Google working on 'Frozen v2' chip to make Gemini faster
What's the story
Google's parent company, Alphabet, is said to be developing a new server chip, internally dubbed "Frozen v2." The tech giant hopes the chip will make its in-house Gemini models run more efficiently. The development of this next-generation chip was first reported by The Information. It is expected to launch sometime in 2028 and could be six to 10 times more efficient than Google's current AI chips.
Efficiency boost
Chip to integrate parts of Gemini's architecture
The Frozen v2 chip is being specially built for the Gemini AI. It promises to improve speed and cut down on computing power needed to answer user queries.
The design would integrate parts of Gemini's AI architecture directly into the chip, allowing it to perform many tasks without repeating calculations.
This would make Gemini faster and more energy efficient by reducing both calculations and data movement within the system.
Company statement
Google responds to Frozen v2 report
Responding to TechCrunch's inquiry about the Frozen v2 chip, Google didn't confirm or deny the report.
The company said, "Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers."
It added that while not every project moves into production, this exploration is central to its full-stack approach.
Future plans
Chip won't replace Google's TPUs
Google engineers believe the Frozen v2 chip could generate six to 10 times more AI tokens with the same amount of electricity compared to Google's latest Tensor Processing Units (TPUs).
However, it won't replace Google's TPUs but rather become a specialized chip in its custom AI hardware lineup.
The company hopes to deploy this new chip by 2028, mainly to solve its internal shortage of AI computing power.