Stock Markets March 25, 2026

Analyst Says Google-Linked Dip in Memory Stocks Is a Buying Opportunity

Market reaction to Google's TurboQuant compression tech sparks sell-off in memory and storage names despite analyst pushback

By Jordan Park SNDK MU WDC STX
Analyst Says Google-Linked Dip in Memory Stocks Is a Buying Opportunity
SNDK MU WDC STX

Shares of memory and storage companies tumbled after Google unveiled TurboQuant, a 3-bit compression approach for AI models. Analysts at Lynx Equity Strategies argued the announcement has been overstated by some commentary, maintained a bullish view on Micron and said the technology does not eliminate demand for DRAM and flash over the next three to five years.

Key Points

  • Google announced TurboQuant, a 3-bit compression technique for AI models, prompting declines in memory and storage stocks.
  • Lynx Equity Strategies' KC Rajkumar argued vector quantization is not novel and disputed claims of an 8x improvement, noting comparisons were made to older 32-bit models rather than current 4-bit inference models.
  • Lynx maintained a $700 price target and a buy rating on Micron, recommending purchases into the pullback and asserting that advanced compression will reduce bottlenecks but not eliminate demand for DRAM and flash over the next three to five years.

Memory and storage stocks slid Wednesday after Google introduced TurboQuant, a new compression algorithm meant to lower memory needs for AI systems. The move came even as the broader technology-heavy Nasdaq 100 posted gains, highlighting a sector-specific market reaction.

The pullback saw SanDisk Corporation (NASDAQ:SNDK) fall 5%, Micron Technology (NASDAQ:MU) lose 4%, Western Digital (NASDAQ:WDC) decline 3.7% and Seagate Technology (NASDAQ:STX) slide 4% on the session.

Google described TurboQuant as a 3-bit compression technique for AI models. In response, Lynx Equity Strategies analyst KC Rajkumar pushed back on the narrative around the announcement in a client note, saying the concept was not as novel as some third-party technology blogs have portrayed.

"Vector quantization is hardly a new idea as LLM models went from 32-bit data in original training models to 4-bit data used in inference models today," Rajkumar wrote. He faulted coverage that suggested an 8x improvement, noting that such comparisons were being made to older 32-bit models rather than the 4-bit quantized inference models now commonly used.

"While there may be an improvement, it is not 8x, in our view," the analyst added.

Rajkumar acknowledged that AI infrastructure providers must continue to innovate to address scaling challenges as token context length expands during inferencing. Still, he argued that these advances in compression do not translate into a near-term collapse in demand for memory and flash, citing supply constraints as a key factor.

"Advanced compression techniques merely reduce bottlenecks without destroying demand for dram/flash," Rajkumar wrote, emphasizing that memory and flash requirements remain tied to broader capacity and supply dynamics over the coming three to five years.

Lynx Equity Strategies left its $700 price target on Micron intact and reiterated a buy recommendation for the stock. The analyst concluded, "We would be buyers into the Google-related pullback today."

The session's coverage also included reference to ProPicks AI, an evaluation tool that analyzes companies including Micron across numerous financial metrics. The original copy cited ProPicks AI's methodology and cited past winners identified by the tool, including Super Micro Computer (+185%) and AppLovin (+157%), as examples of its track record.


Where this matters

  • Semiconductor and storage sectors are directly affected by developments in model compression and memory efficiency.
  • AI infrastructure providers and cloud operators face incentives to optimize memory usage as model context lengths grow.
  • Investor sentiment in technology sub-sectors can diverge from broader market performance, as seen with the Nasdaq 100 advancing while memory names fell.

Risks

  • Market volatility tied to technical announcements - short-term price moves in semiconductor and storage names can be driven by interpretations of AI infrastructure innovations.
  • Misleading comparisons in public commentary - overstated performance claims (for example, an 8x improvement versus older 32-bit models) can foster investor confusion and trigger unwarranted sell-offs in affected sectors.
  • Supply constraints in memory and flash - persistent supply dynamics could sustain demand even as compression techniques improve, creating uncertainty for capacity planning in tech and cloud infrastructure sectors.

More from Stock Markets

Denali Shares Rise After FDA Grants Accelerated Approval to AVLAYAH for Hunter Syndrome Mar 25, 2026 Belgian Equities Advance as Technology, Materials and Healthcare Drive Gains Mar 25, 2026 Paris Markets Close Strong; CAC 40 Gains 1.33% on Broad Sector Strength Mar 25, 2026 Frankfurt closes higher as tech, chemicals and industrials lead gains - DAX up 1.34% Mar 25, 2026 Helsinki Stocks Close Higher as Telecoms, Utilities and Industrials Lead Gains Mar 25, 2026