Model-Specific AI Silicon
Hardware vendors increasingly co-design chips around a specific model architecture, trading general-purpose flexibility for large gains in tokens-per-watt, and this hardware-software co-design becomes a recurring axis of the compute race as inference volume dominates cost.
forming · confidence 38 · Emerging (watchlist) · tracking since July 21, 2026 · updated July 21, 2026
Why the conviction moved
- Jul 21Strengthened
Google is reportedly building a 'Frozen v2' server chip that bakes Gemini's architecture directly into silicon, which engineers project could serve six to ten times more tokens per watt than its newest TPUs by trading flexibility for efficiency.
Source trail
Supporting · July 21, 2026
Google Is Building a Chip That Bakes Gemini's Architecture Into Silicon
Google is reportedly building a 'Frozen v2' server chip that bakes Gemini's architecture directly into silicon, which engineers project could serve six to ten times more tokens per watt than its newest TPUs by trading flexibility for efficiency.
CNBC
Unlock full source trail, score history, and daily updates.
Unlock Trends