Alphabet Inc. Class C's algorithm optimization does not change the trend of memory requirements. Morgan Stanley throws its weight behind storage stocks and reiterates its "buy" rating for Micron Technology, Inc. (MU.US) and SanDisk (SNDK.US).

date
23:15 26/03/2026
avatar
GMT Eight
The latest research report from Daiwa Securities pointed out that although the TurboQuant algorithm launched by Google can reduce some memory usage, it is more likely to drive the evolution of AI applications towards higher computing power requirements, rather than weakening the overall demand for storage and memory.
Market divergence is intensifying around the impact of Alphabet Inc. Class C's new algorithm on the storage industry chain. In its latest research report, Morgan Stanley pointed out that while Alphabet Inc. Class C (GOOGL.US, GOOG.US) launched the TurboQuant algorithm, which can reduce some memory usage, it is more likely to drive the evolution of AI applications towards higher computational power requirements, rather than weaken the overall demand for storage and memory. Morgan Stanley analyst Joseph Moore stated in the report that the widely held belief that "memory usage has been reduced by 6 times" is exaggerated, as this optimization mainly targets KV Cache memory, rather than the overall memory requirement. He noted, "Stocks in the storage sector have fallen recently, in part due to this overinterpretation." The report further mentioned that Alphabet Inc. Class C's latest models, Gemini 3 and 2.5 Pro, have a context window of 1 million tokens, while the earlier Gemini 1.5 Pro in testing could even support up to 10 million tokens, but was not released to the public due to high reasoning costs. Analysts believe that with technologies such as TurboQuant reducing costs, future AI models will tend to use larger-scale contexts and more complex calculations, thereby driving an increase in overall computational power and storage demand. Based on this assessment, Morgan Stanley reiterated its "hold" rating for Micron Technology, Inc. (MU.US) and SanDisk (SNDK.US). Analysts pointed out that the core bottleneck facing AI development currently is still insufficient memory supply, especially in the DRAM sector, where rapid growth in data center demand is squeezing memory supply for end products such as PCs and smartphones. Moore emphasized, "From our industry chain research, there is no evidence to suggest that memory or storage demand is decreasing."