Alphabet's shares increased nearly 3% following reports that Google is developing Frozen v2, a new AI chip aimed at improving power efficiency for its Gemini models.
Technical Innovation Behind Frozen v2
The chip will embed core elements of Gemini's architecture directly into the hardware, reducing the need for extensive calculations and data movement during processing. This integration is expected to enable the chip to serve six to ten times more tokens per unit of power compared to Google’s latest tensor processing units (TPUs), which currently support many of its AI operations.
Frozen v2 is not intended to replace the existing TPU lineup but will serve as a specialized device optimized for Gemini inference workloads. The development represents a step toward co-designing hardware and software to optimize performance for specific AI tasks.
According to reports, Google plans to launch Frozen v2 by 2028, viewing the project as a trial that may not be produced at the scale of TPUs. The chip’s architecture requires consistency in future Gemini models to maintain its efficiency advantages.
Google's efforts to enhance AI hardware come amid growing demand for compute resources. The company has faced capacity constraints in Google Cloud due to increased usage for coding, search, and enterprise AI applications. Google recently agreed to pay SpaceX nearly $1 billion a month to help address this demand.
Alphabet's move to develop this dedicated AI chip follows a market environment where AI compute needs are intensifying. The innovation reflects a broader industry trend to improve the energy efficiency of AI infrastructure.
This material is for informational purposes only and does not constitute financial advice.
