Devices · 5h ago
Google builds 'Frozen v2' chip with Gemini architecture in silicon
Google is developing a server chip called 'Frozen v2' that embeds part of its Gemini AI model's architecture directly into the hardware. Engineers project the chip will deliver 6 to 10 times more tokens per watt than Google's latest TPUs. The move aims to dramatically improve AI inference efficiency for large language models.
Meridian48 take
This could give Google a significant edge in AI inference cost and speed, but the real test will be whether the architectural integration scales beyond a single model generation.
Read the full reporting
Google reportedly developing 'Frozen v2' chip with Gemini's architecture etched into the silicon — engineers project 6 to 10 times more tokens per watt than latest TPUs →
Tom's Hardware
google-aicustom-silicon