Google embeds Gemini architecture into silicon: details about the Frozen v2 chip

Google is preparing to release a specialized processor, Frozen v2, which will be a logical continuation of its tensor processing unit (TPU) line, but with a fundamentally new approach. Unlike universal accelerators, this chip will embed key elements of the Gemini architecture directly into the silicon substrate. This integration will radically reduce the number of computational operations and the volume of data moved when processing user requests.
The project is positioned not as a replacement for TPUs, but as a highly specialized complement to them. The main goal is to alleviate the acute shortage of computing power, which, according to my data, has forced Google Cloud to turn down several deals with external clients. Company engineers predict that Frozen v2 will be able to process six to ten times more tokens per unit of energy consumed compared to Google's latest AI accelerators. This is a colossal leap in efficiency.
However, there is an important nuance: the chip will only be compatible with future versions of Gemini models, provided the basic architecture is preserved. For now, Google views Frozen v2 as an experimental project, so mass production on the same scale as TPUs is not planned. The chip is expected to be deployed in 2028. Notably, on July 20, after this information was published, Alphabet's shares rose by 1.5%.
Against the backdrop of these ambitious plans, Google's AI division is facing serious difficulties. The launch of Gemini 3.5 Pro is delayed, and the company has lost four leading researchers who moved to competitors. Moreover, Chinese models already account for up to 46% of tokens used by American companies, adding further pressure.
Expert opinion: Frozen v2 is a strategic move by Google aimed at creating its own "hardware" advantage in the AI race. However, the delays in release until 2028 look risky: by that time, competitors may offer equally efficient solutions, and current issues with talent and Chinese models will only worsen the situation. The project's success will depend on whether Google can maintain technological leadership amid growing competition.