AI Chip Race Moves Beyond HBM to Advanced Packaging
Advanced packaging is emerging as a critical factor in the global artificial intelligence semiconductor race. Even with sufficient supplies of high-bandwidth memory (HBM) and high-performance GPUs, shortages in the packaging capacity needed to integrate these components can delay shipments of finished AI accelerators. For Samsung Electronics and SK hynix, competitiveness increasingly depends on…
The global artificial intelligence semiconductor race is witnessing a shift from high-bandwidth memory (HBM) to advanced packaging as a critical factor in the industry's competitiveness. As AI accelerators become more powerful and incorporate larger amounts of HBM, the need for reliable and efficient packaging capacity to integrate these components has become increasingly important.
Samsung Electronics and SK hynix are particularly focused on delivering memory that can be seamlessly integrated into customers' AI systems. The advanced packaging industry is projected to surpass conventional packaging revenue by 2026, with capital expenditures reaching around $28 billion, according to research firm Yole Group.
TSMC, through its Chip-on-Wafer-on-Substrate (CoWoS) packaging technology, has established a dominant position in both advanced foundry manufacturing and packaging, with plans to introduce larger 14-reticle platforms to accommodate more computing chips and HBM stacks. However, the increasing demand for advanced packaging has put pressure on TSMC's capacity, driving AI chip companies to explore alternative manufacturing options and diversify their supply chains.
Brief written by urgent.news from Korea IT Times's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
