Google is developing a new server chip, codenamed Frozen v2, that integrates part of its Gemini AI architecture directly into hardware. The company projects the chip will achieve a 6 to 10 times efficiency gain over current designs.
The Frozen v2 Project
Frozen v2 is a custom chip built for data-center workloads. By embedding Gemini's architecture into silicon, Google aims to speed up AI processing while cutting the energy needed per task. The exact design details remain under wraps, but the projected efficiency leap is significant.
Why Efficiency Matters for AI
Running large AI models consumes enormous amounts of electricity and compute time. A chip that delivers 6–10x better efficiency could slash operational costs and reduce the environmental footprint of AI services. Google hasn't said when Frozen v2 will ship or which products will use it first.
The company has not announced a release date or further specifications. For now, the Frozen v2 project signals Google's push to build more specialized hardware for its AI ambitions.
