Nanya's $6.2B Answer to the Memory Squeeze Arrives in 2028

Nanya Technology plans to spend more than NT$200 billion, about $6.2 billion, on capital projects in 2027, roughly four times its 2026 budget of more than NT$50 billion, according to a Cryptobriefing report that credits New Taipei City's investment portal ^[1][13]. That is a real supply-side response to the memory demand surge, and by the company's own schedule it produces no wafers until 2028 ^[8].

The vehicle is a 12-inch DRAM fab in the Taishan District of New Taipei City, where ground was broken in 2022 and the full buildout is costed at NT$480 billion, about $15 billion ^[2]. Phase one targets 30,000 wafers per month by 2028 on 10nm-class process for next-generation DDR, with the site eventually scaling to 45,000 ^[3]. Six years separate groundbreaking from that first capacity milestone ^[18], and the milestone itself is two thirds of the site's eventual output ^[19]. The 2027 capex line alone is roughly 42 percent of the entire Taishan budget ^[21], which is the signature of a tooling year rather than an output year. Separately, Nanya has committed NT$346.6 billion, about $10.7 billion, to EUV-equipped DRAM production across 2026 through 2029 ^[4], an average near NT$86.7 billion annually ^[17].

The funding case sits in one quarter. Nanya reported Q2 2026 revenue of NT$82.55 billion, up 684 percent year over year ^[5], net income of NT$50.19 billion, up 1,324 percent ^[6], and gross margin of 79.5 percent ^[7]. That works out to a net margin around 61 percent ^[14]. Annualise the quarter and the company earns about NT$200.8 billion ^[16], which is the 2027 capex plan almost to the dollar. Nanya is proposing to pay for its catch-up out of the shortage itself, at prices the shortage created.

Two things undercut the optics. The comparisons are flattered by a base: DRAM prices cratered in 2023 and early 2024 as post-pandemic demand cooled and inventories swelled ^[9]. And Cryptobriefing's own framing is that a 79.5 percent gross margin reflects peak-cycle conditions, with prices exposed if AI spending decelerates or if every major producer lands new capacity at once ^[12].

The strategic logic is less about this cycle than the next one. Nanya has historically been the smallest of the major DRAM producers, behind Samsung, SK Hynix and Micron on both capacity and technology, and the EUV spend is its attempt to close that gap ^[10]. Meanwhile the demand it is chasing, high-bandwidth memory for the data centres Microsoft, Google and Amazon are building, is being contested now with tools already installed ^[11]. Capital committed in 2027 does not arbitrate 2026 contract prices. It arbitrates 2029.

Watch whether the NT$50 billion 2026 budget gets revised upward during the year, which would signal Nanya trying to pull capacity forward rather than waiting on Taishan ^[1]. Watch the 2028 date on the 30,000-wafer target for slippage, because a year of delay lands the capacity into whatever the cycle looks like then, not now ^[3]. And watch how much of the NT$346.6 billion EUV commitment is actually drawn in 2026 and 2027, since that four-year number is a plan, not a purchase order ^[4].