is cost: TSMC's entire business model is built on making chips as efficiently as possible, and a $400 million machine that offers a 30% to 50% improvement is harder to justify when older machines still do the job.
The geopolitical dimension is where this story gets sharper. China has been completely cut off from ASML's advanced equipment since 2019, when US pressure led the Dutch government to ban the sales. China still buys ASML's older, less powerful machines, which accounted for 33% of ASML's total revenue in 2025. That share is expected to fall to around 20% in 2026 as new US legislation, called the MATCH Act, pushes to ban even those older machine sales and, critically, the servicing of machines already installed in China. Chip-printing equipment needs constant maintenance; without it, machines already sitting in Chinese factories will gradually stop working reliably.
China's response has been to pour billions into building its own version of the technology. A covert laboratory in Shenzhen, reportedly involving Huawei and thousands of engineers including some former ASML staff, completed a prototype in early 2025 that can generate the right type of light. But generating the light is the easy part. The machine has not produced a single usable chip. Independent analysts tracking the effort put a realistic date for even basic chip production at 2030, and true mass production much later. China's published 7nm chips were made using older machines stockpiled before the export ban took effect, not through any domestic breakthrough.
On the startup front, a San Francisco company called Substrate raised $100 million in late 2025 at a $1 billion valuation to build a rival system using X-ray light generated by a small particle accelerator. Its backers include the CIA's venture arm and Peter Thiel's Founders Fund. The company claims its approach could produce chips at a tenth of the current cost, and it wants to build its own chip factories rather than sell the machines. It has demonstrated promising early patterns in a lab setting, but has no product for sale and no production timeline before 2028 at the earliest. Industry observers are watching with interest but considerable skepticism, noting that going from a lab demonstration to a working chip factory is a decade-long journey on a good day.
For business operators, the practical reality is this: the chips that power the AI tools you use, or may use soon, depend on a supply chain concentrated in a handful of companies and countries. ASML's new machine helps ensure that supply of advanced chips keeps growing through the 2030s. The companies and governments that can access it will be able to build more powerful AI infrastructure. Those that cannot will be working with what they have. That gap is widening, not closing.