Key Takeaways
- Memory chip stocks including Micron and Sandisk tumbled over 10% following Google’s TurboQuant announcement last month
- Google’s new technology promises to slash AI memory requirements by as much as 6x
- Morgan Stanley views the recent decline as a normal correction rather than fundamental weakness
- Memory chips have emerged as the primary constraint in AI infrastructure expansion, surpassing GPUs
- The investment bank maintains Overweight recommendations on Micron and Sandisk with targets of $520 and $690 respectively
Morgan Stanley remains confident in memory semiconductor stocks despite recent market turbulence that shook investor sentiment in late March.
The iShares Semiconductor ETF experienced approximately 10% decline over the previous month. Multiple factors contributed to the downturn, including valuation worries, demand uncertainties, and emerging AI innovations.
Google introduced its compression technology TurboQuant on March 24. The breakthrough promises to slash memory requirements for AI model operation by as much as six-fold. The announcement triggered investor anxiety.
Both Micron and Sandisk experienced declines exceeding 10% in the subsequent trading sessions. Micron’s shares settled at $357 on March 27, though the stock maintained a 25% gain for the year.
Morgan Stanley’s Joseph Moore challenged the negative sentiment in a research report issued March 26.
Moore confirmed Overweight ratings for both Micron and Sandisk. The firm’s price objectives remain unchanged at $520 and $690.
According to Moore, the decline represents “a healthy pricing in of durability concerns” instead of genuine demand deterioration. The financial institution maintains that memory manufacturers’ fundamentals are “more durable than the market thinks.”
Memory Chips Eclipse GPUs as Primary AI Constraint
Over the previous 24 months, Nvidia’s graphics processing units dominated headlines as the critical component driving AI infrastructure investment. While this remains accurate, Morgan Stanley identifies memory as the new limiting element.
“Memory is a bottleneck, increasingly the bottleneck, to AI builds,” the research team stated. They observed customers now making advance payments for high-volume orders, indicating severe supply constraints.
DRAM surplus has completely evaporated, Moore noted. “Everywhere we look we see indications that it is a true bottleneck,” he explained.
AI could represent “well north of 50%” of total semiconductor expenditure, the bank projected. Production increases will struggle to match such robust demand levels.
Morgan Stanley’s Assessment of TurboQuant Impact
Morgan Stanley directly evaluated Google’s TurboQuant announcement, arguing investors misinterpreted its significance.
The compression approach applies exclusively to KV Cache memory rather than total memory consumption. “They are just talking about KV Cache memory, not memory overall,” the analysts clarified.
KV Cache utilizes high-bandwidth memory, a specialized and constrained category. Morgan Stanley characterized TurboQuant as “normal course productivity improvement” rather than a demand-threatening innovation.
The bank acknowledges that gross margins approaching 81% won’t persist indefinitely. However, it identifies few catalysts for near-term margin compression.
Morgan Stanley emphasized robust free cash flow prospects for memory sector companies. The firm stressed that “duration is all that matters,” and relevant metrics “all appear positive.”
Micron and Sandisk retained their Overweight ratings as of March 26, 2026.
