Top chipmakers embrace ASML's $400M machines, agree to crucial chipmaking change
Samsung and TSMC have announced plans to adopt ASML's high-NA EUV lithography machines, signaling the semiconductor industry's broader embrace of this advanced chipmaking technology Intel, Samsung, TSMC, and ASML are collaborating on a critical photomask standard change from 6-inch to 12-inch masks, potentially boosting high-NA EUV chip production productivity by 40 percent High-NA EUV machines, costing up to $400 million each, use larger mirrors and improved optical systems to create smaller, m
Analysis
TL;DR
- Samsung and TSMC have announced plans to adopt ASML's high-NA EUV lithography machines, signaling the semiconductor industry's broader embrace of this advanced chipmaking technology
- Intel, Samsung, TSMC, and ASML are collaborating on a critical photomask standard change from 6-inch to 12-inch masks, potentially boosting high-NA EUV chip production productivity by 40 percent
- High-NA EUV machines, costing up to $400 million each, use larger mirrors and improved optical systems to create smaller, more precise circuit features on silicon wafers
- Samsung and SK Hynix plan to begin producing memory chips with high-NA EUV by 2028, while TSMC targets 2030 for its most advanced logic chips
- China faces significant barriers to accessing this technology due to export restrictions, with experts estimating it could take 15 years to develop a domestic EUV lithography capability
Why It Matters
The adoption of high-NA EUV lithography by the world's largest chipmakers represents a pivotal inflection point in semiconductor manufacturing, directly impacting the performance and efficiency of chips powering AI data centers and consumer electronics. The collaborative shift to 12-inch photomasks demonstrates how industry-wide standard changes can unlock substantial productivity gains, which is critical given the ongoing global memory chip shortage driving up prices across consumer devices.
Technical Details
- High-NA EUV (high-numerical-aperture extreme ultraviolet) lithography machines use bigger mirrors and improved optical systems to more precisely collect and focus light, enabling smaller feature sizes on silicon wafers compared to older deep ultraviolet lithography
- The technology involves an extremely complex laser system that zaps molten tin into plasma at temperatures around 200,000°C, with mirrors collecting and redirecting the light into lithography tools for projecting circuitry patterns
- The proposed photomask standard change from 6-inch to 12-inch masks could improve high-NA EUV chipmaking productivity by approximately 40 percent, as confirmed by ASML CTO Marco Pieters
- ASML's EUV machines cost up to $400 million each and are the only source of this technology globally, with Zeiss Group serving as the exclusive supplier of mirrors and optical systems
- Intel has already begun high-volume manufacturing using high-NA EUV for Panther Lake mobile processors starting July 2026, while Samsung and SK Hynix target 2028 for memory chips and TSMC targets 2030 for advanced logic chips
Industry Insight
The consolidation of major chipmakers around a single lithography provider (ASML) and a unified photomask standard reinforces the extreme concentration of power in the semiconductor equipment supply chain, making ASML an indispensable strategic asset with significant geopolitical leverage. The 40 percent productivity gain from the 12-inch photomask transition could meaningfully alleviate the global memory chip shortage that is inflating consumer electronics prices, but the multi-year timelines (2028-2030) for widespread adoption mean supply constraints will persist in the near term. China's exclusion from high-NA EUV technology and the 15-year timeline estimated for domestic development underscores the deepening technological gap and suggests that US-Dutch export controls are effectively slowing China's semiconductor advancement, though it may also accelerate China's push for self-sufficiency in older-generation lithography.
Disclaimer: The above content is generated by AI and is for reference only.