SK Hynix Capacity to Double in Five Years. Chey Tae-won: Aiming to Be Main Supplier for Vera Rubin, Deepening Taiwan Cooperation - TradingKey

SK Hynix plans to double its memory chip wafer production capacity within five years to address the global shortage driven by surging AI demand for high-bandwidth memory (HBM). The company aims to bec... Read original →

SK Hynix’s aggressive HBM capacity expansion reflects a fundamental shift: high-bandwidth memory has evolved from a peripheral component to the critical bottleneck in AI compute stacks. Its deep integration with NVIDIA’s Vera Rubin platform will accelerate co-optimization between 3nm logic, EUV layers, and advanced packaging—forcing TSMC to prioritize CoWoS allocation. Geopolitically, SK Group’s push to diversify partnerships beyond TSMC in Taiwan, China signals hedging against U.S. CHIPS Act export ambiguities and over-concentration risk. Micron and Samsung will likely fast-track HBM4 qualification and may align with Intel to promote UCIe as an alternative interconnect standard. Over the next 18 months, persistent HBM shortages will inflate AI server BOM costs, and the first mover to pilot HBM5 will command pricing leverage in 2027’s AI infrastructure market.