Nanya Technology Earnings Call: Memory Prices to Keep Rising in Second Half, Capex to Surge Over NT$200 Billion Next Year

Nanya Technology (2408-TW) held its earnings call today (10th), where CEO Pei-Ying Lee stated that due to the construction of a new fab, advanced process development, and growing demand for customized memory, the company's capital expenditure for this year is expected to exceed NT$50 billion. For next year, the internal preliminary plan is to raise capex to over NT$200 billion. Although the budget still awaits board approval, the investment scale is expected to more than double compared to this year.

Lee indicated that the new fab's full-capacity monthly output is planned at 45,000 wafers, with total capital expenditure estimated at approximately NT$480 billion. So far, initial investments have approached but not yet reached NT$100 billion, with the majority of construction and equipment investments to be concentrated in the coming years.

According to current plans, the first phase of the new fab will reach a monthly capacity of 30,000 wafers by 2028, with further expansion to 36,000 wafers by 2029. The timeline for the subsequent expansion from 36,000 to 45,000 wafers will depend on the deployment needs for AI infrastructure, customized memory, and advanced packaging technologies such as wafer bonding, and no definitive schedule has been set yet.

Regarding whether such massive capital expenditure would require external fundraising, Lee stated that Nanya currently has a healthy financial position, and existing funds are sufficient to support future investments. There is no strong need for fundraising at this stage, although the company has not ruled out evaluating other financing tools based on long-term capital planning.

On memory pricing, Nanya's average selling price in Q2 increased by over 60% compared to Q1. With sales volume remaining roughly flat, this drove quarterly revenue up 68.2% to NT$82.549 billion. Gross margin climbed to 79.5%, and earnings per share reached NT$14.66, both setting new historical highs for the company.

Looking ahead to Q3, while Lee did not specify the exact average selling price, he clearly stated that market prices in Q3 remain higher than in Q2, and the business outlook for Q3 and Q4 continues to improve. He noted that Nanya's existing capacity is already near full utilization, and in the short term, before the new fab ramps up, there will be little room for significant volume growth. Therefore, future operational growth will primarily come from rising memory prices and improved product mix.

Lee highlighted that strong demand from AI servers and cloud data centers is driving continued growth in HBM, server DDR5, and other high-capacity memory products, which is squeezing capacity for general-purpose DRAM used in smartphones, PCs, automotive, and consumer electronics.

Currently, memory shortages are no longer limited to high-end products—supply tightness and price increases are occurring across DDR5, DDR4, and even DDR3. The supply shortage is expected to persist for several quarters or more.

On whether price increases might gradually moderate, Lee stated that prices still depend on supply-demand gaps and end-market cost tolerance. Some low-cost, small-end products may reduce capacity or procurement due to rising memory costs, but overall market demand remains strong, and the supply gap has not noticeably narrowed.

Using the TV market as an example, he noted that high-end products can better absorb memory price hikes, while low-end products face greater cost pressure. End markets may further consolidate toward premium products, and prices for smartphones and TVs could adjust upward due to rising memory costs.

Currently, Nanya's long-term supply agreements (LTAs) cover about 50% of its capacity, with contract durations ranging from one quarter to several years. The company is gradually shifting short-term contracts to longer-term supply agreements to enhance predictability of future capacity and pricing.

In terms of product mix, revenue from AI infrastructure and server-related products accounted for over 20% in the first half. DDR5 currently contributes about 10% of revenue, and LPDDR5 is expected to begin customer validation in the second half.

Lee believes that AI demand has triggered structural changes in the memory industry. Beyond GPUs, AI computing chips such as CPUs, TPUs, and ASICs are also continuously driving demand for high-speed, high-capacity DRAM. Before large-scale new capacity comes online, the memory market will remain tight, and prices are expected to continue their upward trend.