Google is reportedly building Frozenv2, a Gemini-specific chip that could be 10x more efficient
新智元 · wechat · 2026-07-21
A WeChat report says Google is working on a server chip codenamed Frozenv2 that would hard-code parts of Gemini’s architecture into silicon.
- The chip is described as not running the model on top of hardware, but embedding key decision paths directly into the chip.
- According to the report, it could be 6–10× more efficient per watt than Google’s latest TPU.
- Earliest deployment is said to be 2028.
- The story frames the move as Google’s response to a severe compute shortage and an industry-wide push toward more specialized inference silicon.
Related event: Google Reportedly Developing Gemini-Specific Chip Frozen v2(10 posts)→
More from Infra
- Anthropic S-1: only $19B total compute spend since inception, overbuild warning — bookwormengr · 2026-10-27
- CoreWeave, You.com and NVIDIA lift BrowseComp 22.9% in 3-week RL post-training sprint — RichardSocher · 2026-10-27
- Grid queues make power architecture an AI-factory decision; Exowatt and Redwood offer fixes — ingliguori · 2026-10-27
- Two CVSS 9.9 Wasmtime flaws allow sandbox escape, one ARM64-only since 2020 — basedjensen · 2026-10-27
- AI Infra Meme: While Countries Need TW-Scale Power for Datacenters, People Argue About Lightbulbs — mgostIH · 2026-10-27
- NVIDIA reportedly stops taking RTX 5090 orders from board partners, even backorders — chemist_slime · 2026-10-27