Bernstein Asia Semiconductors & Global Memory: 2026 Is Still All About AI
Summaries are public. PDF access requires an active membership; all members have the same access. Sign in
Bernstein emphasizes that artificial intelligence remains the singularly important investment theme across the Asian semiconductor and global memory sectors for 2026. Demand for AI accelerators continues to heavily outpace supply, creating persistent bottlenecks across advanced logic wafers, high-bandwidth memory, and Chip-on-Wafer-on-Substrate packaging capacities. The firm notes the rapid ascent of Tensor Processing Units, with Broadcom expected to drive the majority of 2026 shipments, supplemented by MediaTek late in the year. Nevertheless, NVIDIA is projected to retain approximately 60% of the global packaging supply, with its next-generation Rubin platform slated for a late 2026 release. Taiwan Semiconductor Manufacturing Corporation remains Bernstein's top pick, benefiting from both graphics processing unit training and edge AI inference, with a target price of NT$1,800 based on 20 times forward earnings. TSMC's 2026 capital expenditure is projected at $47 billion, representing a 15% year-over-year increase, although the capital intensity ratio will decline as revenue growth accelerates. In the memory sector, Bernstein raised target prices for Samsung Electronics, SK Hynix, and Micron Technology to reflect stronger pricing momentum. Memory prices are expected to remain robust throughout most of 2026 before supply and demand eventually normalize in 2027. Conversely, mature node foundries such as United Microelectronics Corporation face headwinds from elevated memory costs and broader inflationary pressures.