After two consecutive days of sharp declines, the storage sector began to

Li Jia

Source: Wall Street News

Google’s TurboQuant Algorithm Breakdown Sparks Split in Memory Chip Sector: Flash Memory Stocks Under Continuous Pressure, Kioxia and Others Give Back Large Earlier Gains; HBM Suppliers Samsung and SK Hynix Stabilize Quickly. Morgan Stanley points out that the technology mainly optimizes memory efficiency in the AI inference stage, with limited impact on core HBM demand, while posing a greater threat to NAND flash memory. Bloomberg Intelligence believes memory optimization in inference hits NAND flash harder, whereas core memory demand for GPUs (HBM/DRAM) remains largely unaffected.

A Google algorithm breakthrough boosting AI inference efficiency is triggering a structural divergence in the memory chip industry: flash memory companies face sustained stock price pressure, while high-bandwidth memory (HBM)-related names have held relatively steady.

U.S.-listed memory chip stocks declined for two consecutive sessions, with SanDisk plummeting 11% at Thursday’s close. On Friday, selling pressure spread to Asian markets, with Korean memory stocks extending losses: SK Hynix fell more than 5% intraday, and Samsung Electronics dropped over 4%.

However, supported by the central role of HBM in AI training, Samsung Electronics and SK Hynix — key HBM suppliers for NVIDIA’s AI accelerators — stabilized rapidly. Samsung nearly fully recovered its losses, and SK Hynix narrowed its decline to 1%, showing greater resilience than flash memory makers. In contrast, stocks such as Kioxia, which had surged more than 600% over recent months, continued to slide.

Morgan Stanley analyst Tiffany Yeh noted in a research report that Google’s “TurboQuant” technology significantly improves AI inference efficiency by compressing memory footprint and data movement, but “will not weaken demand for core memory chips such as HBM.” The market is gradually realizing that the technology poses a far greater threat to flash memory firms than to the HBM sector.

### Flash Memory Sector Bears Brunt, Earlier Gains Sharply Erased

Driven by expectations of widespread AI adoption, flash memory and storage product manufacturers attracted heavy investor inflows in recent months. Since late August, SanDisk’s stock had rallied more than 1,000% at one point, while Kioxia jumped over 600%, outperforming traditional memory giants including Samsung Electronics, SK Hynix, and Micron Technology by a wide margin.

However, market sentiment reversed this week. After recognizing the far-reaching impact of Google’s technological breakthrough, investors sold off shares of these companies first. Google announced that its TurboQuant algorithm can reduce the memory required for specific stages of large language model operations by at least five-sixths, significantly lowering the overall operating cost of AI systems. Markets fear this will reduce memory procurement demand from hyperscale data center operators such as Meta, in turn dragging down memory chip prices in the smartphone and consumer electronics sectors.

Bloomberg Intelligence analyst Jake Silverman wrote: “Since model weights are stored in GPU memory, demand for HBM and DRAM produced by Micron is likely unaffected. In contrast, NAND flash demand will face deeper long-term pressure.”

### HBM Demand Logic Intact, Samsung and SK Hynix Stabilize

In contrast to the ongoing decline in the flash memory sector, HBM-related stocks have shown strong resilience during this selloff. Samsung Electronics fully recouped its losses on Friday, and SK Hynix’s share price returned roughly to pre-selloff levels.

Analysts say the divergence is supported by clear logic. During the training phase of large AI language models, GPU demand for HBM is highly concentrated, and TurboQuant optimizes memory efficiency only in the inference stage, leaving core training-side HBM demand untouched. Samsung and SK Hynix became market favorites early in the AI investment boom thanks to their HBM offerings, and this dominant position has not been shaken by the latest algorithm breakthrough.

### Analysts: Short-Term Volatility Masks Long-Term Industry Narrative

Notably, the memory chip selloff comes amid a broader macro backdrop where valuations of technology stocks are under scrutiny. Inflation concerns stoked by Middle East tensions have made markets cautious toward high-valuation stocks, with investors highly sensitive to news flow and profit-taking likely to trigger swings at any time.

Ed Gomes, Chief Investment Officer at SGMC Capital, stated that hardware demand driving AI deployment and adoption represents a long-term structural theme that “will evolve over years or decades, not days or weeks.” He views the TurboQuant-driven selloff as “short-term noise that presents a good buying opportunity for high-quality names.”

Still, analysts warn that as AI efficiency algorithms continue to iterate, the split within the memory chip industry may intensify further. Whether flash memory companies can revive their earlier high-growth expectations remains to be seen.

---

## Risk Warning and Disclaimer

The market is subject to risks; investment involves caution. This article does not constitute personal investment advice and does not account for the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Any investment based on this article is at your own risk.