Google's Frozen v2 Chip Boosts Efficiency for Gemini Models
Alphabet's stock gained more than 3% on Monday after reports emerged that Google is developing a new server chip, called Frozen v2, designed to run Gemini models efficiently.
Frozen v2 is reportedly six to ten times more power efficient than Google's latest Tensor Processing Units (TPUs) by embedding parts of Gemini directly into the hardware. Unlike TPUs, which are built for general-purpose AI tasks, Frozen v2 is being designed specifically for Gemini and would reduce computation and data movement needed to generate AI responses.
The new chip family is seen as a complementary addition rather than a replacement for Google's existing TPU lineup. The project aims to reduce AI infrastructure costs and lessen the company's reliance on Nvidia's GPUs.