Citi sees memory market undersupply worsening through 2031

Investing.com -- Citi analysts project that continual learning in artificial intelligence will drive significant memory demand across multiple product categories, leading to worsening undersupply through 2031. The bank made this forecast in a research note published Monday.

The investment bank expects continual learning to become a central theme in AI development over the next five years. This AI learning approach reinforces models with training on new tasks and knowledge, creating persistent need for both model updates and access to previously acquired data.

Citi forecasts HBM bit demand to increase 62% year-over-year to 75.2 billion gigabits in 2027 and 69% to 127.0 billion gigabits in 2028. The bank also projects eSSDs demand to expand as retaining previously learned data requires substantial storage capacity.

For DRAM, Citi projects global demand to grow 30% in 2027 and 35% in 2028, driven by continual learning affecting both server DDR5 and HBM demand. The bank forecasts DRAM supply-demand ratios of -8.7% in 2027 and -9.7% in 2028, with limited supply growth of 19% and 22% respectively.

On NAND, Citi expects demand to grow 29% in 2027 and 33% in 2028, exceeding supply growth of 21% and 25%. The bank projects NAND supply-demand ratios of -6.1% in 2027 and -5.5% in 2028, compared to -0.8% in 2026.

Citi named Samsung Electronics, SK Hynix, Micron, Sandisk, and Kioxia as top memory picks. For semiconductor equipment and materials, the bank highlighted Montage, AMAT, Lam Research, TES, Eugene Tech, and TechWing.

You May Also Be Interested In

• Goldman Sachs Positive on Meta Platforms Inc. (META): 'We see a key advantage for Meta being scale & distribution'
• Wall Street's new pair trade: Long software, short chips
• Chamberlain University names Jason Dunne as chief academic officer