technology 6 min read

Taiwan's EUV Bet Reshapes the Memory Chip Race

Nanya and Winbond are reinvesting DRAM windfalls into EUV lithography, narrowing the technology gap with Korean rivals. The move could rewrite the memory market balance by 2030.

  • Semiconductors
  • Memory Chips
  • South Korea Tech
  • EUV Lithography
  • Taiwan Tech

The memory industry thought it had sorted out who does what.

Samsung and SK Hynix would chase high-margin HBM and server DRAM. Micron would fill gaps. Nanya and Winbond — together holding roughly four percent of global DRAM sales — would make do with older nodes and lower margins. It was a comfortable division, enforced not by policy but by economics: advanced memory required capital, know-how, and EUV equipment that Korea already had locked down.

That equilibrium is cracking. Not dramatically today, but decisively enough to matter by 2030.

Taiwan’s windfall is funding a technology leapfrog.

The catalyst is simple: AI-driven demand for HBM and high-capacity server DRAM has pushed Samsung, SK Hynix, and Micron to reallocate capacity toward those products. General-purpose DRAM — DDR3, DDR4, the workhorse chips used in industrial equipment, automotive systems, and networking gear — is getting starved. Prices are up. Margins are up. And Taiwan, which had been largely ceded that space, is suddenly sitting on a revenue boom.

Nanya’s third-quarter revenue jumped 61.9 percent quarter-over-quarter to NT$133.6 billion, even as unit shipments fell. Average selling prices rose more than 60 percent. Gross margin hit 85.7 percent.

Winbond was even more explosive: revenue surged 278.2 percent year-over-year to NT$82.3 billion.

These are not survivability numbers. They are expansion numbers.

Two very different EUV plays.

Nanya and Winbond are pursuing similar goals through different timelines. Both have publicly committed to bringing EUV lithography into their DRAM production lines, a category where neither has previously competed meaningfully against the Korean triplet.

Nanya’s board approved up to NT$346.6 billion in equipment investment for its new fifth-generation fab, Phase One, covering 2026 through 2029. EUV tooling is explicitly included. Wafer starts are targeted for late 2027, with capacity ramping to 30,000 wafers per month by end of 2028 and 50,000 by end of 2030. The company is also developing next-generation 1c, 1d, and 1e DRAM process nodes alongside the EUV transition. Separately, Nanya plans a 3D packaging and test facility at the Pingtung Science Park by late 2028, leveraging silicon-through-via and wafer-stacking techniques — technologies relevant to HBM but not equivalent to HBM production itself.

Winbond, meanwhile, has spent roughly three and a half years in preparation with ASML. It expects first EUV equipment delivery around two years from now and aims to introduce a 14-nanometer DRAM process by 2030 — though it acknowledges that timeline depends on smooth development and the actual mass-production date remains uncertain. Winbond plans a staged approach: develop the process using existing ArF immersion lithography first, then layer in EUV.

Both companies are playing the long game. Neither will outproduce Samsung or SK Hynix in advanced memory this decade. But both are systematically erasing the technological wall that has kept them in the legacy DRAM lane.

Why this matters beyond Taiwan.

The immediate implication is a potential reversal in cost dynamics. EUV lithography reduces the number of exposure steps needed to pattern certain circuit features compared to ArF immersion. Fewer process layers mean fewer opportunities for defects, higher throughput per wafer, and lower unit costs. If Nanya and Winbond can successfully migrate to EUV at 14 nm and beyond, they won’t just be making cheaper commodity DRAM — they’ll be producing it at margins that make further capacity expansion viable without the same reliance on price spikes.

That changes the calculus for the Korean giants. Samsung and SK Hynix have dominated advanced DRAM not because they are the only ones who can use EUV, but because they arrived first, accumulated yield experience, and built deep customer lock-in. The Taiwan players are now running the same race with similar equipment. The gap is narrowing from both ends.

The timing is the critical variable. Memory markets are cyclical by nature. The current supply squeeze benefiting Taiwan will ease as Korean and Micron capacity comes online over the next two to three years. By 2030, if Nanya and Winbond are running EUV-equipped fabs at scale, the industry could face a new wave of competition precisely when the big investors need the highest returns to justify their capex. Price cycles that favor producers over buyers could re-emerge — and Taiwan, with its lower operational costs and newer facilities, would be positioned to capture disproportionate upside.

ASML is the chokepoint no one can sidestep.

No analysis of Taiwan’s EUV ambition is complete without acknowledging ASML’s bottleneck. The Dutch company is the sole supplier of EUV lithography tools, and demand is already outstripping planned capacity. ASML has signaled that customer orders for 2027–2028 exceed what it can currently deliver. With both Korea and Taiwan pushing for equipment simultaneously, along with growing foundry demand, allocation decisions will carry enormous geopolitical weight.

This is where the US-Japan-Korea-Taiwan triangle becomes operationally consequential. Equipment access is no longer a commercial question alone. It sits at the intersection of export controls, alliance coordination, and industrial policy. Washington has quietly encouraged Tokyo and Seoul to prioritize their own memory firms for EUV allocations. Taiwan, outside the formal US security umbrella but deeply embedded in the semiconductor supply chain, faces a more complicated trajectory. Whether it secures timely EUV deliveries will determine whether its ambition translates into market share or merely expensive ambition.

Who wins, who loses, and what comes next.

For Samsung and SK Hynix, the short-term threat is minimal. They hold years of EUV production experience that Taiwan cannot replicate overnight. Yield optimization, process integration, and customer relationships in high-performance memory require time. The Korean firms also benefit from massive government support — South Korea’s CHIPS-style incentives and domestic supply-chain subsidies dwarf anything available to Nanya or Winbond.

But the medium-term picture is less comfortable. If Taiwan achieves competitive EUV-based DRAM production by 2030, the Korean firms lose their structural advantage in the memory segment they’ve increasingly depended on to subsidize their HBM bets. Every dollar Taiwan earns in advanced commodity DRAM is a dollar that doesn’t flow to the Korean ecosystem.

For Micron, the dynamic is more ambiguous. The American company benefits from tariffs and IRA incentives that disadvantage both Korean and Taiwanese competitors. But if global DRAM pricing softens due to expanded capacity — including new Taiwan supply — Micron’s volume-dependent business model faces headwinds regardless of where the capacity originates.

For Taiwan itself, the stakes are existential in a different sense. The island’s semiconductor sector is anchored by TSMC’s foundry dominance. Memory has always been secondary. But adding advanced DRAM capability diversifies the ecosystem, deepens the talent pool, and makes Taiwan strategically indispensable to multiple corners of the chip value chain simultaneously. That kind of redundancy is exactly what Washington wants to see — and exactly what Beijing fears.

The memory cycle turns roughly every three to five years. We are arguably midway through an upcycle driven by AI. What Taiwan is doing right now — converting temporary windfalls into permanent capability — means the next downcycle may not look like the last one. Instead of Korea dictating terms from the top, the market could fragment into overlapping tiers: Korea dominating HBM, Taiwan contesting advanced commodity DRAM, and everyone else scrambling for whatever margins remain.

That fragmentation favors no single country. It favors whoever builds the fastest.

Bottom line.

Taiwan’s EUV investment is not an immediate challenge to Samsung or SK Hynix. But it is a structural one. The financial foundation exists. The equipment path is defined. The only unknown is time — and in memory semiconductors, time is everything.