Woofun AI reports that Google is engineering the "Frozen V2" processor to optimize inference for its Gemini AI model. The new hardware is projected to deliver six to ten times the efficiency of existing TPU units, with a rollout scheduled for as early as 2028.