← Back

TSMC, Samsung Embrace High-NA EUV, Deepening AI Chip Race

Sep 8, 2026
TSMC, Samsung Embrace High-NA EUV, Deepening AI Chip Race

TSMC and Samsung have committed to integrating ASML’s next-generation High-NA extreme ultraviolet (EUV) lithography machines, a pivotal move solidifying the future roadmap for sub-2nm process nodes. This decision, driven by insatiable AI-driven demand for more powerful chips, ensures the foundry giants can meet the performance requirements of designers like Nvidia and Apple. As Intel Foundry has already purchased and is currently installing its own High-NA EUV tool, this commitment from its primary rivals officially transforms the next generation of process technology into a three-way race, fundamentally reshaping the strategic landscape for semiconductor leadership. The adoption of High-NA EUV, priced at over $350 million per machine, creates a formidable capital barrier to entry, further consolidating the market among the top three players. This technology enables the etching of even finer circuit patterns, directly translating to more transistors per wafer and thus higher performance and energy efficiency for AI accelerators. The primary winners are the foundries themselves, securing their roadmaps, and hyperscalers like Amazon and Microsoft, who will gain access to more powerful custom silicon. The loser is Moore’s Law’s affordability; the soaring cost ensures only the most profitable chip designs can justify leading-edge production. Looking forward, the critical variable is not just adoption, but operational mastery. The first foundry to achieve high-yield, high-volume production with High-NA EUV will gain a significant, multi-year competitive advantage. Watch for initial yield reports in the next 18-24 months, as these will be the key indicators of leadership. This trajectory suggests that by 2027, the performance gap between chips made on High-NA EUV and older processes will become a chasm, forcing a strategic recalculation for any hardware company not aligned with TSMC, Samsung, or Intel.