Applied Materials says efficiency in moving data, rather than computing speed, is the main bottleneck for AI semiconductor performance and that the industry solution is 3D scaling. It described 3D scaling across logic, memory and packaging, and said capacity expansion should accompany demand. The company detailed approaches for DRAM, a shift in HBM stacking from micro-bumps to hybrid bonding, and panel interposer development, while highlighting cooperation with Samsung Electronics and SK Hynix in Korea.

SK Hynix has completed negotiations on long-term supply agreements with about 10 customers, including core clients, it said. The contracts typically run five years and include financial arrangements such as deposits alongside purchase commitments. The company said tight memory supply-demand conditions are likely to persist, supporting customers’ push for mid- to long-term supply stability. It is also negotiating 2027 HBM supply volumes and prices with key customers.