Reactions from ranked influencers
11 posts
Google may be preparing to freeze parts of Gemini’s architecture directly into silicon. Informally called "Frozen v2," the chip reportedly targets 6–10× more tokens per watt than Google’s newest TPUs. Deployment is planned for as early as 2028. The motivation is immediate: Google’s AI compute shortage has reportedly become severe enough that its Cloud division has turned down outside customers. The efficiency comes with rigidity. Future Gemini models could use Frozen v2 only while retaining the same underlying architecture. TPUs reduced Google’s dependence on Nvidia. Frozen v2 would go further, tying Gemini’s architecture directly to the silicon running it.
@_xjdr I'd imagine search ai is a huge cost driver, knowing what kind of traffic search runs, remembering the impact instant had. And so is probably worth it on that basis alone. And if they can make that model smarter, too...
Google is developing a new chip named 'Frozen v2' specifically designed to run the Gemini family of models more efficiently. The new hardware is expected to arrive in 2028. This was originally reported by The Information. Gemini probably helped design this.
Love my Google friends but maybe make a good model first and _then_ focus on inference efficiency after people actually want to use it ?
Google is developing a new chip named 'Frozen v2' specifically designed to run the Gemini family of models more efficiently. The new hardware is expected to arrive in 2028. This was originally reported by The Information. Gemini probably helped design this.