South Korea Announces $520B Public-Private Chip Investment; Samsung, SK Hynix Build 4 HBM Fabs

South Korean President Lee Jae Myung announced an 800 trillion won ($520 billion) coordinated public-private investment plan on Monday, June 29, to expand semiconductor production capacity alongside Samsung Electronics and SK Hynix, framed as essential to maintaining Seoul's competitive edge in the global AI race. The initiative, announced in a televised state address with the CEOs of both companies flanking him, aims to position Korea to capitalize on surging demand for memory chips and HBM accelerators as AI training and inference workloads scale globally.

The centerpiece is construction of four new production facilities—two each from Samsung and SK Hynix—in the southwestern region near Gwangju, far from existing Seoul-area clusters. Samsung will also build HBM packaging facilities in Chungcheong, targeting the critical bottleneck: high-bandwidth memory chips required for AI model training and inference are now the scarcest link in the GPU supply chain, with SK Hynix supplying the vast majority of HBM going into NVIDIA accelerators. The government will expedite permitting, potentially bringing fab completion forward by 12 years—from the mid-2040s to the mid-2030s.

The $520 billion headline figure dwarfs the U.S. CHIPS Act ($52 billion in direct subsidies), though the comparison is imperfect: much of Korea's total is private capex that the state is coordinating, with government contributions including subsidies, faster approvals, and regional infrastructure. SK Hynix committed $15 billion to new facilities in February; this plan substantially accelerates and expands that commitment as AI memory demand projections have climbed sharply.

For practitioners: this move reflects a strategic choice by major economies to secure domestic memory-chip sovereignty and HBM supply. Korea's dominance in memory—with Samsung and SK Hynix holding roughly 70% of global DRAM and 80%+ of HBM—makes this an asymmetric advantage. Architects deploying large-scale AI at cost-conscious scales should monitor HBM availability and pricing closely; Korea's expansion could ease constraints but timing and yield ramp matter enormously.