Google is reportedly developing a new AI server chip, codenamed Frozen v2, that could bring some features of its Gemini AI model directly into hardware. The goal is to make AI models faster and more power-efficient for users.
According to reports, the Frozen v2 chip is expected to solve AI computing capacity issues that have limited some Google Cloud deals with external customers. Google plans to deploy the chip by 2028, although its design is still being finalised. The new chip is expected to be 6 to 10 times more efficient than Google's current custom AI chips in terms of AI performance per unit of power.
A Google Cloud spokesperson said the company is continuously researching new technologies and designing both its hardware and software together to improve performance and efficiency.
The Frozen project is expected to create a new family of Google-designed AI chips that will work alongside the company's existing Tensor Processing Units (TPUs), not replace them. While TPUs will continue to handle general AI workloads, the Frozen chips are expected to process specific AI inference tasks more efficiently.
Google's latest effort reflects the growing demand for AI infrastructure. As AI investments continue to increase, technology companies are focusing on building more powerful and energy-efficient hardware to support advanced AI applications.
