Alphabet Inc. Class C (GOOG.US, GOOGL.US) released a new algorithm aimed at reducing memory usage, causing storage concept stocks to fall in response. SanDisk (SNDK.US) dropped by 8% at one point.

date
22:35 25/03/2026
avatar
GMT Eight
Google recently released a series of new algorithms aimed at significantly reducing the memory consumption required for large language models and vector search engines to run. This move has put collective pressure on the stock prices of the storage industry chain.
As artificial intelligence technology continues to advance, computing power and storage requirements have become a focus of the industry. Alphabet Inc. Class C (GOOG.US, GOOGL.US) recently released a series of new algorithms aimed at significantly reducing the memory usage required for large language models (LLM) and vector search engines, causing pressure on stock prices in the storage industry chain. The core technologies introduced by Alphabet Inc. Class C this time include TurboQuant, Quantized Johnson-Lindenstrauss Transform, and PolarQuant. Among them, TurboQuant optimizes the compression efficiency in the vector quantization process to address the memory burden brought by processing high-dimensional data; the Johnson-Lindenstrauss Transform significantly reduces the data size while maintaining data relationships unchanged through mathematical dimensionality reduction techniques, achieving efficient representation with "zero additional memory overhead"; PolarQuant utilizes polar coordinate mapping to project data onto a fixed structure, skipping the traditional data normalization steps and further reducing resource consumption. Alphabet Inc. Class C research scientists Amir Zandieh and Vice President and Senior Researcher Vahab Mirrokni stated that with the widespread application of AI in scenarios such as large language models and semantic searches, the importance of underlying vector quantization technology will continue to increase. These algorithms are expected to fundamentally improve the operational efficiency of AI systems and reduce dependence on high-end storage hardware. In response to this news, stocks of storage and hard drive-related companies fell in early trading on Wednesday. Micron Technology, Inc. (MU.US) fell by about 4%, Western Digital Corporation (WDC.US) fell by about 4.4%, Seagate Technology Holdings PLC (STX.US) fell by about 5.6%, and SanDisk (SNDK.US) fell by 8% at one point. Market concerns arise that a decrease in AI model reliance on memory may weaken long-term demand for storage chips and hard drive products. At the same time, SanDisk announced that it had reached a private subscription agreement with semiconductor company Nanya Technology and will be making equity investments, showing that the industry is accelerating integration and layout internally. Alphabet Inc. Class C stated that the results related to TurboQuant will be officially released at the International Conference on Learning Representations in Rio de Janeiro in April this year. Analysts believe that if the relevant technologies are widely applied, they may have a profound impact on the structure of AI hardware demand, reshaping the growth logic of the storage industry.