Embedding the model directly into the chip, allowing silicon to save computing power and data transfer - Google is quietly refining a new server chip codenamed Frozen v2 along this path. According to multiple media reports, the chip aims to run the Gemini model more efficiently, with reports indicating that part of the Gemini architecture will be permanently embedded in the chip, thereby reducing the amount of computation and data movement required when answering user questions. Google engineers expect that, compared to its latest generation general-purpose AI chip, the TPU (Tensor Processing Unit), the Frozen chip can provide 6 to 10 times more token processing capability per unit of power.
The release of this chip is scheduled for 2028, but Google has clearly stated that it will not replace general-purpose TPUs, instead becoming a more specialized and targeted branch in its custom chip product line. Following the announcement, the capital market reacted quickly: Google's Class A shares (GOOGL) rose nearly 3.7% in early trading, while Class C shares (GOOG) saw a peak increase of about 3.9%.
In response to media inquiries, Google stated in a statement that the company's team continues to research and test various innovative technologies, aiming to provide users and customers with the best performance and highest efficiency. It also emphasized that although not all projects will go into mass production, this rigorous exploration is a core component of its full-stack strategy. Google further said that by co-designing hardware and software at the fundamental level, the entire system is deeply integrated and highly optimized for real-world workloads. However, Google currently positions Frozen v2 as an experimental project and has no plan to mass-produce it like the TPU.
Analysts believe that behind this project is the increasingly tight computing power situation within Google. Previously, insufficient computing power not only intensified internal resource competition, but reportedly even forced Google Cloud to reject some external customer business. Just last month, Google agreed to pay nearly $1 billion per month to SpaceX to fill the computing power gap and fulfill its computing power commitments to enterprise customers. Frozen v2 is clearly a move to alleviate this gap.
