Google's new memory breakthrough is moving chip markets - and could transform AI

TurboQuant reduces the working memory needed for vector quantisation without sacrificing accuracy, offering a material reduction in inference costs.