Google plans to release a specialized chip called Frozen v2, according to The Information.
This processor will integrate elements of the Gemini architecture directly into its silicon structure, reducing the number of computations and the amount of data transferred needed to process user requests.
According to The Information, the project is intended as a complement to the tensor processing unit (TPU) lineup, rather than a replacement. The chip is also expected to help alleviate the shortage of computing power, which has reportedly led Google Cloud to turn down some deals with external clients.
Google engineers estimate that the device will be able to process six to ten times more tokens per unit of energy compared to the company's latest AI accelerators.
Media reports indicate that the chip will only work with future versions of the Gemini models if Google maintains the core architecture. Currently, the company views Frozen v2 partly as an experimental project and does not plan to produce it in the same volumes as its general-purpose TPUs. The tech giant aims to deploy the chip by 2028.
According to TradingView, on July 20, following the publication of this news, Alphabet's shares rose by 1.5%.
Hourly chart of Alphabet's shares on the Nasdaq. Source: TradingView.In recent months, Google's AI division has faced challenges. The launch of the next release, Gemini 3.5 Pro, has been delayed, compounded by the loss of four leading researchers who left for competitors, while Chinese models are gaining traction among American companies. These models account for up to 46% of tokens used by U.S. companies.
As a reminder, in July, the head of DeepMind suggested that advanced AI models should be tested before release.