DRAM Extends Its Slide as Morgan Stanley Tells Clients to Sell Chips, Buy Cloud

The Rotation Call

$DRAM is a wrapper on the memory chip complex, and that complex is exactly where the AI trade is unwinding fastest. On July 6, Morgan Stanley's Mike Wilson told clients it was time to 'sell chips, buy cloud', arguing momentum in semiconductors is fading and that capital is rotating out of overextended silicon names into hyperscalers like Microsoft, Amazon and Meta. He singled out memory as bearing the brunt of the correction — and said it may not be over. When the most-watched equity strategist on the Street names your sector as the crowded exit, the ETF that packages that sector into a single ticker wears it directly.

Fundamentals Are Fine, Positioning Isn't

Here is the tension worth sitting with: the memory tape and the memory business are telling opposite stories. The stocks slid into a bear market, down more than 20% from June highs, after a Korean brokerage pegged SK Hynix's Q2 profit roughly 8% below consensus on softer HBM4 shipments. Yet the physical market is still tightening: TrendForce expects Q3 DRAM contract prices to rise 13–18% quarter-over-quarter, and Samsung is reportedly pushing customers for hikes of up to 20%. What broke isn't demand — it's the multiple. This was a group that had more than doubled since its April launch, and parabolic moves need parabolic news to hold. Absent that, gravity and profit-taking do the rest.

Why the ETF Amplifies the Move

DRAM isn't a diversified semi fund — it's concentrated beta to a handful of memory makers led by Samsung, SK Hynix and Micron. When those three fall together on the same catalyst, there's nothing in the basket to cushion it, which is why the ETF prints double-digit down days that no single chip stock does. The path from here is fairly binary: this is a positioning reset in a still-undersupplied market, so a real bounce likely needs a clean HBM4 print or fresh contract-price confirmation to force the profit-takers to cover. Until then, $DRAM trades as the purest expression of whether the memory trade was early to the top — or just early to unwind.