Alphabet is developing a new server chip known as “Frozen v2” to improve the efficiency of its in-house Gemini models, according to a report from The Information. The chip is expected to be released in 2028 and could offer efficiency gains of six to 10 times compared to Google’s current AI chips, based on the number of tokens generated per unit of power.
Google did not confirm the report but stated, “Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers,” in a response to TechCrunch. The company added that while not all projects move into production, the exploration process is integral to its full stack approach, which focuses on optimizing hardware and software integration.
Demand for proprietary AI chips is increasing as companies seek to enhance the performance of their models while addressing global AI computing shortages. This trend comes amid a market shift to reduce reliance on Nvidia, a dominant chipmaker in the AI sector. In June, OpenAI announced its own custom chip called Jalapeño. Earlier this month, reports surfaced of Anthropic exploring a chipmaking partnership with Samsung.
Investor apprehension regarding Alphabet’s significant planned spending, estimated between $180 billion and $190 billion, has also impacted perceptions of its AI strategy. Following the report on the Frozen v2 chip, Google’s stock rose roughly 3% on Monday morning, indicating a positive response from the market ahead of its earnings report later this week.
