Industry Analysis
The TPCA Show 2026 is less a trade event than a structural warning: the AI hardware bottleneck is migrating from lithography to ABF substrates. Unimicron's chairman headlining the keynote confirms substrate capacity has become the new chokepoint in accelerator supply chains.
Technical cascade: Next-gen AI chips under CoWoS-L and CoPoS packaging push substrate layer counts from 16 to 24+, expanding per-die area by 3-5x. Upstream CCL suppliers (Isola, Panasonic) hit material limits; downstream, NVIDIA, AMD, and custom ASIC programs (Google TPU, AWS Trainium) are locked in a zero-sum capacity race.
Compliance exposure: US export controls on advanced packaging inputs are fragmenting the market into two tracks. Fabs operating in mainland China serving US AI customers face 15-25% compliance cost inflation, forcing painful capacity reallocation.
Competitive dynamics: Ibiden and Shinko are scaling 20+ layer lines; AT&S leverages EU AI sovereignty narratives. The existential threat is vertical integration—if TSMC or NVIDIA internalize substrate production, Unimicron's margin architecture gets structurally compressed.
12-24 month outlook: Substrates will transition from commodity components to strategic assets, replicating the EUV pricing-power dynamics of 2019-2022.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.