Citi forecasts that demand for memory semiconductors will outpace supply over the coming years, with shortages expected to worsen through 2031 as artificial intelligence applications require increasing amounts of memory and storage capacity.
In a research note published Monday, the bank said continual learning could become a central feature of AI development over the next five years.
This process allows AI models to incorporate new information and tasks while maintaining access to previously learned data, increasing requirements for both processing-related memory and long-term storage.
The bank expects demand for high-bandwidth memory to rise substantially, forecasting a 62% year-over-year increase to 75.2 billion gigabits in 2027 and a further 69% expansion to 127.0 billion gigabits in 2028.
Citi also anticipates additional demand for enterprise solid-state drives as AI systems accumulate and retain larger volumes of information.
The bank’s DRAM projections indicate that demand will grow faster than production capacity. Global demand is forecast to increase 30% in 2027 and 35% in 2028, while supply growth is expected to reach 19% and 22%, respectively.
As a result, Citi forecasts DRAM supply-demand ratios of -8.7% in 2027 and -9.7% in 2028.
The NAND market is expected to experience a similar divergence between supply and consumption.
Citi projects NAND demand growth of 29% in 2027 and 33% in 2028, exceeding anticipated supply increases of 21% and 25%.
The bank estimates that the NAND supply-demand ratio will move from -0.8% in 2026 to -6.1% in 2027, before reaching -5.5% in 2028.
Alongside its industry forecasts, Citi identified several publicly traded companies in its semiconductor research.
Its preferred memory stocks include Samsung Electronics (USOTC:SSNHZ), SK Hynix (NASDAQ:SKHY), Micron (NASDAQ:MU), Sandisk (NASDAQ:SNDK) and Kioxia (USOTC:KXIAY).
The bank also highlighted semiconductor equipment and materials companies Montage (USOTC:MNTCF), Applied Materials (NASDAQ:AMAT), Lam Research (NASDAQ:LRCX), TES, Eugene Tech and TechWing.
Citi’s forecasts indicate that the adoption of continual learning could increase demand across multiple memory categories, with projected supply growth remaining insufficient to meet consumption through the forecast period.
Leave a Reply