According to a report by TechCrunch, Alphabet, the parent company of Google, is currently in the process of developing a cutting-edge server chip, internally codenamed Frozen v2. This innovative chip is specifically designed to boost the operational efficiency of the Gemini model. By integrating a portion of the Gemini model's foundational architecture directly into the hardware, the Frozen v2 chip significantly reduces both computational demands and data transmission volumes. Remarkably, it boasts a token processing capability per unit of power consumption that is 6 to 10 times greater than that of current TPUs (Tensor Processing Units). The anticipated deployment date for this groundbreaking technology is set for 2028. Rather than replacing TPUs, the Frozen v2 chip is envisioned to function as a specialized supplementary branch, enhancing the overall capabilities of Alphabet's AI infrastructure.