Google Reportedly Developing 'Frozen v2' AI Chip

Google is reportedly developing a next-generation AI server chip, internally codenamed Frozen v2, that could integrate selected capabilities of its Gemini artificial intelligence models directly into hardware. According to reports, the chip is being designed to improve AI inference efficiency and overcome computing capacity limitations that have affected Google Cloud's ability to serve some external customers. Engineers are still finalizing the design, with deployment expected by 2028. The company believes the chip could be six to ten times more power-efficient than its current custom AI processors for specific workloads.

The proposed Frozen v2 chip is expected to complement Google's existing Tensor Processing Units (TPUs) rather than replace them, creating a separate family of AI accelerators optimized for inference tasks. Google stated that it continues to innovate by co-designing hardware and software to maximize performance and efficiency. The development comes amid growing global investment in AI infrastructure, as technology companies race to build more powerful and energy-efficient chips to support the next generation of artificial intelligence services.