NEW YORK — Google is developing a new custom artificial intelligence server chip designed to run its Gemini AI models far more efficiently, according to a report published as the company seeks to reduce computing costs, ease internal capacity shortages and strengthen its position in the rapidly expanding AI infrastructure race. The project, internally known as “Frozen v2,” is still under development and has not been officially announced by Google.
According to people familiar with the project, the new chip would incorporate portions of Google’s Gemini AI architecture directly into the hardware itself rather than relying entirely on software running atop general-purpose AI processors. By embedding parts of the model into the silicon, Google aims to significantly reduce power consumption while increasing the number of AI requests each chip can process.
The reported design could make the processor six to ten times more efficient than Google’s latest custom AI chips when measured by AI tokens processed per unit of electricity, representing a potentially major advance in lowering the cost of operating large language models. Engineers are reportedly still finalizing the design, and deployment is not expected before 2028.
The project reflects one of the biggest challenges facing artificial intelligence companies today: computing capacity. Demand for AI services has grown so rapidly that even major technology companies have struggled to secure enough processing power. Reports indicate Google’s internal shortages have at times forced Google Cloud to decline potential customer contracts because available AI infrastructure was fully utilized.
Rather than replacing Google’s existing Tensor Processing Units (TPUs), Frozen v2 is reportedly intended to complement them by handling specific Gemini inference workloads more efficiently. The strategy would allow Google to lower operating costs while expanding the amount of AI computing available across Search, Workspace, Cloud, Android and other Gemini-powered services.
The development comes as competition among AI infrastructure providers intensifies. Alphabet, Microsoft, Amazon, Meta and OpenAI continue investing billions of dollars in custom hardware, advanced data centers and semiconductor technologies designed to reduce dependence on third-party processors while improving AI performance.
For businesses, more efficient AI hardware could ultimately reduce cloud computing costs while allowing companies to deploy larger and more sophisticated artificial intelligence applications. Faster, cheaper AI processing may also accelerate adoption across healthcare, finance, manufacturing, cybersecurity and customer service.
The reported project also underscores the increasing importance of vertical integration in artificial intelligence. Instead of relying solely on outside chip manufacturers, technology companies are increasingly designing specialized processors tailored specifically to their own AI models, allowing software and hardware to be optimized together.
Investors welcomed the report, with Alphabet shares rising more than 3% during Monday’s trading session, reflecting optimism that improved AI efficiency could strengthen Google’s competitive position while reducing long-term operating expenses.
The report follows news last week that Google delayed the release of its latest Gemini AI model while engineers continued improving its coding performance and overall capabilities. Together, the developments illustrate Google’s effort to strengthen both the software and hardware foundations of its AI ecosystem before the next generation of products reaches consumers.
Although Google has not confirmed specific details of Frozen v2, the reported initiative highlights how the global AI race is increasingly shifting beyond software models toward the specialized infrastructure required to operate them efficiently at massive scale. Companies capable of reducing AI computing costs while improving performance are expected to gain significant competitive advantages as enterprise AI adoption continues accelerating.
JBizNews Desk | New York
© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.
