Google designs new AI chip that bakes Gemini directly into silicon
Crypto Briefing 2026-07-20 15:07:48
Context: Google has unveiled a new AI chip, dubbed Frozen v2, which integrates elements of its Gemini model directly into the silicon. This innovative design promises significant efficiency gains, with expectations of 6-10x improvements over previous Tensor Processing Units (TPUs). The chip is slated for deployment as early as 2028.
Key Facts
- The Frozen v2 chip embeds Gemini model elements into silicon, a design innovation that is expected to yield 6-10x efficiency gains over Google's previous Tensor Processing Units (TPUs).
- The Gemini model, now integrated into the Frozen v2 chip, is a key component of Google's AI infrastructure, and its embedding into silicon is anticipated to significantly enhance performance.
- The Frozen v2 chip is expected to be deployed as early as 2028, marking a significant milestone in Google's efforts to advance its AI capabilities through custom hardware design.