Alphabet's new server chip, internally named "Frozen v2," is projected to be 6 to 10 times more efficient than Google's current AI chips, measured by tokens generated per unit of power. The chip is slated for release in 2028, according to anonymous sources cited by The Information. This development follows Google's planned AI expenditures of $180 billion to $190 billion, with the news reportedly assuaging investors and boosting the company's stock by 3%.
Hyperscalers are aggressively pursuing custom AI silicon to reduce NVIDIA dependence and improve inference efficiency, shifting compute demand towards internal design and away from merchant silicon.