AI

Samsung and TSMC Commit to ASML's $400M High NA EUV Machines

The world's leading chipmakers will deploy next-generation lithography tools starting in 2028 to meet AI chip demands.

Omega Editorial· September 8, 2026· 3 min read

Samsung and TSMC have both committed to deploying ASML's High NA extreme ultraviolet lithography machines, marking a significant milestone for the next generation of chipmaking technology as AI applications drive demand for more advanced semiconductors.

The announcements position ASML's most sophisticated tools—which cost approximately $400 million each—as the industry standard for future chip production. Samsung plans to use the machines for DRAM memory production starting in 2028, while TSMC will deploy them for advanced logic chips, citing "increasingly complex transistor architectures required for AI applications."

Why it matters

The dual commitments from Samsung and TSMC remove a key uncertainty that has weighed on ASML's growth outlook. With the world's two largest chipmakers now on board alongside Intel, ASML gains visibility into demand for its most expensive and technically advanced product line. The move also signals that the semiconductor industry sees a clear path to continue shrinking chip features beyond current limits—critical for maintaining the performance improvements that AI workloads demand.

Technical advantages driving adoption

High NA EUV machines represent a leap forward in lithography precision. These tools can print smaller and more intricate circuit patterns onto silicon wafers compared to standard EUV systems. For Samsung, the technology will "extend the DRAM scaling roadmap" and improve manufacturing efficiency. TSMC expects usage to increase as transistor designs grow more complex to support AI chip requirements.

The technology arrives as chipmakers face physical limits in how small they can make features using existing methods. High NA EUV provides a path to continue advancing chip density and performance.

Market implications

Barclays analysts called the announcements "a positive" that "should provide more visibility on adoption which has been a key debate." The firm noted that ASML now faces a significant decision on whether to expand EUV production capacity beyond recently announced targets, given strong demand signals.

ASML shares have climbed 120% over the past year, with investors closely watching High NA adoption as a driver of future growth. The company plans to add approximately 30% capacity for EUV production overall in 2027 but has not provided recent forecasts for High NA machine sales specifically.

Intel became the first customer to use High NA EUV technology in production in July, according to ASML.

Industry collaboration on photomasks

Beyond the machine commitments, Samsung and TSMC will join ASML in an industry initiative to develop next-generation 12-inch photomask technology, upgrading from the current 6-inch format. Photomasks serve as the stencils that print circuit patterns onto wafers during manufacturing. ASML said the larger format will improve productivity and reduce chipmaking costs.

These details were first reported by CNBC.

#asml#euv lithography#tsmc#samsung#semiconductor manufacturing#ai chips

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in AI

AI· 3 min read

SEMIFIVE Ships 500mm² AI Inference Chip on Samsung 4nm Process

Korean ASIC designer completes mass production of HyperAccel's data center accelerator, marking expansion beyond security and HPC markets.

Via AI Watch · Sep 8, 2026
AI· 3 min read

Microsoft Project Zenith Brings 30B+ Parameter AI Models to PCs

New Windows 11 configuration targets developer hardware with unified memory and bandwidth to run large language models entirely offline.

Via AI Watch · Sep 8, 2026
AI· 3 min read

Arm Unveils AI-Native Chip Platforms for Mobile, Robotics, Cloud

New architectures target agentic AI workloads with specialized compute for edge devices, physical systems, and data centers.

Via AI Watch · Sep 8, 2026