Morgan Stanley Upgrades Memory Stocks: Sharpest Correction Over, AI Demand to Drive Next Rally; Bullish on Samsung, SK Hynix

Morgan Stanley has released a new research report on the memory industry, stating that the significant correction in memory stocks since the beginning of the year is nearing its end. The market's focus is gradually shifting from price cycles to capital returns, with catalysts such as share buybacks, free cash flow, and long-term supply agreements (LTAs) expected to become key drivers for the next phase of stock appreciation.

Shawn Kim, an analyst known for his bearish stance, believes current valuations offer an attractive tactical entry point. He views the recent pullback as merely a normal fluctuation within the broader AI super-cycle.

The report notes that Morgan Stanley had already warned in early July that memory stocks could experience short-term corrections due to slowing DRAM price increases and overly concentrated market positions. Since then, while market sentiment toward AI capital spending and demand remains optimistic, attention has turned to signs of moderating DRAM price growth starting in 4Q26, rising inventory levels, and increasing new supply—leading to caution regarding future earnings surprises. However, analysts believe that after the recent correction, memory stock valuations have significantly improved, and corporate capital return policies are likely to accelerate.

Morgan Stanley emphasizes that AI-driven demand does not conflict with traditional memory cycles. Although year-on-year DRAM price growth may have peaked, the scale of this AI build-out cycle far exceeds historical norms, with prices surging up to 700% year-over-year—about seven times higher than previous peaks. DRAM has become a critical bottleneck in AI infrastructure deployment, yet related stocks are currently trading at only about 3x forward 12-month P/E (NTM P/E), barely reflecting any long-term growth premium. The report argues that if investors begin to believe AI will extend the industry’s profitability cycle, valuation multiples still have room to expand.

For individual companies, Morgan Stanley maintains its target prices for Samsung Electronics and SK Hynix but adjusts earnings estimates. SK Hynix’s 2026 EPS forecast is raised by 13%, reflecting gains from asset disposals in Q2. Samsung Electronics sees a 10% downward revision in 2026 EPS due to weakness in consumer electronics, though forecasts for 2027–2028 are only slightly adjusted.

The report suggests the memory industry will enter the late-cycle phase starting in 4Q26, with companies placing greater emphasis on capital returns, LTAs, stable profitability, and free cash flow rather than relying on rapid price increases for operating leverage. Recent earnings forecast revisions have notably cooled, indicating the most pessimistic period may already be behind us.

Industry surveys show that 3Q26 DRAM contract prices rose approximately 15% quarter-on-quarter during initial transactions, slightly below Morgan Stanley’s original 20% estimate. 3Q26 NAND prices increased around 20% quarter-on-quarter. As price growth slows in 4Q26, customer pre-buying enthusiasm has declined, but additional capacity is expected to shift from consumer products to enterprise SSDs (eSSDs), which, combined with continued growth in AI server demand, should absorb the incremental supply.

The report further states that the market has already fully priced in the risk of slowing EPS growth over the next 12 months. However, AI-driven demand is more likely a structural shift rather than a simple cyclical trend. Therefore, applying traditional cyclical valuation methods to the memory sector may no longer be appropriate. If companies actively enhance shareholder returns, the market could re-rate the entire industry.

Morgan Stanley maintains a long-term bullish view, believing that AI capital expenditures, the rapid development of Agentic AI, and projected 25%–50% profit growth for Samsung and SK Hynix by 2027 will support fundamental strength. The report estimates that target prices for both companies imply over 60% upside from current levels. In the near term, after heavy deleveraging in July, buying momentum has returned, and investors are expected to continue favoring DRAM and niche memory types like DDR4 and SLC NAND—those most directly benefiting from AI capex—over memory module makers.