Google announces revolutionary Frozen v2 chip: Gemini architecture embedded directly into silicon
Tech giant Google is preparing to release a new-generation specialized processor — Frozen v2. This chip marks a fundamentally different approach to hardware acceleration of artificial intelligence. Instead of relying solely on software optimization, Google engineers decided to embed key architectural elements of their flagship Gemini model directly into the silicon substrate.
How it works and why it is needed
Integrating the architecture at the transistor level allows for a radical reduction in the volume of data moved and the number of computational operations required to process user requests. According to preliminary estimates, Frozen v2 will be able to process six to ten times more tokens per unit of energy consumed compared to current generations of Google's tensor processing units (TPUs). This is not a replacement for TPUs, but rather a highly specialized addition to them, designed to handle the most resource-intensive tasks.
Strategic context and timeline
The project is launching amid an acute shortage of computing power, which, according to my data, has forced Google Cloud to reject some contracts with external clients. Obviously, the company is betting on vertical integration to bypass the limitations of traditional chips. However, there is an important nuance: Frozen v2 will only work effectively with future versions of Gemini, provided the model's basic architecture remains unchanged. For now, the project is considered experimental, and mass production on the scale of universal TPUs is not planned. The chip is expected to be deployed in 2028.
Market reaction and industry challenges
The news about Frozen v2 has spurred investor interest: on July 20, Alphabet (GOOG) shares rose by 1.5% on the Nasdaq. However, behind this announcement lie serious problems. The launch of Gemini 3.5 Pro is delayed, and Google has lost four leading researchers who moved to competitors at Anthropic and OpenAI. Meanwhile, Chinese AI models are steadily capturing market share: they already account for up to 46% of all tokens processed by American companies.
My expert commentary: Frozen v2 is a bold but risky move. If Google manages to maintain architectural continuity with Gemini, such a chip could give the company a tremendous advantage in energy efficiency and speed. However, dependence on a single architecture and delays in new model releases cast doubt on the project's long-term viability. The AI market does not tolerate stagnation, and 2028 is a very long way off by this industry's standards.