Google’s Frozen v2 chip set to supercharge Gemini AI models

Google is developing the Frozen v2 AI chip, optimized for its Gemini models, promising significant efficiency and performance improvements.

The new Frozen v2 chip aims to deliver 6-10 times better performance per watt compared to existing chips.
Custom designs will eliminate unnecessary components, optimizing costs and efficiency for AI workloads.
Google plans to roll out the Frozen v2 chip to its data centers by 2028.