source: techcrunch ai: google is working on a new ai chip designed to make gemini more efficient
level: business
alphabet is creating a new server chip called frozen v2 to run its gemini ai models more efficiently. the chip could be six to ten times more efficient than current google ai chips, measured by tokens generated per unit of power. it is expected to launch in 2028, according to a report from the information citing anonymous sources. google did not confirm the chip but said it constantly researches hardware and software co-design for better performance.
ai companies are increasingly building their own chips to improve model efficiency and reduce reliance on nvidia. openai announced its first custom chip, jalapeño, in june, and anthropic is reportedly in talks with samsung for a chipmaking partnership. google's move follows a broader industry trend where firms seek to control hardware costs and address ai computing shortages. efficiency has become a key selling point as concerns about ai spending grow.
news of the frozen v2 chip boosted alphabet's stock by about 3% ahead of its earnings report. investors have worried about google's planned ai spending of $180 to $190 billion. the more efficient chip suggests those investments may yield better returns. google's full-stack approach, co-designing hardware and software, aims to optimize real-world workloads and maintain competitiveness in the ai market.
why it matters: custom ai chips can lower operational costs and energy use, making large-scale ai deployment more sustainable and less dependent on external suppliers.
source: techcrunch ai: google is working on a new ai chip designed to make gemini more efficient