Google developing Gemini-specific chip called Frozen v2

Google is developing Frozen v2, a custom server chip that embeds Gemini architecture directly into silicon, targeting 6-10x better energy efficiency than current TPUs and deployment by 2028. The project aims to resolve Google's internal capacity bottleneck that has forced Google Cloud to decline external customer business. Google just locked Gemini's architecture into silicon to gain 6-10x efficiency - but that means the chip can only run Gemini as-is, and any founder selling Gemini inference optimisation or acceleration becomes redundant the moment this ships in 2028. You have two years to own the problem before Google owns the solution.