High-NA EUV is no longer a slide in a roadmap deck. The first one is installed and running at a customer site, and at roughly $ 380 million per copy, it is the most expensive tool the chip industry has ever bought. ASML shipped the inaugural High-NA system to Intel over the past few months, a machine so large that it arrived in hundreds of crates and took months to assemble inside the fab. In the same year, the Dutch government finally did what I kept saying it would and restricted immersion DUV to China. The most advanced machine ever built and the tightest controls ever written landed on ASML in the same twelve months.
The Netherlands has brought in its own licensing rules covering the advanced immersion tools, the NXT:2000i class and up, and as of this year those machines need export licenses The Hague will no longer grant for China. The EUV-versus-DUV line that held for years is simply gone. Some sales already in progress have been pulled. I wrote two years ago that this was coming and that it was only a question of timing, and the timing turned out to be about eighteen months. China spent that window buying every DUV tool it could, which is why ASML’s China revenue has actually been climbing into the restrictions, a last surge of orders booked before the door swung shut.
The EXE:5000 going into Intel lifts the numerical aperture of the optics from the 0.33 of standard EUV to 0.55, sharpening the smallest features it can print and pushing the resolution limit down another step. That sounds incremental. It is not. Reaching 0.55 forced a near-total redesign of the optical system, with mirrors so large and so precise that Zeiss spent years just building the capability to make them. The result is a tool the size of a bus, priced around 380 million dollars before installation, that only ASML can build and only Zeiss can supply the optics for. The supply chain for the most advanced machine on earth runs through two European companies and nobody else.
To keep the optics buildable at 0.55, ASML went with an anamorphic design that magnifies the image differently in the two directions, which halves the printable field on each exposure. A smaller field means a High-NA tool images less area per pass than a standard EUV tool, so raw throughput per exposure actually drops even as resolution improves. Whether it lowers cost per wafer comes down to whether one High-NA pass beats two standard EUV passes, and that math is not settled. The early customers are buying these machines partly to find out. I would not take it as given that High-NA pays off quickly. Spending 380 million on a tool to answer a question nobody has answered yet is a strange kind of confidence.
Peter Wennink, who led ASML through the entire EUV ramp and the opening of the export-control era, stepped down this year, and Christophe Fouquet took over as chief executive. Fouquet came up through the EUV side of the business, which tells you where the company sees its future. He inherits ASML at the peak of its powers and in the most exposed political position it has ever held. The roadmap is the part of his job he can actually engineer. The rest of it is keeping customers across rival blocs all convinced ASML is still on their side, while three governments treat the company as a strategic asset they each want to steer.
Every time ASML builds a more capable machine it climbs further beyond any possible competitor, and every step beyond raises the political price of owning something that essential. That pattern has run for almost twenty years, and it is only speeding up. High-NA is the most extreme machine yet. The DUV restrictions are the widest controls yet. They landed in the same year because they are the same story told from two ends. Fouquet’s real inheritance is the question no one running a lithography company should ever face: what do you do when your product becomes too important for your own customers to be trusted with it? I do not think he knows yet, and I am not sure anyone does.