Analysts and researchers say Google's TurboQuant compression algorithm to make LLMs more efficient is more likely to expand memory chip demand than reduce it (Daniel Tudor/Financial Times)