The machine that prints the world’s most advanced chips is nearing the edge of what its optics can do, and the company that builds it has now put a number on what comes next. According to a peer-reviewed paper reported by Reuters on October 8, the successor to ASML’s High NA EUV lithography system, a technology known as Hyper NA, will need roughly a decade before it is ready.
The paper was written by engineers at ASML and Zeiss, the German optics firm that supplies the mirrors inside the machines. It lays out what Hyper NA could deliver: patterns as small as 5 nanometers, more than a third finer than what the current High NA systems can print.
The significance sits in the physics. For years, the chip industry has pushed the limits of how small it can draw a circuit, and EUV lithography, which uses extreme ultraviolet light, is the tool that made the latest rounds of shrinking possible. High NA, the current generation, is only beginning to be adopted, and the paper suggests its successor is still far enough away that chipmakers should plan around a long gap.
One striking finding in the paper is that Zeiss can already make mirrors with the precision Hyper NA would require. The machine itself, the engineers wrote, would be only slightly larger than today’s High NA systems, and the existing light source could be reused as-is. That narrows the list of hard problems to solve, even if the timeline remains long.
ASML has begun research on Hyper NA but has not committed to production. The company has a history of spacing out its lithography generations, and each new system has been more expensive and more complex than the last. High NA machines cost about $400 million apiece, and the customers that can afford them are few.
Those customers have already begun buying in. Intel, Samsung, SK Hynix and TSMC have all started adopting High NA systems, betting that the machines will be needed to keep their roadmaps moving over the coming years. The paper’s ten-year horizon for Hyper NA tells them how long they will be living with the current generation.
The timeline also matters for the broader question of how long Moore’s Law can be stretched. Chipmakers have turned to creative workarounds as raw shrinking has slowed, stacking chips, chiplets and new materials to keep performance climbing. EUV lithography remains the engine for the densest logic, and a decade-long wait for the next optical leap pushes more of that burden onto packaging and design.
Zeiss’s role is easy to overlook but hard to overstate. The company makes the mirrors that focus the ultraviolet light, and those mirrors are among the most precise objects ever manufactured. The paper’s claim that the mirrors for Hyper NA are already achievable suggests the gating factor is not the optics but everything else, the machine’s scale, its engineering and the decision to fund it.
The economics are part of the decision. Each lithography generation costs more to develop, and the customers who can pay for it grow fewer. ASML has said in the past that it will proceed when the technology and the demand justify it. A ten-year horizon gives it time to judge both.
For the chip industry, the paper is less a promise than a boundary. It tells engineers and executives what the next optical generation might allow and how long they will have to wait for it. Until then, the work of pushing chips forward will fall to the tools already installed, and to the companies patient enough to plan a decade out.
The reaction among analysts has been measured. The paper confirms what many in the industry suspected, that the next step in EUV is real but distant, and that the current generation of tools will carry the load through the late 2020s and beyond. It also reinforces ASML’s position: the company remains the only maker of the machines at this frontier, and its roadmap, however slow, still sets the pace for everyone else.
The wait is not idle. ASML and Zeiss will spend the intervening years on the engineering that turns a research result into a tool that can be sold and serviced at volume. The paper is the first public marker on that path, a statement that Hyper NA is worth pursuing and that the industry should begin to think about what it will make possible.
The paper does not settle the debate over whether Hyper NA will ever be built, and neither company has made that commitment. What it offers is a clearer picture of the distance to the next generation. It also shows how much of the industry’s roadmap, from smartphones to AI accelerators, still passes through a small cluster of companies in the Netherlands and Germany that few consumers have ever heard of.
