Google's TurboQuant AI breakthrough sends memory stocks tumbling on reduced storage demand

Memory Stocks Slide As Google's New AI Efficiency Breakthrough May Slash Data Storage Needs - SanDisk (NASDAQ:SNDK)

Google Unveils TurboQuant Algorithm On Tuesday, Google researchers introduced "TurboQuant." This set of advanced quantization algorithms enables massive compression for large language models (LLMs). According to the Google blog, the technology "optimally addresses the challenge of memory overhead in vector quantization." The breakthrough focuses on the key-value (KV) cache. Google describes this as a "digital cheat sheet" that stores frequently used information. High-dimensional vectors capture complex AI data but consume vast amounts of memory. Google reported that TurboQuant reduces key-value memory size "by a factor of at least 6x" without sacrificing model accuracy. Enhanced Efficiency For LLMs Google stated these techniques allow "building and querying large vector indices with minimal memory." Price Action: SanDisk shares were down 3.17% at $649.20, Western Digital shares were down 1.29% at $291.80, and Micron Technology shares were down 3.46% at $381.86 at the time of publication on Wednesday, according to Benzinga Pro data. Photo: Nor Gal / Shutterstock This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

MU, WDC, SNDK fall: Why Google's TurboQuant is rattling memory stocks By Investing.com

Investing.com -- Memory stocks fell Wednesday despite broader technology sector strength, with shares dropping after Google unveiled TurboQuant, a new compression algorithm that could reduce memory requirements for AI systems. SanDisk Corporation (NASDAQ:SNDK) fell 5.7%, Micron Technology (NASDAQ:MU) dropped 3%, Western Digital (NASDAQ:WDC) declined 4.7%, and Seagate Technology (NASDAQ:STX) slid 4%. The declines came as the Nasdaq 100 advanced. Google introduced TurboQuant, a compression technology designed to reduce memory consumption in large language models and vector search engines. The algorithm addresses bottlenecks in key-value cache, which stores frequently accessed information in AI systems. According to Google's announcement, TurboQuant can compress key-value cache to 3 bits without requiring training or fine-tuning while maintaining model accuracy. Testing on open-source models including Gemma and Mistral showed the technology achieved a 6x reduction in key-value memory size. The algorithm also demonstrated up to 8x performance increase over unquantized keys on H100 GPU accelerators. The technology works through two steps: applying the PolarQuant method for high-quality compression by rotating data vectors, and using the Quantized Johnson-Lindenstrauss algorithm to eliminate residual errors. Google said traditional vector quantization methods add 1 to 2 extra bits per number in memory overhead, partially negating compression benefits. TurboQuant will be presented at ICLR 2026, while PolarQuant is scheduled for presentation at AISTATS 2026. Google tested the algorithms across benchmarks including LongBench, Needle In A Haystack, ZeroSCROLLS, RULER, and L-Eval. The technology has applications beyond AI models, including vector search capabilities that power large-scale search engines. Memory stocks have rallied significantly year to date, making them vulnerable to developments that could reduce demand.