Unimicron spent NT$10.088 billion to acquire an existing facility. The real signal is that the capacity expansion bottleneck has shifted from construction to material availability.
Renovation schedules stretch into next year, but you need to move in next month. No one waits for the construction crew; they buy a finished property and retrofit the utilities. The package substrate industry is taking this exact path for capacity expansion.
On September 30, Unimicron, a Taiwan-based IC package substrate manufacturer, announced it had acquired land and a factory in Hukou Township, Hsinchu County, formerly owned by Bridgestone Taiwan, via public tender for NT$10.088 billion. The land area is 19,072 ping, the building area is approximately 17,601 ping, totaling over 36,000 ping. For a company producing ABF substrates and PCBs, this is a sum equivalent to a significant portion of annual capital expenditure spent on real estate.
NT$10.088 billion
Total transaction amount
Land area
36,000 ping
Combined land and building area
These three figures correspond to the total transaction amount, land area, and the combined area of land and buildings. Note the last item—the additional 17,600 ping beyond the land area consists entirely of existing structures, which represents the 'time' actually purchased in this deal.
If the only issue were land, Taiwan does not lack industrial land; it lacks suitable land. Unimicron stated plainly that its substrate bases in Taiwan include Shanying, Xinfeng, and more recently Yangmei and Guangfu. Yangmei Plants 2, 3, and 5 will be built sequentially over the next three years, but existing bases currently have no new land available for expansion.
More critical is how capacity is reserved. Unimicron notes that customers typically book five years of capacity in advance. This means substrate capacity is not sold on the spot market but is locked in before construction is complete. Even with Yangmei and Guangfu sites ramping up production over the coming years, they will not fill the five-year reservation orders already signed.
Taiwanese media reporting on this deal cited an industry consensus: tech companies increasingly prefer buying existing industrial facilities. However, properties meeting location, power supply, fire safety regulations, and free of soil contamination are scarce. Consequently, buyers act quickly when suitable properties appear. Unimicron’s public tender concluded on the last trading day of September.
Capital is not the tightest constraint. Unimicron disclosed during its September 8 earnings call that long-term contract prepayments from customers pushed cash on hand from NT$60 billion in the first quarter to NT$93 billion, with management stating it has since exceeded NT$100 billion. The fact that customer funds arrive before facilities are acquired marks a fundamental departure from capacity expansion patterns in the previous upcycle.
This dynamic admits two interpretations. One view holds that demand for high-end ABF substrates is robust, compelling suppliers to race for speed in capacity expansion. A more sober reading notes that globally, only Unimicron and Japan’s Ibiden possess strong mass-production capabilities and yields in this segment; the fact that even one of them is no longer building its own capacity suggests that new capacity lead times can no longer keep pace with customer schedules. Both perspectives point to the same conclusion: time is scarcer than money.
Zooming out one layer reveals that facilities are not the hardest constraint—buildings can be purchased off the shelf, but materials cannot. High-end ABF substrates require high-performance glass fiber cloth, with T-glass being the most constrained category. According to market research firm TrendForce, Nittobo holds approximately 90% of the global supply share for this material.
Expansion timelines illustrate the issue most clearly. Nittobo’s new fab in Taiwan, China, focused on glass fiber yarn, targets mass production as early as the second half of 2026. However, yarn must be woven into cloth, and the cloth must pass substrate manufacturer certification, meaning actual delivery is likely not until the first half of 2027. The new Japan-based facility faces an even later timeline, with production starting in 2027 and shipments only in 2028. Citing industry sources, Taiwan’s Economic Daily reported that T-glass shortages will persist at least through 2027.
Unimicron is not the only one hampered by this material constraint. Kinsus stated in August that the T-glass shortage has cost the company approximately 10% to 15% of potential monthly revenue. Consequently, it plans to install equipment at its Taiwan, China, site in 2027 to expand high-end ABF capacity by 25%. Material shortages ultimately manifest as lost revenue, a metric far more concrete than any supply-demand forecast.
Another often-overlooked ratio: T-glass accounts for approximately 15% to 20% of BT substrate costs, but only 5% to 10% of ABF substrate costs. This means price increases in this material do not deliver the most severe cost shock to ABF; rather, they determine a different outcome—whether the product can be manufactured at all. High costs can be negotiated; material scarcity can only be endured.
⚠️ Note
The 2027 constraint on high-end ABF shipments will not be substrate factory capacity, but the supply cadence of upstream glass fiber cloth. The rationale lies in a one-tier difference in time scales: facilities and equipment can be deployed within 12 to 18 months, whereas the journey from glass fiber yarn to cloth and then to customer certification spans two to three years. This assessment can be overturned at any time—if public data from substrate manufacturers in the second half of 2027 shows declining capacity utilization and a loosening of material supply, it would indicate that the bottleneck has shifted back to the demand side, invalidating my judgment.
No one can currently answer when the facility acquired by Unimicron will reach full capacity. In the first half of next year, glass fiber cloth shipments will begin from Nippon Electric Glass’s fab in Taiwan, China. How much of that output is allocated to high-end ABF versus BT will likely reveal more about the priority hierarchy than any capacity expansion announcement.