NewsBite analysis · AI, Hardware
The Economic Times
Google plans new chip to run Gemini models more efficiently: Report
Google is in the process of creating a new server chip informally named 'Frozen v2' that will integrate components of its Gemini AI model directly into hardware. The move aims to resolve an ongoing AI computing capacity crunch which has caused strife within the company and restricted Google Cloud's ability to accept new customer contracts. The projected efficiency gains of the chip, reportedly up to six to ten times more than existing models, would enable better utilization of power and resource management for AI applications. Google anticipates deploying the chip as early as 2028 as development continues.
MonitorConfidence: HighLong term horizon
Published 20 Jul 2026, 14:30 UTC3 min read
Visit original article ↗Lead image — The Economic Times
Summary
TL;DR
- New chip 'Frozen v2' aims to enhance Gemini AI efficiency.
- Projected launch slated for 2028.
- Could be 6-10 times more efficient than current chips.
- Targets internal tensions and computing capacity issues.
- Integration of the Gemini model directly into the hardware.
Interesting facts
- The chip will not replace existing tensor processing units (TPUs).
- Shares of Alphabet rose 3% following the news.
- Google Cloud's restrictions on external contracts heightened the need for this development.