As the global semiconductor industry grapples with the physical limitations of traditional silicon scaling, the focus has shifted toward advanced packaging as the primary engine of performance gains. High-profile analyst Ming-Chi Kuo has released a new forecast regarding TSMC’s next-generation advanced packaging technology, dubbed CoPoS, which is now slated for mass production in the second half of 2028. This timeline provides a critical window into the industry's transition toward glass substrates, a shift necessitated by the extreme thermal and structural demands of next-generation AI and high-performance computing chips.
One of the most significant points of contention in the supply chain has been whether the adoption of glass substrates would render traditional materials like Ajinomoto Build-up Film (ABF) obsolete. Kuo’s research suggests a more nuanced reality: glass will not replace ABF. Instead, the two materials are expected to coexist in a sophisticated hybrid architecture. In this design, chip interconnectivity is achieved through a combination of the chip-side Redistribution Layer (RDL), glass-through vias (TGV) for vertical copper interconnects, and standard ABF buildup layers.
This structural interdependence effectively debunks the narrative of a zero-sum game between glass processing and existing organic film technologies. By utilizing glass for its superior flatness and thermal stability while retaining ABF for its proven buildup capabilities, TSMC is charting a path that prioritizes reliability over radical substitution. This approach ensures that the existing supply chain for organic substrates remains relevant even as the industry integrates more exotic materials to handle the power-hungry requirements of trillion-parameter AI models.