Google's 'Frozen v2' Chip Could Slash Gemini AI Costs by 2028
Google is reportedly developing 'Frozen v2,' a custom AI chip 6-10x more efficient than TPUs, aimed at boosting Gemini model performance and reducing inference costs, expected by 2028.
Market context
Follow the story. See which markets it affects. Explore the context.
Google is reportedly developing 'Frozen v2,' a custom AI chip 6-10x more efficient than TPUs, aimed at boosting Gemini model performance and reducing inference costs, expected by 2028.