Google · 2026-07-20 · rumor
Google 'Frozen v2' — custom Gemini-only chip promised 6–10× more efficient
The Information reports Google is building 'Frozen v2', a server chip that hard-codes parts of Gemini's model structure into silicon for 6–10x more tokens per watt than current TPUs. Deployment targeted for 2028.
Google is reportedly baking parts of Gemini's architecture directly into a custom server chip, targeting 6–10x more tokens per watt.
What is it?
Frozen v2 is a Google server chip that would embed parts of the Gemini model's structure directly into the silicon, per a July 20 report from The Information relayed by TechCrunch and CNBC. Unlike Google's general-purpose TPUs, which run any model, Frozen v2 would be built to serve Gemini and only Gemini — trading flexibility for a much shorter compute path per token.
How does it work?
The idea, per The Information's sources, is that hard-wiring Gemini's layers into hardware removes many programmable steps a TPU would otherwise perform, cutting energy per generated token. The chip is described as an internal Google project targeting 6–10x better tokens-per-watt than current TPUs. Google declined to confirm the specifics but said its teams are 'constantly researching and experimenting' with new performance and efficiency work.
Why does it matter?
If Frozen v2 hits its efficiency target, Google could serve Gemini traffic at a fraction of today's power cost — easing an internal compute crunch that reportedly forced Google Cloud to turn away outside business and pushed it into a roughly $1B/month SpaceX capacity deal. It also deepens the hyperscaler drift away from Nvidia GPUs. Alphabet's stock closed up ~3% on the report.
Who is it for?
AI infra watchers, TPU/Nvidia analysts, Google Cloud customers
Try it
Nothing to try — internal, 2028 deployment