Winbond's Business Update: Customers Seek Capacity Through 2030 as Module B Expansion Kicks Off

Memory manufacturer Winbond (2344-TW) held its business update today (6th), where CEO Chen Pei-ming stated that memory supply remains tight, with multiple customers seeking long-term agreements extending to 2029 and even 2030, and requesting early reservation of production capacity at the new Kaohsiung Module B facility. In response to long-term demand, the company has officially launched the Module B capacity expansion plan, scheduled to break ground in January 2027, begin equipment installation in 2029, and achieve mass production by the end of 2029.

Chen noted that the current monthly capacity of Module A in Kaohsiung will expand from 15,000 wafers to 24,000 wafers, with pilot production expected by year-end. However, after the Module A expansion, factory space will be nearly fully utilized. To meet strong customer demand for post-2029 capacity, the company has therefore initiated the construction of Module B.

According to current plans, Module B is scheduled to begin construction in January 2027. If progress proceeds smoothly, equipment installation could start as early as January 2029, adopting a staggered approach with equipment arriving and being commissioned incrementally. Larger-scale production is expected to commence in Q4 2029, with actual output and revenue contributions primarily materializing in 2030.

Chen indicated that many customers are already seeking to negotiate post-2029 supply arrangements beyond existing long-term contracts, and some have even inquired about signing long-term supply agreements for Module B capacity as early as 2030—reflecting intense global competition for mature and niche memory capacity.

Winbond believes that beyond AI-driven HBM demand, DDR5, LPDDR5, future LPDDR6, as well as the company’s core offerings such as niche DRAM, CUBE, and customized memory, are all facing supply constraints. The company anticipates that memory supply in 2027 may be tighter than in 2026.

The cleanroom area of Module B is estimated at approximately 30,000 square meters—about double the current Kaohsiung site’s 15,500 square meters. If fully equipped in the future, it could support a monthly capacity of 50,000 to 60,000 wafers. However, the company will not complete all equipment investments at once, instead opting for phased expansion based on customer demand, product mix, and long-term supply agreements.

Chen stated that Module B might initially deploy 10,000, 15,000, or 20,000 wafers of capacity, gradually increasing based on market conditions. Future products will include standard DRAM, CUBE, Wafer-on-Wafer, customized ASIC memory, and silicon capacitors. Actual equipment configuration will be adjusted according to customer forecasts and commitments.

In terms of process technology, Module B will support next-generation 14nm and future 12nm DRAM processes, with plans to introduce EUV equipment. The company will first develop a non-EUV version of the 14nm process, later preparing for 12nm mass production using EUV technology.

Given that typical semiconductor equipment lead times range from 12 to 14 months—and EUV equipment can take up to 3 to 3.5 years—Winbond must proactively plan equipment procurement and capacity deployment. Module B will adopt a phased equipment installation approach to avoid excessive upfront capital outlays and better align capacity with actual customer demand.

Regarding capital expenditures, Winbond estimates total CapEx for 2026 at approximately NT$39.5 billion, with around NT$7.5 billion already invested in the first half, primarily for equipment expansion and process upgrades at the Kaohsiung plant.

As for the total investment for Module B, Chen admitted the scale will be "very large." Citing an industry peer investing NT$350 billion to build 45,000 wafers per month, if Winbond’s initial phase targets 15,000 wafers, the investment could be roughly one-third of that figure. However, the final amount will depend on process nodes, equipment volume, and the extent of EUV adoption.

On existing capacity, the Kaohsiung plant currently operates at about 15,000 wafers per month, expected to rise to 24,000 by year-end. Current 16nm monthly capacity stands at only about 2,000 wafers. The company’s ideal goal is to increase 16nm output to 16,000 wafers per month, though the transition speed depends on customer demand for 20nm products.

With increased capacity and the introduction of 16nm processes, Winbond expects bit output from the Kaohsiung plant to potentially double by 2027. The company will also maximize effective output from existing facilities before Module B ramps up, through process scaling, yield improvements, and product mix optimization.

Winbond emphasized that when signing long-term contracts in the future, it will prioritize strategic customers with stable demand and 5–10 year competitiveness who are willing to co-invest in capacity, rather than simply chasing order volume—ensuring large-scale capital spending on Module B aligns with long-term order commitments.