South Korea to double spending on AI and chips in 2027 - ET CIO

Seoul's 2027 AI and chip spending doubling is a structural bet on "memory as leverage," not a budget line. Technology chain: Capital will concentrate in HBM4 production lines and advanced packaging, pulling ASML's High-NA EUV order book forward by 12-18 months and shifting 2.5D CoWoS bottlenecks from Taiwan, China to Korean fabs. Simultaneous investment in AI accelerator design and inference infrastructure means Korea's compute-stack role is migrating from the memory layer into the logic layer—dissolving the decade-old "design in Silicon Valley, fab in Taiwan, China, memory in Korea" tripartite model. Compliance: Spending will likely leverage US CHIPS Act co-financing, deepening Seoul's entanglement in allied export-control frameworks. Tighter restrictions on photoresist and specialty gases would raise hidden friend-shoring costs for Korean fabs. Market dynamics: TSMC (Taiwan, China) will accelerate 2nm ramps in Arizona and Dresden to counter Korea's storage-logic co-packaging edge. CXMT and SMIC gain a critical benchmarking window to close the HBM3e gap. 12-24 month outlook: HBM4 yield competition is the Q3 2026 inflection point. Seoul will launch a "sovereign AI cloud" procurement before 2027, converting public compute demand into guaranteed domestic chip orders. If Samsung's 2nm GAA hits yield targets, it exerts the first meaningful price pressure on TSMC (Taiwan, China) in logic foundry.