Google Said to Be Developing 'Frozen' Chip That Bakes Gemini Architecture Into Silicon

Google is developing a new server chip, known internally as "Frozen v2," that would hardwire parts of its Gemini AI models directly into silicon to make inference dramatically more efficient, according to a report from The Information. The chip is positioned as a specialized line separate from Google's existing Tensor Processing Units and is not expected to reach deployment until around 2028.