Tech Beetle briefing IN AI

Google reportedly working on a new AI chip, here is how it may improve Gemini models

Essential brief

Alphabet, Google's parent company, is reportedly developing a new AI chip called Frozen v2, aimed at improving the speed and power efficiency of its Gemini AI models. Expected to launch in 2028, Fr

Key topics

google reportedly working ai chip here improve gemini models Alphabet Google AI Frozen Gemini AI

Key facts

Alphabet is developing Frozen v2, a new AI chip expected to launch in 2028.

Frozen v2 could be six to ten times more power-efficient than current Google AI chips.

Custom AI chips help companies improve performance and reduce reliance on external suppliers.

Alphabet plans to invest up to $190 billion in AI technologies this year.

Highlights

Frozen v2 is designed to enhance the speed and energy efficiency of Google's Gemini AI models.

The chip may generate significantly more AI tokens per unit of power compared to existing chips.

Google emphasizes co-designing hardware and software for optimized AI workloads.

Other AI companies like OpenAI and Anthropic are also developing custom AI chips.

Alphabet's stock rose about 3% following reports of the Frozen v2 chip development.

Why it matters

The development of Google's Frozen v2 AI chip highlights the increasing importance of custom hardware in advancing AI performance and efficiency. As AI models demand more computing power, companies designing their own chips can optimize operations, reduce costs, and lessen dependence on dominant suppliers like Nvidia. This trend is shaping the future of AI infrastructure and competitive dynamics in the technology sector.

Alphabet, the parent company of Google, is reportedly working on a new AI chip named Frozen v2, designed to enhance the performance and energy efficiency of its Gemini AI models. The chip is anticipated to be released in 2028 and is expected to significantly outperform Google's existing AI hardware. According to reports, Frozen v2 may achieve between six and ten times the efficiency of current Google AI chips, measured by the number of AI tokens generated per unit of power consumed. This improvement could allow Gemini models to deliver faster responses while using less energy.

Google has emphasized its ongoing commitment to innovation through co-designing hardware and software to optimize real-world AI workloads. While not all projects reach production, this approach aims to maximize system integration and efficiency. The development of Frozen v2 aligns with a growing trend among AI companies to create custom chips rather than relying solely on third-party manufacturers.

This shift is driven by increasing demand for AI computing power and the desire to reduce operational costs and dependency on suppliers like Nvidia, which currently dominates the AI chip market. Other companies are also pursuing similar strategies; for example, OpenAI introduced its custom chip Jalapeno in June, and Anthropic is reportedly negotiating a chipmaking partnership with Samsung.

Alphabet has been investing heavily in AI technologies, with plans to spend between $180 billion and $190 billion this year. Following news of the Frozen v2 chip, Alphabet's stock rose by approximately 3 percent. This investment underscores the company's focus on advancing AI capabilities through hardware innovation.

The development of Frozen v2 represents a significant step in Google's efforts to improve AI model efficiency and maintain competitiveness in the rapidly evolving AI industry.

Key topics in this update include google reportedly working, ai chip here, and improve gemini models.