Hanmi National (Seoul) – According to an analysis released by global financial institution JP Morgan, the rapid increase in computational power in the field of artificial intelligence (AI) is expected to lead to a shortage of memory semiconductors for two years. As GPUs and CPUs perform AI calculations, data storage and transmission speed become crucial; the current market may experience supply–demand imbalances beyond expectations.
JP Morgan strategist Jay Kwon described the memory semiconductor shortage as “a structure in which supply and demand cannot fully align, causing prices and quantities to rise simultaneously.” He forecast that even though the total addressable market (TAM) for memory is expanding, supply will not keep pace, leading to continued shortages over two years. The strategist noted that expansion of computation from GPU to CPU significantly increases memory demand, and this shortage would heighten uncertainty for investors.
Memory semiconductors are essential for AI servers and high‑performance equipment; particularly large memories such as HBM (High Bandwidth Memory) stack multiple DRAMs vertically to accelerate data transmission. When shortages occur, AI infrastructure investment rises while cost structures may be affected. In Korea, Samsung Electronics and SK Hynix are regarded as major suppliers.
Samsung Electronics announced on July 30 that it has entered a supply agreement with global data‑center customers and disclosed ongoing negotiations. This contract is deemed directly relevant to the expansion of HBM demand by cloud providers’ AI infrastructure. Accordingly, memory prices have been rising, and recent quarters show a trend of price adjustments.
Kwon suggested long‑term supply agreements as a solution that could ease shortages. Long‑term contracts set quantities and terms for suppliers and customers in advance, reducing imbalances. Thus the memory shortage may indirectly influence cost structures of AI servers and high‑performance computing equipment.