Google's 'Frozen v2' chip aims to make Gemini 10x more efficient by 2028
Alphabet’s new chip promises 6-10x efficiency boost for Gemini models.
Alphabet, Google's parent company, is designing a new server chip internally dubbed "Frozen v2" to make its Gemini AI models far more efficient. According to a report from The Information, the chip is slated for release around 2028 and could deliver six to ten times better efficiency—measured by tokens generated per watt—than Google’s current AI chips. While Google did not directly confirm the report, a spokesperson told TechCrunch: "Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers." The company emphasized its full-stack approach of co-designing hardware and software for real-world workloads.
The move reflects a broader industry trend where AI giants are creating custom silicon to reduce dependence on Nvidia, which dominates the AI chip market. OpenAI announced its first custom inference chip, "Jalapeño," in June, and Anthropic is reportedly in early talks with Samsung for a chipmaking partnership. The Frozen v2 news came at a critical time for Alphabet, which plans to spend $180–$190 billion on AI infrastructure. The efficiency promise appears to have reassured investors, as Google's stock rose about 3% on Monday ahead of its earnings report.
- Google's Frozen v2 chip is expected by 2028 and claims 6-10x efficiency improvement over current Google AI chips.
- The chip is designed specifically for Gemini models, generating more tokens per watt of power.
- Google stock rose 3% on the news as investors view custom chips as a way to control AI spending and reduce reliance on Nvidia.
Why It Matters
Custom AI chips like Frozen v2 are critical to cutting costs and breaking Nvidia's hardware monopoly.