Morgan Stanley: Memory stocks at the top of the cycle appear to have the lowest valuations, but in reality, they are in the most dangerous position.

October 13 — Morgan Stanley's latest report points out that Samsung and SK Hynix currently trade at forward P/E ratios of only about 3 to 4 times, which appears cheap but may conceal a valuation trap at the cycle top. Historically, memory stocks have tended to peak ahead of earnings peaks, by about 4 months in 2008 and 2012, and by about 8 months in 2018. Although the AI-driven memory supercycle is still ongoing, with tight supply and demand, rising prices, and improving earnings still providing support, as stock prices have surged and earnings expectations have been revised upward in a concentrated manner, optimistic expectations have already been fully priced in, and further upside faces a test.

Morgan Stanley believes that the stronger the fundamentals of the memory industry, the more easily the risk of a cycle top is underestimated, and high prosperity may not translate into sustained excess returns. The report warns that high profits will eventually attract new capacity, with supply expected to be gradually released starting in the second half of 2027, potentially triggering a new cyclical turning point. Asian memory stocks have cumulatively risen about 9 times since the rally began in September 2025, while consensus EPS estimates have been revised up by an average of 432%. The low P/E ratios of Samsung and SK Hynix may be exactly the typical trap of cyclical stocks near peak earnings. To support the upside implied by current stock prices, DRAM contract prices would need to rise by another roughly 70% quarter over quarter, which is very difficult. In addition, the evolution of data center architecture may change the distribution of returns across the industrial chain: memory accounts for about 29% of a single Nvidia NVL72 rack, but falls to 19% when expanded to 8 interconnected racks, while networking equipment rises from 10% to 20%. Memory manufacturers may not benefit proportionally from the growth in AI infrastructure investment.