On July 20th, US tech news outlet The Information disclosed, citing sources, that Google is in the process of developing a cutting-edge server chip. This chip will seamlessly incorporate the design of its Gemini AI model, thereby significantly elevating the service efficiency of AI models. Internally dubbed 'Frozen v2', this chip is engineered to address the current scarcity of AI computing power that Google is grappling with. This deficit in computing power has resulted in internal resource limitations, compelling Google Cloud to decline collaborations with certain external clients.

According to projections by Google staff, the new chip will exhibit a remarkable 6 to 10-fold increase in token processing capability per unit of power consumption, in comparison to Google's existing latest-generation, in-house-developed AI chips. At present, engineers are meticulously refining the primary functions of the new chip and devising how its diverse components will synergize. Google has set its sights on deploying this chip as early as 2028.