Google's new server chip, informally named 'Frozen V2', is expected to deliver dramatic improvements in AI processing efficiency. According to The Information, the chip could be six to 10 times more efficient than Google's latest custom AI chips based on the number of AI tokens served per unit of power. This significant efficiency gain is achieved through direct integration of Gemini model blueprints into the hardware, reducing data movement and decision-making during inference processes. The development comes as Google addresses internal AI computing capacity shortages that have created tensions within the company and prompted Google Cloud to decline deals with outside customers.
The chip incorporates elements of Google's Gemini model directly into the hardware, as reported by The Information. With the blueprints built into the chip, the amount of data movement is expected to be reduced, along with the number of decisions needed to be taken during inference. This integration approach is designed to enhance the overall efficiency of AI model operations and processing capabilities, representing a fundamental shift in how AI hardware is designed and deployed. The 'Frozen' project specifically aims to create a new set of homegrown chips apart from Google's tensor processing units (TPUs), rather than replacing them entirely. The semiconductor is separate from Google's tensor processing units (TPUs) and would reduce the amount of data the chip has to move around, making it faster at responding to queries.
Google is looking to deploy the chip by 2028 at the earliest, though engineers are still finalizing the design and the amount of model information that will be hardwired, according to The Information. The development represents a significant step in the company's AI infrastructure strategy, with the chip specifically aimed at addressing an AI computing capacity crunch that has fueled internal tensions and prompted Google Cloud to decline deals with outside customers. The 'Frozen' project is designed to create a new set of homegrown chips apart from Google's tensor processing units (TPUs), rather than replacing them entirely.
Shares of Alphabet were up 3.3% in early trading following the announcement, as reported by The Information. The positive market reaction reflects investor confidence in the company's AI infrastructure development initiatives and their potential impact on future competitive positioning in the AI market. A Google Cloud spokesperson stated that "Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for ... By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized." The development comes amid reports that Google delayed the launch of its latest Gemini AI model after it fell short of internal goals, with the company working to improve its capabilities, particularly in coding.
