AI is moving from large-scale model training toward inference and agent-based services. GPUs, HBM, and 2.5D interconnects have powered the first wave of AI acceleration, but the future is diversifying—bringing AI accelerators, CPUs, general-purpose memory, and on-die AI semiconductors into play. From a memory-centric view, the demand for massive capacity, ultra-fast speed, and wide I/O bandwidth is clear. Yet the most urgent challenge is low power. This talk will explore the current limits of DRAM core die technology in HBM and highlight new device structures designed to break through these bottlenecks. We will look at the shift beyond the traditional 1T1C DRAM cell toward oxide-semiconductor-based 2T0C architectures, advances in bonding methods, and the potential of glass substrates. We will also examine strategies for achieving low-power operation—from smarter architectures to heterogeneous integration of processors and memory—and discuss how these innovations open opportunities across devices, materials, equipment, and packaging. By framing these breakthroughs within a broader AI semiconductor technology framework, the session will provide a clear view of both the technical challenges and the strategic directions shaping the next era of global semiconductor innovation.

※ 연사정보

AI is moving from large-scale model training toward inference and agent-based services. GPUs, HBM, and 2.5D interconnects have powered the first wave of AI acceleration, but the future is diversifying—bringing AI accelerators, CPUs, general-purpose memory, and on-die AI semiconductors into play. From a memory-centric view, the demand for massive capacity, ultra-fast speed, and wide I/O bandwidth is clear. Yet the most urgent challenge is low power. This talk will explore the current limits of DRAM core die technology in HBM and highlight new device structures designed to break through these bottlenecks. We will look at the shift beyond the traditional 1T1C DRAM cell toward oxide-semiconductor-based 2T0C architectures, advances in bonding methods, and the potential of glass substrates. We will also examine strategies for achieving low-power operation—from smarter architectures to heterogeneous integration of processors and memory—and discuss how these innovations open opportunities across devices, materials, equipment, and packaging. By framing these breakthroughs within a broader AI semiconductor technology framework, the session will provide a clear view of both the technical challenges and the strategic directions shaping the next era of global semiconductor innovation.

※ 연사정보

※ 연사정보

As monolithic scaling approaches the physical and economic limits of Moore’s Law, Advanced Semiconductor Packaging (ASP) has transitioned from a routine backend process to the primary driver of microelectronic innovation. Driven by an unprecedented surge in demand for Artificial Intelligence (AI), high-performance computing (HPC), and 5G infrastructure, ASP has emerged as the new strategic layer in the global technology ecosystem, reshaping hardware architecture and market competition.
Through heterogeneous integration, advanced packaging techniques—such as 2.5D/3D stacking and fan-out wafer-level packaging—bypass single-die fabrication constraints. By combining disparate, optimized chiplets into a unified system, ASP fulfills the skyrocketing commercial need for massive interconnect densities, ultra-low latencies, and extreme thermal efficiencies.
Ultimately, semiconductor leadership is no longer defined solely by front-end lithography nodes, but by the ability to architect complex, multi-die systems. Packaging has effectively become the critical differentiator, supply chain constraint, and the ultimate bottleneck for future technological supremacy.

※ 연사정보

Biography
Jung-Hoon Lee is a data scientist in the Digital Solutions team at the Electronics business of Merck KGaA, Darmstadt, Germany. He has successfully executed several data-sharing use cases with customers worldwide in the semiconductor sector. He holds a bachelor’s degree in physics and chemistry from POSTECH in South Korea and a master’s degree in physics from the University of Hamburg in Germany. With extensive experience in manufacturing and R&D, Jung-Hoon leverages his expertise to drive impactful data science solutions within the organization.