Semiconductor stocks have come under pressure. Over the past month, the sector tracked by the iShares Semiconductor ETF (SOXX) has dropped roughly 10%, as investors grow cautious about valuations, demand trends, and the sustainability of recent gains.
On March 24, Alphabet’s Google (GOOGL) rolled out TurboQuant, a new compression method that could reduce the amount of memory required to run AI models by 6x, causing a sell-off in memory names, with Micron (MU) and Sandisk (SNDK) down more than 10% in the days following the news.
The recent pullback in those stocks may look like the start of a bigger downturn. But Morgan Stanley says it is not. Here’s what the firm sees.
Memory is now the bottleneck in AI growth
Over the past two years, investors have focused on GPUs from companies like Nvidia as the key driver of AI growth. That is still true. But memory is now the limiting factor.
In a March 26 research note sent to TheStreet, Morgan Stanley argues that in a normal cycle, a selloff reflects investors taking profits on concerns about flat spot pricing, rising capital spending, and productivity gains.
“But this is anything but normal,” the analysts said. “Memory is a bottleneck, increasingly the bottleneck, to AI builds (and agentic CPU builds), which appears durable.”
Getty Images
“We see the recent selloff as a healthy pricing in of durability concerns – capex, demand destruction, productivity, etc.,” the analysts wrote.
They added that the strength of memory companies is “more durable than the market thinks, with memory supply remaining a gating factor for AI.”
Morgan Stanley maintains overweight ratings on memory names Micron and SanDisk, with price targets of $520 and $690, respectively.
What about Google’s TurboQuant?
Morgan Stanley says Google’s TurboQuant does not have the effect investors think it does.
The firm notes the reduction applies only to a specific part of memory, not overall usage. “They are just talking about KV Cache memory, not memory overall,” the firm said.
The KV cache is often held in high-bandwidth memory, whose contents are fixed and cannot change.
Related: Morgan Stanley resets Alphabet stock forecast on Waymo growth
More importantly, the improvement does not reduce the need for memory across AI systems.
“This is an evolutionary development, with basically no surprises for memory,” the analysts said.
They added that improvements like compression have been an ongoing focus across the industry for years.
In fact, better efficiency could increase demand over time. As costs come down, companies can run larger and more advanced models.
More AI Stocks:
- Morgan Stanley sets jaw-dropping Micron price target after event
- Bank of America updates Palantir stock forecast after private meeting
- Morgan Stanley drops eye-popping Broadcom price target
Google itself has tested much larger context windows for its Gemini models but held back in part due to cost constraints.
Morgan Stanley’s takeaway is that TurboQuant is more about enabling capability than cutting demand.
“Result – shortages are intensifying and customers are prepaying for large volume deals given conviction that these shortages will be sustainable,” the analysts wrote.
Earnings will remain strong
Morgan Stanley believes earnings for memory companies could remain elevated for several quarters, supported by strong pricing and limited supply growth.
“As AI evolves, we expect compute architectures to become more memory-intensive,” Micron’s CEO Sanjay Mehrotra said in the latest earnings call. “This is why we strongly believe that Micron is one of the biggest beneficiaries and enablers of AI.”
Related: Jim Cramer drops unexpected take on Microsoft stock
On March 18, Micron reported a record gross margin of 81% for the fiscal second quarter of 2026. Its revenue for the quarter reached $23.86 billion, almost tripled from a year earlier.
“We don’t think 81% gross margin is the new normal, but it’s also hard to see that getting worse in the next several quarters – perhaps even through the next two years,” Morgan Stanley wrote.
Micron stock is up 25% year-to-date despite this week’s tumble, closing at $357 on March 27.
Related: Redfin reveals shift in home prices, housing market
