Google researchers introduced TurboQuant, a compression algorithm that reduces memory requirements for large language models by at least sixfold while delivering up to eight times faster inference. The announcement sparked profit-taking in global memory and storage stocks, with SK Hynix and Kioxia each tumbling as much as 6.4%, though analysts argue efficiency gains may ultimately boost demand under the Jevons paradox.

Broadcom has secured OpenAI as a custom AI accelerator customer and projects AI semiconductor revenue will double to $8.2 billion in 2026. This deal signals a shift toward bespoke AI hardware solutions that could influence Alphabet’s future infrastructure investments and highlight the competitive pressure from non-Nvidia chip suppliers.