Memory stocks are having a moment. Shares of Micron Technology Inc. (MU), Sandisk Corp. (SNDK), SK hynix Inc. (SKHY), and Western Digital Corp. (WDC) all moved higher in premarket trading Thursday, continuing a trend that has investors excited about the intersection of artificial intelligence and memory chips.
The reason? AI servers are hungry—really hungry—for high-bandwidth memory (HBM), DRAM, NAND flash, and storage. As hyperscale cloud providers expand data-center capacity, memory makers are enjoying firmer pricing and sustained demand. And according to analysts at Morgan Stanley, this isn't just a passing phase.
AI Spending Keeps Memory Demand High
Morgan Stanley's Andrew Slimmon appeared on CNBC Wednesday and made a straightforward case: memory chips and compute power are scarce, and that scarcity isn't going away anytime soon. "There's a scarcity of memory chips. There is a scarcity of compute power," Slimmon said. "I don't think that will be resolved in the near future."
Slimmon acknowledged that some memory stocks got a bit overheated earlier this year, but he argued that the recent pullbacks have made valuations more attractive. He also pushed back on comparisons to past boom-and-bust cycles like the dot-com era or commodity cycles. His point? The dot-com bubble ended only after supply caught up with demand—and right now, supply is still playing catch-up. The AI infrastructure buildout, he said, is still in its early stages.
Morgan Stanley Sees Multi-Year Tailwind
Fellow Morgan Stanley analyst Joseph Moore echoed that view, saying the recent weakness in memory stocks has created an interesting entry point. Speaking on CNBC Wednesday, Moore noted that memory remains "at the center of every methodology for AI training and inference" and argued that recent sell-offs reflect short-term concerns, not a deterioration in long-term demand.
"We're seeing these second-derivative sell-offs," Moore said. "But the true north here is the strength of the data center and the belief from data center customers that memory's going to be a binding constraint on the ability to ramp AI for really multiple years."
Moore estimates that memory inflation could total about $80 billion this year as demand for HBM, DRAM, and NAND outpaces supply. And while higher prices are raising costs for electronics makers and cloud providers, there's no practical substitute for the memory required to train and run AI models. He added that NVIDIA Corporation (NVDA) CEO Jensen Huang recently told investors that memory shortages are likely to persist for years, forcing the industry to improve efficiency rather than rely on additional supply.
Moore expects the shortage to ease only if AI spending slows materially—a scenario Morgan Stanley does not currently anticipate. If AI investment remains strong, the supply-demand imbalance could last three to four years.
Price Action: In premarket trading Thursday, Sandisk shares were up 1.62% at $1,625.11, Micron shares were up 2.98% at $988.05, and SK hynix shares were up 6.55% at $176.10, according to market data.
