The impact of artificial intelligence (AI) is profound, completely reshaping the technological landscape we know. This new generation of algorithms promises to simplify various time-consuming tasks, thereby significantly improving productivity. With the global race to adopt AI technology, demand for data centers is exploding, and foundries are rapidly mass-producing semiconductors to support AI operations. High-bandwidth memory (HBM), DRAM, and NAND flash memory chips play an indispensable role in AI data processing, but even with continued increases in chip production, demand remains high, and prices remain stubbornly high.

Alphabet's Google recently announced a major breakthrough in compression technology, which promises to further improve the efficiency of artificial intelligence models, thereby reducing the demand for certain scarce memory chips. This news may have a potential negative impact on Micron Technology and SanDisk.

Algorithm Breakthrough: TurboQuant Reshapes AI Efficiency

In a blog post last week, Google officially announced that its research team had successfully developed an AI memory compression algorithm called TurboQuant. Google scientists stated in the related research paper: "We introduced a series of advanced, theoretically sound quantization algorithms capable of large-scale compression of large language models and vector search engines."

The most impressive advantage of this algorithm is that its integrated application can reduce the memory usage of AI models by at least 6 times, while increasing the running speed by up to 8 times, and maintaining zero loss of accuracy. This achievement completely redefines the operational efficiency of artificial intelligence technology. In other words, with this algorithm, the number of memory chips required can be reduced by 83%.

In an article published on the X platform, Cloudflare co-founder and CEO Matthew Prince stated, "This is Google's DeepSeek." He added, "There's still significant room for optimization in AI inference in terms of speed, memory usage, power consumption, and multi-tenant utilization."

Industry Impact: NAND Hit Hardest, Micron and SanDisk Under Pressure

Preliminary analysis indicates that the launch of Google's new algorithm will not have an equal impact on all types of memory. NAND flash memory will be the most directly affected, while DRAM and HBM will be largely unaffected. This situation is particularly unfavorable for SanDisk, as almost all of its revenue comes from its NAND flash memory business; in contrast, Micron Technology is much less dependent on the NAND market, with approximately 21% of its second-quarter revenue coming from flash memory.

Previously, the severe shortage of memory chips had been driving up market prices, a phenomenon also mentioned by Micron Technology CFO Mark Murphy in the company's second-quarter earnings report released earlier this month. He stated, "DRAM...prices have risen by about 60%, mainly due to tight supply in the industry...NAND...prices have risen by over 70%, mainly due to tight supply in the NAND industry."

Industry insiders believe that if Google's TurboQuant algorithm ultimately achieves its intended application effect, it will effectively reduce market demand for certain types of memory chips. This decrease in demand will eventually lead to a drop in the prices of related memory chips, thereby impacting the sales of Micron Technology and SanDisk's products.

Turning Point: Memory Price Drops May Stimulate Demand

However, this change is not entirely bad news. Some commentators have pointed out that the decline in memory chip prices may actually stimulate market demand growth. This is because lower memory costs will further reduce the application costs of artificial intelligence technology, which will encourage more companies to adopt AI technology, thereby increasing the overall usage and market demand for memory chips.

Currently, the ultimate direction of these changes remains to be seen.