Memory chip rally faces peak rate of change as EPS revisions top 89%

Morgan Stanley said the memory chip industry is approaching a "peak rate of change" across pricing, inventory and earnings revisions, signaling near-term volatility before a potential 35% to 40% earnings surge in 2027.

"The rate of change in pricing is peaking, but the cycle itself is not yet over," Morgan Stanley analysts wrote in a July 6 research note. The bank expects memory earnings to grow 35% to 40% in 2027.

Three indicators are converging at peak levels: DRAM price year-over-year growth has narrowed from first-quarter highs, inventory cycle improvements are flattening, and earnings-per-share revision breadth has reached about 89%, a level that historically leaves limited room for further upgrades. The bank identified three core debates driving the uncertainty — whether AI compute capacity is genuinely oversupplied, whether token consumption trends are shifting toward cheaper alternatives, and why long-term supply agreements have failed to trigger valuation re-rating.

The immediate risk is a rotation out of crowded memory positions into laggard sectors, with hyperscaler cloud capital expenditure guidance during the upcoming second-quarter earnings season serving as the decisive catalyst. Memory stocks have already corrected about 17% in the current drawdown, compared with prior bull-market pullbacks of 15% to 32% since the generative AI cycle began in late 2022.

Three Debates Define the Uncertainty

The first debate centers on whether AI compute capacity is genuinely oversupplied. An unconfirmed report that one of the largest AI capital expenditure buyers may have excess compute available for sale has fueled bearish narratives, though Morgan Stanley characterized this as infrastructure monetization rather than true oversupply. The second debate involves a shift in enterprise token consumption — companies that once encouraged maximum token generation are now seeking cheaper alternatives, including open-source large language models from Chinese developers, to control IT budgets. The third debate questions why long-term supply agreements have not driven valuation re-rating, with the bank noting that market participants remain skeptical because past LTAs were either renegotiated or forced customers to accept unwanted inventory.

Hyperscaler Earnings Hold the Key

Morgan Stanley said hyperscaler cloud capital expenditure guidance during the second-quarter earnings season will matter more for memory stock direction than comments from memory companies themselves, because chipmakers at this point in the cycle are likely to maintain optimistic outlooks. UBS on Monday lifted its DDR contract pricing forecasts to 32% quarter over quarter in the third quarter from 17%, while Citi added Micron Technology Inc. to its 90-day upside catalyst watch and Bank of America reiterated a buy rating with a $1,550 price target. Samsung Electronics Co. reports quarterly results Tuesday, offering the next major read on high-bandwidth memory pricing and demand, while SK Hynix Inc. is set to list on the Nasdaq on July 10.

Investment Angle

Memory stocks have already corrected about 17% in the current drawdown, compared with prior bull-market pullbacks of 15% to 32% since the generative AI cycle began in late 2022. The bank's long-term thesis remains intact — it expects memory earnings to grow 35% to 40% in 2027 — but near-term positioning risk is elevated. Micron trades at 22 times trailing earnings with a forward multiple of 7 times, while SanDisk carries a trailing price-to-earnings ratio of 60 times. The key swing factor is whether hyperscaler cloud capital expenditure guidance in the coming weeks confirms or contradicts the peak-rate-of-change thesis.

This article is for informational purposes only and does not constitute investment advice.