Google Reportedly Developing ‘Frozen v2’ AI Chip to Boost Gemini Efficiency

Barchart·07/20/2026 19:16:28

Alphabet ($GOOGL) is developing a new AI server chip, internally dubbed "Frozen v2," that would embed elements of its Gemini AI model directly into hardware to improve inference efficiency, according to The Information. The chip is reportedly targeted for deployment in 2028 and is intended to complement, rather than replace, Google's Tensor Processing Units (TPUs) as the company works to expand AI computing capacity.

• The chip could deliver six to ten times more AI tokens per unit of power than Google's latest TPU chips, according to the report.
• Frozen v2 would hardwire portions of Gemini's architecture into silicon to reduce computation and data movement during AI inference.
• The project is reportedly aimed at easing internal compute shortages that have limited Google Cloud's ability to serve some enterprise customers.
• Google is expected to deploy the chip around 2028 and currently views the project as a specialized addition to its custom AI chip portfolio.

Relevant Companies

• Alphabet ($GOOGL) – Developing the Frozen v2 chip to improve Gemini inference efficiency and expand AI infrastructure.
• SpaceX ($SPCX) – Recently reached a reported compute capacity agreement with Google as the company works to address AI infrastructure constraints.
• NVIDIA ($NVDA) – Custom AI chip development by hyperscalers continues to shape demand for third-party AI accelerators.

Editor’s Note: This is a developing story. This article may be updated as more details become available.

This article contains syndicated content. We have not reviewed, approved, or endorsed the content, and may receive compensation for placement of the content on this site. For more information please view the Barchart Disclosure Policy here.