Alphabet Inc.'s (GOOGL) (GOOG) next AI chip may be less about building faster hardware and more about keeping pace with surging AI demand. On the company's second-quarter earnings call, CEO Sundar Pichai revealed that Google's AI model APIs are now processing approximately 22 billion tokens every minute, up from 16 billion just one quarter earlier.

He also acknowledged that the company remains "supply constrained" as AI demand continues to outstrip available computing capacity.

Enter Frozen v2

Those comments help explain reports that Alphabet is developing Frozen v2, a next-generation AI chip designed to run Gemini models more efficiently.

According to The Information, Frozen v2 integrates parts of Gemini's architecture directly into the hardware. Engineers reportedly believe the chip could process six to 10 times more AI tokens per unit of power than Google's latest custom AI chips, potentially allowing the company to serve far more AI requests without a proportional increase in infrastructure.

The broader takeaway is that Google's AI hardware strategy is increasingly being driven by demand rather than technological ambition alone. As Gemini adoption accelerates across Search, Cloud and enterprise products, the company is racing to build infrastructure that can keep up.

For investors, Frozen v2 represents more than another AI chip. It is Google's attempt to solve a problem created by its own success: processing tens of billions of AI tokens every minute while easing growing compute constraints.