Nanya Technology spent years as the also-ran of the memory business, the Taiwanese DRAM maker that Samsung, SK Hynix and Micron never had to worry about much. The AI boom has changed the arithmetic. The company now plans to spend around $6bn in 2027 as it races to expand capacity into a memory shortage that shows little sign of easing, according to Reuters.
That figure, if it holds, would mark a sharp escalation. Nanya set its 2026 capital expenditure at NT$52bn, roughly $1.6bn and itself more than triple the year before, with about 70% earmarked for a new plant. A jump toward $6bn the following year would be the most aggressive build-out in the company’s history.
The money is aimed largely at a single building. Nanya’s new 5A fab is scheduled to begin equipment move-in during the first quarter of 2027, and once running it is expected to roughly double the company’s total production capacity, with the initial phase alone adding more than 30,000 wafers a month.
The timing is deliberate. Nanya has warned that the AI-driven memory shortage is likely to persist through at least 2027, with some in the industry stretching that to 2028, a scarcity that has sent contract prices vertical and reached ordinary buyers, from cloud operators to Apple, which pulled its cheapest Mac Mini from sale as memory costs surged.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
Nanya’s own accounts show what the boom has done. First-quarter revenue in 2026 surged more than 580% year on year and gross margins climbed to around 68%, striking figures for a company that was posting losses not long before.
Customers are helping fund the expansion. Nanya raised NT$78.72bn, about $2.5bn, through a private placement backed by firms that depend on its output, including SK Hynix’s Solidigm unit, Japan’s Kioxia, Cisco and SanDisk, several of them locking in supply through multi-year agreements.
The strategy is a bet on climbing the value chain. Nanya has been pushing into the AI server supply chain with customised high-bandwidth memory, the specialised DRAM that sits beside AI accelerators, and domestic brokerages expect server-related products to make up more than 60% of its sales by 2027.
That would be a reinvention rather than an adjustment. Part of Taiwan’s Formosa Plastics group, Nanya has historically leaned on commodity DRAM for PCs and consumer devices, the lowest-margin corner of the market. Redirecting capacity toward AI memory is an attempt to escape the price war it has spent most of its life losing.
It is chasing rivals who moved earlier and bigger. SK Hynix is spending $51bn on a single new memory plant, Micron has broken ground on a $9bn expansion in Hiroshima, and Samsung is heading for an 18-fold profit jump on the same wave, a reminder of the scale Nanya is trying to match.
Whether $6bn is enough to close that gap is another question. Nanya remains a fraction of the size of the three giants that dominate DRAM, and the spending buys capacity that will not come fully online for years, well after the current shortage may have shifted.
That is the industry’s oldest risk, and Nanya is not immune to it. Memory is famously cyclical, and a synchronised expansion across every major maker raises the prospect that the shortage flips to glut just as the new fabs switch on, squeezing prices exactly when the debt taken on to build them comes due.
For now, demand looks like the safer bet. Nanya, like its larger rivals, is wagering that AI’s appetite for memory outlasts the time it takes to pour concrete and install tools, and that the shortage it keeps warning about will still be there when its 5A fab finally comes to life.
