Google develops custom Frozen v2 chip to boost Gemini efficiency

The project signals a shift in hardware strategy for the tech giant, which currently relies heavily on market-dominant Nvidia. By co-designing hardware and software, Google intends to optimize its systems for real-world workloads, potentially reducing the massive capital expenditures that have recently unsettled investors. While the company declined to confirm the report, a spokesperson noted that rigorous exploration of new architecture remains central to their full-stack approach.

Industry peers are following a similar path to secure computing independence. OpenAI recently unveiled its Jalapeño inference processor, and Anthropic is exploring a chipmaking partnership with Samsung. For Google, the stakes remain high: after committing nearly $190 billion to AI infrastructure this year, the company faces mounting pressure to demonstrate tangible returns. The market reacted favorably to the news of the Frozen v2, with Alphabet shares climbing 3% on Monday ahead of the upcoming earnings report.