Archive · 21 Jul 2026
Google Developing Frozen v2 AI Chip
Google is working on a new AI chip called Frozen v2.
This chip is special because it's designed to make Gemini, Google's AI model, work much faster and use less electricity.
Unlike other AI chips that can do many different tasks, Frozen v2 is made just for Gemini.
This means it can answer more questions while using less power, which is a big deal because AI can be very energy-hungry.
However, because it's so specific to Gemini, if Google changes Gemini a lot in the future, this chip might not be as useful.
Still, investors are excited because it could help Google keep up with the growing demand for AI.
Google is developing a new AI chip called Frozen v2, designed specifically for Gemini AI models.
Frozen v2 is reported to be 6-10 times more efficient than current TPUs in AI tokens per unit of power.
The chip embeds parts of Gemini's architecture into the silicon to reduce unnecessary calculations and energy use.
Google views Frozen v2 as an experimental project due to its specialization, which may limit its usefulness for future AI model changes.
The news of Frozen v2 sent Alphabet shares up about 3% on Monday, reflecting investor excitement.
- Who
- Google (Alphabet Inc.)
- What
- Development of a new AI chip called Frozen v2
- When
- Currently in development, expected to start manufacturing in 2028
- Where
- Internally at Google
- Why
- To improve the efficiency and performance of Gemini AI models, addressing the growing demand for AI computing power and reducing energy consumption
Proponents of Specialized AI Chips
Advocates for General-Purpose AI Chips
Specialization vs. Flexibility
Proponents of Specialized AI Chips
Specialized chips like Frozen v2 offer significant efficiency gains for specific AI models, reducing power consumption and improving performance.
Advocates for General-Purpose AI Chips
General-purpose chips provide flexibility, allowing them to adapt to various AI models and future advancements without becoming obsolete.
- Chip Name
- Frozen v2
- Purpose
- Optimize Gemini AI models
- Efficiency Gain
- 6-10x more AI tokens per unit of power
- Current Status
- Experimental project
- Impact on Shares
- Alphabet shares up 3% on Monday
- Primary Use Case
- Gemini AI models
- Key Benefit
- Reduced energy consumption
- Trade-off
- Limited flexibility for future AI model changes
- Expected Manufacturing Start
- 2028
Sources
Google's secret 'Frozen v2' AI chip could make Gemini up to 10 times more efficient, report says
Google plans new AI server chip ‘Frozen v2’ to boost Gemini efficiency: Report
Google plans new AI Chip to improve Gemini AI performance: All details
Google Working On New AI Chip To Run Gemini Models More Efficiently
