Google Developing 'Frozen v2' AI Chip to Boost Gemini Efficiency

Jakarta (ANTARA) — Google’s parent company, Alphabet, is reportedly developing a new server chip to improve the operational efficiency of its Gemini artificial intelligence (AI) model.

As reported by TechCrunch on Monday (20 July) local time, anonymous sources said the chip is known internally as “Frozen v2” and is targeted for launch in 2028.

According to the report, Frozen v2 could potentially be six to ten times more efficient than Google’s current AI chips, measured by the number of tokens that can be generated for each unit of power used.

“Our team continues to research and experiment with a variety of new innovations to deliver maximum performance and efficiency for our users and customers,” Google said in a statement.

The company added that not all projects under development will proceed to the production stage, but such exploration is an important part of developing technology that integrates hardware and software.

Developing in-house AI chips has become an increasingly common trend among technology companies in recent years.

Beyond improving the efficiency of AI models, the move is also intended to reduce dependence on external chip supplies amid soaring global demand for AI computing capacity.

In addition, several companies are seeking to lessen their reliance on NVIDIA, which has long dominated the AI chip market and serves as the main supplier for many AI model developers.

In June, OpenAI announced its first custom chip for AI inference needs, named Jalapeño. Meanwhile, earlier this month, Anthropic was reported to be exploring a chip development partnership with Samsung.

The report on the development of the more efficient Frozen v2 chip is said to have boosted investor optimism in Google’s AI strategy. Following the report’s publication, Alphabet shares reportedly rose around 3 per cent in Monday morning trading (20 July) local time.