Morgan Stanley: The Sharpest Correction in the Memory Sector Is Over, Valuations Look Attractive
Miles Bennett
Morgan Stanley analyst Shawn Kim sees the most violent leg of the memory-chip sell-off as finished, with stocks trading at roughly 3× forward P/E and pricing in almost no growth — a tactical re-entry window.
What caused this sell-off in the first place?
Kim flagged the risk as early as July 6: DRAM's second derivative — the rate of change in price momentum — was about to peak, positioning was crowded, and momentum trades faced a pullback.
This means → the drop was not a surprise but a predictable phase of cycle aging.
Earnings-revision breadth peaked in late June and has since retreated in step with the sell-off. In plain terms = the most panicked selling is done, and a bounce from the lows is reasonable.
Where are we in the cycle now?
The report places the memory cycle in a late-stage transition during Q4 2026. This reflects a key shift: cycle-driven operating leverage can no longer outrun gradually softening prices.
In plain terms = the "rising volumes and rising prices" phase is over. From here, what matters is capital returns, locking in margins through long-term agreements, and free-cash-flow generation.
Memory stocks currently trade at roughly 3× forward P/E, pricing in almost zero growth premium. This means → if AI-driven demand proves structural rather than cyclical, traditional normalized-earnings frameworks may undervalue these companies, and multiple expansion itself could become the main return driver.
What are prices doing right now?
Industry checks show Q3 DRAM contract prices rose roughly 15% quarter-on-quarter in early trades — slightly below the prior expectation of 20%. NAND prices rose roughly 20% QoQ.
Incremental pricing momentum is slowing into Q4, and customer procurement urgency is easing.
In plain terms = prices are still rising, but the rate of increase is narrowing — the textbook signature of a late cycle.
Which names does Morgan Stanley prefer?
Stock picks center on where capex flows and where bottlenecks are tightest: DRAM and niche memory (DDR4, SLC NAND — capacity-constrained products with strong pricing power) over memory-module assemblers.
SK Hynix 2026 EPS was raised 13% (reflecting a Q2 asset-disposal gain); Samsung Electronics 2026 EPS was cut 10% (weaker consumer business). Price targets for both are unchanged.
Longer term, the report stays bullish on 25–50% earnings growth for Samsung and SK Hynix in 2027, with targets implying over 60% upside from current prices.
What has to go right for this bull case to work?
Two variables will determine whether the thesis delivers: first, whether late-cycle dynamics actually trigger accelerated capital returns (buybacks, dividends, trimming inefficient capex); second, whether structural AI demand holds up long enough to support a re-rating.
This means → if AI demand turns out to be a cyclical pulse rather than a structural shift, today's low multiples are not "cheap" — they are fair.
In plain terms = Morgan Stanley is handing investors a conditional entry ticket — whether the conditions are met will take another two to three quarters of data to confirm.
Content is for reference only, not financial advice.