Nanya’s $6.2B Capex Plan Aims to Ease AI Memory Supply Crunch by 2028

For supply chain managers reliant on high-bandwidth memory, Nanya Technology’s T$200B capex plan signals a long-awaited capacity expansion. The new fab, set to produce 30,000 wafers monthly by 2028, could ease the stubborn memory shortage that has constrained AI server production. With customers like Nvidia and Google, this investment is a critical node in the AI hardware supply network.

Taiwanese memory chipmaker Nanya Technology has unveiled a massive capital expenditure plan of over T$200 billion ($6.2 billion) for 2027, roughly four times its expected 2026 outlay, in a bold bet on the structural demand shift driven by artificial intelligence. The announcement came alongside stunning second-quarter 2026 results that underscore the severity of the current AI-fueled memory shortage. Nanya's unaudited Q2 revenue reached T$82.55 billion, a 684% jump from a year earlier, while net income soared 1,324% to T$50.19 billion. Gross margin flipped dramatically from a negative 20.6% in Q2 2025 to 79.5% in the latest quarter, reflecting pricing power and tight supply conditions.

“

Nanya's unaudited Q2 revenue reached T$82.55 billion, a 684% jump from a year earlier, while net income soared 1,324% to T$50.19 billion.

The spending plan, though still pending board approval, is centered on a new fabrication plant whose total investment at full capacity will reach about T$480 billion. The first phase is scheduled to come online in 2028 with a monthly capacity of 30,000 wafers, eventually scaling to 45,000 wafers per month. This capacity expansion is crucial because memory chips—particularly high-bandwidth memory (HBM) used in AI accelerators—are in acute shortage. Nanya counts Nvidia, Qualcomm, and Google among its customers, tying its fortunes directly to the infrastructure build-out supporting large language models and AI inference. President Pei-Ing Lee told an online press briefing that structural changes from AI are underpinning a stronger long-term memory industry outlook and that the current supply deficit is expected to persist for several more quarters.

The context is a global semiconductor industry racing to add capacity. Rivals Samsung Electronics and SK Hynix are likewise ramping investments, with South Korea making a concerted national push to expand output. Lee characterized this competitive escalation as positive for the broader ecosystem, signaling confidence in sustained demand. Nanya's bet is not without risk: the multi-billion-dollar commitment comes with execution challenges, including potential construction delays, equipment supply bottlenecks, and the ever-present geopolitical tensions surrounding Taiwan. Yet the company's market capitalization of around $47 billion suggests investors are already pricing in a prolonged upcycle.

From a market perspective, Nanya's 684% revenue growth and near-80% gross margin are extraordinary even in the notoriously cyclical memory industry. These figures illustrate that AI-driven demand is not a short-term spike but a structural reordering of computing architecture. Traditional DRAM cycles were historically driven by PC and smartphone demand; now, the explosive growth of data center AI training and inference workloads has created a new, sustained demand vector. Nanya's heavy investment signals that it expects this demand to maintain pricing power for years, enough to justify a four-fold capex increase.