RISING TIDES
Observations from the IMC Research Process

The Memory Trade Reloaded

Memory has become a determinant of AI scalability, driving demand and pricing power for incumbent players.

• From Enabler to Backbone: Since October when we last highlighted the memory theme, its role in AI infrastructure has become increasingly more important. Memory is now foundational as AI systems scale beyond training into real-world deployment. Memory constraints, instead of compute, are increasingly defining system architecture, performance, and cost.
• Inference has Reshaped Demand: AI demand has moved decisively from training-centric to inference-driven workloads. Real-time agents, enterprise applications, and multi-modal models require persistent access to vast datasets. This shift is structurally more storage-intensive, reinforcing demand for high-capacity HDDs for cost-efficient data retention and NAND-based SSDs for fast-access inference layers.
• Memory Leaders Have Seen Strong Momentum: Seagate (STX) and Western Digital (WDC) remain central to hyperscale storage expansion, where cost-per-bit efficiency is critical. In NAND and DRAM, Micron (MU), Samsung Electronics (SGLKR), SK Hynix (HYLKR), and SanDisk (SNDK) are benefiting from rising utilization and improving margins. Growing memory complexity is also lifting the equipment ecosystem, with LAM Research (LRCX), KLA (KLAC), and Advantest seeing increased demand for etch, metrology, and testing.