Alphabet shares closed 1.51% higher after a report that the company is developing a new server chip called “Frozen v2” designed to run Gemini AI models more efficiently by permanently embedding parts of the model architecture directly into silicon.
According to The Information, Google engineers project the chip could serve between six and ten times more tokens per unit of power than the company’s current tensor processing units. Frozen v2 would become a specialized branch of Google’s custom-chip portfolio rather than replace its general-purpose TPUs.
The project is aimed at easing a major internal compute shortage that has forced Google Cloud to turn away outside business. Last month, Google agreed to pay SpaceX nearly $1 billion a month to help bridge the gap. The company is targeting 2028 for Frozen v2 deployment.
The trade-off is reduced flexibility — the chip would work with future Gemini models only if Google maintains the same underlying architecture. Google currently views Frozen v2 partly as a trial run and does not plan to produce it at TPU scale.
Google’s AI efforts face broader challenges including a delayed next Gemini Pro release, senior researcher departures, and growing competition from Chinese models that now account for 45% of U.S. company token use.
