Taiwan's Nanya Tech/Winbond Adopt EUV, Challenging Korean Dominance in DRAM
nashnova research
Nanya Technology and Winbond Electronics plan to bring EUV lithography into DRAM mass production before 2030, with combined monthly capacity reaching 50,000 wafers — the first direct challenge to Korean makers' cost moat from the legacy DRAM lane.
What exactly are these two Taiwanese chipmakers doing?
Winbond has worked with ASML for over three and a half years and plans to adopt a 14 nm-class EUV DRAM process before 2030. It expects its first EUV tool within roughly two years.
Nanya plans to spend up to NT$346.6 billion (≈US$10.6 billion) between 2026 and 2029 to build out Phase 1 of its 5A fab, including EUV equipment. Wafer starts are set for H2 2027.
Nanya's capacity targets: 30,000 wafers/month by end-2028, 50,000 by end-2030, while advancing through its 1c, 1d, and 1e DRAM generations.
This means → Taiwanese firms are no longer defending low-end lines. They are betting real capital on EUV — lithography that uses extreme-ultraviolet light to pattern chips — pushing into Korean makers' technical heartland.
Why challenge now?
Samsung, SK hynix, and Micron have shifted capacity toward HBM — high-bandwidth memory built for AI servers — and server DRAM, tightening legacy DDR3/DDR4 supply.
That supply gap improved Taiwanese firms' pricing power; higher profits fund the EUV and advanced-node investments.
In plain terms = Korean makers chased AI memory's fat margins, leaving a gap in the "old battlefield." Taiwanese firms are using that windfall to upgrade their arsenal.
Should Korean makers worry in the short term?
Short-term threat is limited. Samsung and SK hynix have accumulated years of EUV mass-production experience across multiple DRAM nodes; Taiwanese firms are still in the equipment-installation phase.
ASML holds a near-monopoly on EUV lithography tools. Foundries and DRAM makers compete for the same machines, creating uncertainty over whether Taiwanese firms can secure tools on schedule.
Q2 2026 global DRAM revenue share: Samsung 39.4%, SK hynix 24.9%, Micron 23.3%; Nanya just 1.7%, Winbond 0.6%.
This means → The share gap exceeds 20×. Taiwan's challenge sounds more like distant footsteps than a close-quarters fight.
How could the landscape shift in the medium to long term?
Between 2028 and 2030, Taiwanese fabs come online and process nodes keep shrinking. New legacy-DRAM supply could trigger a fresh round of price competition.
If EUV adoption effectively raises die yield per wafer, Taiwanese firms' unit costs will fall meaningfully, sharpening their cost edge.
Korean makers' counter-strategy: keep shifting the product mix toward HBM and high-capacity server DRAM — high-value products that sustain a profit tier above the legacy fight.
This reflects a deeper signal: the DRAM industry is splitting into two lanes — "AI premium" and "legacy commodity" — each with its own home-court advantage.
Where is the real tipping point?
Whether Taiwanese firms can complete EUV production ramp-up before 2030 is the critical checkpoint for any real rewrite of the competitive order.
In plain terms = buying the tools is only step one. Running yields up to a profitable level is what counts — and Korean makers spent years getting there.
If Taiwanese firms hit their timeline, cost competition and share battles in legacy DRAM will heat up markedly. If they slip, the Korean moat holds.
市场有风险,内容仅供研究参考,不构成投资建议。
