Morgan Stanley: Memory Super Cycle Not Over, Pullbacks Are Buying Opportunities
Claire Weston
Morgan Stanley's latest report calls the recent memory-sector pullback a healthy reset within an ongoing bull run — the AI-driven super-cycle is still accelerating, earnings upgrades remain strong, and stocks trading at roughly 5× forward P/E could re-rate to 8–10×.
Why isn't this pullback a topping signal?
Memory stocks staged a parabolic rally from their March lows, while leveraged-ETF exposure swelled and hedge-fund and retail positioning grew crowded — a price reset was inevitable.
This means → the trigger was overheated positioning, not deteriorating fundamentals — the cycle itself has not turned.
Morgan Stanley raised bear-case valuations for SK Hynix and Samsung by 175% and 58%, respectively. In plain terms = even the "worst-case" valuation anchor is rising sharply, signaling stronger institutional confidence in where the floor sits.
How cheap are these stocks right now?
Samsung trades at roughly 5.2× forward P/E; SK Hynix at about 4.7× — both in the low end of historical cycle ranges.
The report forecasts DRAM prices rising 20%–30%+ in Q3 2026, enough to keep year-on-year growth accelerating.
If long-term agreements reach over 70% of total supply over the next 3–5 years, DRAM companies could theoretically re-rate from ~5× to 8–10× P/E. This means → the higher the share of locked-in supply, the more predictable earnings become, and the higher the multiple the market will pay.
How is AI fundamentally changing the demand logic for memory?
Memory has shifted from a consumer-electronics "accessory" to a core input for AI "intelligence production" — higher HBM bandwidth (high-bandwidth memory, a stacked-memory architecture designed for AI chips), larger DRAM/NAND capacity translates directly into more tokens produced per unit of time.
In plain terms = memory demand used to follow the smartphone and PC upgrade cycle — a roughly fixed market. Now, as AI models grow larger, context windows lengthen, and more users run more AI agents in parallel, there is no natural ceiling.
Every new GPU generation is bottlenecked by memory, not compute: per-system DRAM is rising from 80 GB on the A100 to 288–768 GB on Rubin GPUs — a 4–7× increase — while AI chip shipments are growing at roughly 60% annualized.
Could falling prices actually boost demand?
Morgan Stanley argues that AI demand has a different price elasticity than past cycles: lower DRAM prices → lower inference costs → faster AI deployment → new incremental demand.
This means → a price decline no longer just makes the same product cheaper — it expands the total addressable market, creating a self-reinforcing positive feedback loop on the demand side.
This reflects a fundamental identity shift for memory — from "cyclical consumable" to "infrastructure input." The slope of the demand curve has changed.
Supply discipline — the key variable for how long the bull run lasts?
The report acknowledges that memory prices will eventually decline: current investment plans begin ramping production in H2 2027, making a cycle reversal inevitable.
Historically, cycle busts were almost never caused by weak demand — the problem was the industry over-expanding capacity while filling a demand gap. Current management guidance points to more disciplined capacity additions, and cleanroom plus EUV equipment constraints also limit supply growth.
In plain terms = supply discipline is the biggest "faith test" of this bull run — if manufacturers resist the urge to over-build, the cycle runs longer with a higher peak. But the report concedes that supply discipline has never been sustainably maintained in memory history.
What risks should investors watch?
Morgan Stanley expects the cycle could approach its peak near year-end, but with AI-agent demand accelerating since January 2026, the top may still be several quarters away.
Key risk checklist: ① a technology-efficiency breakthrough that sharply reduces memory usage; ② a cooling of AI competition; ③ demand growth diverging from its exponential trajectory; ④ tightening liquidity.
Liquidity has already tightened notably versus Q1, and it typically leads the equity market by about three months. This means → even if fundamentals stay positive, sustained funding contraction could cause stock prices to peak before earnings do — a variable worth tracking closely.
Content is for reference only, not financial advice.