Samsung Delays High-NA EUV Mass Production to 1nm Node, Expected Around 2030
Nashnova编辑部
Samsung has pushed High-NA EUV mass production from its original 2nm target to the 1nm node around 2030, with traditional EUV still running alongside it — the three leading foundries now chart divergent paths for a tool that costs $400 million per unit.
How far was the delay — and why?
Samsung originally planned to bring in High-NA EUV at 2nm and 1.4nm. The new target is A10 and beyond (1nm-class), with mass production expected around 2030.
Samsung researcher Park Chang-min disclosed the updated roadmap at the NGL 2026 conference on August 11, 2026. The reason: the underlying technical work is not yet finished.
This means → Samsung's High-NA timeline now trails Intel's by roughly four years and has slipped at least one full node from its own prior plan.
Why does "adoption" not mean "full replacement"?
Samsung stated explicitly: even at the 1nm node, traditional EUV (0.33 numerical aperture) multi-patterning remains the primary patterning method. High-NA enters only on select layers.
In plain terms = High-NA EUV widens the "lens" from 0.33 to 0.55, theoretically cutting the number of patterning passes needed to print finer features — but Samsung says the technology cannot yet carry the full workload.
The bottleneck goes beyond the scanner itself: each unit costs roughly $400 million (double a conventional EUV tool), and supporting technologies — masks, pellicles — are still maturing.
What paths are Intel and TSMC taking?
Intel has moved first: it used High-NA EUV on its Intel 18A process to ship parts of the "Panther Lake" Core Ultra Series 3 processor — the world's first high-volume logic product made with the technology.
TSMC chose a third path: keep extracting value from conventional EUV rather than switching to High-NA. Deputy co-COO Kevin Zhang called the ability to maintain an aggressive scaling roadmap on existing EUV tools a core R&D advantage.
TSMC has not publicly disclosed which node will first use High-NA.
Three divergent roadmaps — what does this signal for the market?
The three leading foundries are now split: Intel already in production, Samsung delayed to ~2030, TSMC holding steady. This means → High-NA EUV adoption is tracking well behind the equipment maker's expectations.
This reflects a deeper reality: when a single tool costs $400 million and the ecosystem around it is still catching up, "more advanced" does not automatically mean "more economical" — foundries vote with their capital budgets.
For equipment supplier ASML, the three-way divergence directly reshapes its High-NA order book and shipment timeline — demand certainty is falling.
Content is for reference only, not financial advice.