Samsung to triple HBM prices next year - 매일경제

Samsung's decision to triple HBM pricing is not a cost pass-through—it is a strategic repricing signal following its 2024 qualification lag behind SK Hynix. The technical bottleneck has shifted from DRAM die to TSV stacking and hybrid bonding. This aggressive pricing implies Samsung's HBM4 yield has crossed a critical threshold, and the company is attempting to reclaim pricing power through a technology-premium narrative. The downstream impact is immediate: HBM now accounts for over 35% of AI accelerator BOM costs, and a threefold price delta will directly fracture hyperscaler TCO models, forcing accelerated adoption of HBM4 or 3D-stacked alternatives. In the competitive arena, SK Hynix—holding roughly 50% market share—will almost certainly adopt a stable-price, volume-lock strategy, securing long-term contracts with NVIDIA and AMD during Samsung's repricing window. Micron will likely exploit the gap with cost-competitive HBM3E for second-tier AI silicon customers. More critically, Samsung's aggressive pricing will accelerate multi-sourcing consolidation among hyperscalers, paradoxically eroding its own negotiating leverage. 12-24 month outlook: HBM is transitioning from a commodity component to a strategic bottleneck resource for AI compute. If the market absorbs this pricing, memory chips complete their migration from cyclical goods to quasi-strategic assets. If customers collectively resist, it signals a structural inflection in AI capex by H2 2025. Pricing power, not process node, is the true battleground ahead.