Semiconductor Slides into Bear Market, but Morgan Stanley Urges Buying Memory Stocks: Don’t Be Scared by Past Cycles, AI Demand to Support, Micron Rally Just the Beginning

After the Philadelphia Semiconductor Index (SOX) officially entered a bear market last week, memory-related stocks rebounded on Monday. Morgan Stanley analyst Joseph Moore stated that the current memory upcycle, driven by AI data center construction, is fundamentally different from past cycles, and investors should not misjudge the industry trend due to short-term weakness in the consumer electronics market. The AI-driven DRAM supply-demand imbalance is expected to persist for many years.

On Monday, Micron Technology (MU-US) and SanDisk (SNDK-US) both rose over 5%, while SK Hynix’s (SKHY-US) American Depositary Receipts (ADR) gained more than 2%, reflecting market appetite for memory stocks on weakness.

In a recent report, Joseph Moore emphasized that the current semiconductor cycle is entirely driven by AI data center buildouts. Therefore, while weakness in PC, smartphone, and consumer electronics markets may still emerge, affecting spot prices and inventory levels, these factors are insufficient to alter the overarching AI demand trend.

He noted that recent market signals dragging down memory stocks largely stem from PC and smartphone markets, not from actual AI infrastructure demand, creating misleading "false signals" for investors.

Morgan Stanley pointed out that there are no signs of easing in the memory supply shortage caused by data center demand—on the contrary, it continues to worsen. The firm forecasts that DRAM prices in Q3 2024 will rise at least 25% from Q2, primarily driven by continued heavy procurement of high-bandwidth memory (HBM) and DRAM products for AI servers.

Moore warned that using past memory cycles as a basis for selling now would cause investors to miss a fundamental industry shift. He explained that AI model training and inference are consuming DRAM at massive scale, leaving significantly less memory supply available for other end markets.

DRAM has become one of the biggest bottlenecks in AI development, affecting not only PC and smartphone manufacturing but also pushing cloud service providers to pay premium prices for memory rather than hoard inventory—indicating persistently strong demand.

Morgan Stanley estimates that the global AI-driven memory supply-demand gap could persist through 2028. The firm sees no evidence that memory supply will catch up with AI demand, meaning the memory industry will remain in a prolonged state of undersupply.

Although memory manufacturers are signing long-term supply agreements with customers and AI chip designs are gradually improving memory efficiency—potentially reducing the volatility of the industry cycle—Morgan Stanley believes this could actually extend the current memory upcycle, making it more favorable for long-term stock performance.

In terms of investment strategy, Moore remains most bullish on compute chip stocks, viewing Nvidia (NVDA-US) and Broadcom (AVGO-US) as the most attractive core AI plays. However, he also noted that memory stocks have begun to catch up in performance, making now a good entry point for memory exposure.

Several investment banks have recently maintained positive outlooks on AI memory. As demand for high-bandwidth memory (HBM) grows and AI servers require increasing DRAM and NAND Flash capacity, major memory suppliers like Micron, SK Hynix, and Samsung Electronics are expected to benefit from the AI investment boom over the coming years.

Analysts believe that despite the semiconductor sector’s recent sharp pullback due to valuation corrections and profit-taking, AI data center construction remains the core driver of memory demand growth. As long as AI infrastructure expansion continues, the current memory bull cycle is likely to persist.