Alphabet’s ‘Frozen v2’ Chip: A Potential Game-Changer for Gemini AI Efficiency
Alphabet, Google’s parent company, is in the process of developing a revolutionary server chip. This chip could significantly decrease the energy cost of running its Gemini AI models. The news of this development led to a surge in the company’s stock, with a rise of as much as 3% on Monday.
As reported by The Information, the chip, internally known as “Frozen v2”, incorporates elements of Google’s Gemini AI architecture directly into the silicon. This approach differs from the traditional reliance on general-purpose processors. The innovative design eliminates much of the processing overhead associated with conventional GPUs and TPUs. This reduction leads to fewer computational steps and less data movement needed to respond to user queries. Google engineers anticipate that the chip could yield six to ten times more AI tokens per unit of power than the company’s current custom TPUs.
The deployment of Frozen v2 is planned for as early as 2028. It is designed to be a distinct, specialized product line that complements Google’s existing Tensor Processing Units, rather than replacing them. The project also aims to alleviate a significant internal computing capacity crunch. This crunch has reportedly led Google Cloud to turn away enterprise customers. Alphabet plans to invest between $180 billion and $190 billion in AI infrastructure. Investors view the more efficient chip as a promising sign that these substantial investments could yield significant returns. Google has neither confirmed nor denied the report.
Source: TechCrunch — Google is working on a new AI chip designed to make Gemini more efficient