Google building next-gen AI chip to boost Gemini efficiency

A report suggests Google is working on a new in-house AI processor, codenamed Frozen v2, aimed at significantly improving Gemini's performance and energy efficiency as competition intensifies over custom AI hardware and infrastructure.

Alphabet, Google’s parent company, is reportedly working on a next-generation artificial intelligence processor that could significantly improve the efficiency of its Gemini AI models, reflecting the industry's growing focus on proprietary hardware to strengthen AI capabilities and reduce reliance on external chip suppliers.

According to reports, Google is developing a server processor internally known as Frozen v2, with production targeted for 2028. The reports, citing people familiar with the project, said the chip is expected to deliver a substantial leap in efficiency over Google's current AI processors by generating significantly more AI output while consuming less power.

While the company did not confirm details of the reported project, Google said research into advanced hardware remains an integral part of its long-term AI roadmap.

"Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers," a Google spokesperson said. "While not every project moves into production, this rigorous exploration is central to our full stack approach. By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads."

Custom silicon emerges as a strategic AI priority

The reported initiative reflects a broader industry shift toward designing specialised AI chips that are closely integrated with proprietary software platforms. Technology companies are increasingly investing in custom silicon to improve processing efficiency, optimise operating costs and gain greater control over the infrastructure required to train and deploy advanced AI models.

The move also comes at a time when demand for AI computing resources continues to rise rapidly. Nvidia remains the leading supplier of AI accelerators, but growing dependence on a single vendor has prompted major technology companies to explore alternative hardware strategies through in-house chip development.

Alphabet is not alone in pursuing this approach. OpenAI recently introduced its first custom inference processor, Jalapeño, while reports have indicated that Anthropic is evaluating potential chip development partnerships to support the expansion of its Claude AI platform.

Investors focus on efficiency alongside AI expansion

As AI companies continue investing heavily in data centres and computing infrastructure, efficiency has become an increasingly important measure of long-term competitiveness. Investors are now looking beyond raw computing performance to assess whether these large capital investments can deliver sustainable economic returns.

Alphabet has already outlined ambitious spending plans to expand its AI ecosystem, with investments expected to support infrastructure, cloud services and AI research. Against that backdrop, reports of a more energy-efficient in-house processor were viewed positively by the market, with Alphabet shares gaining around 3% following publication of the reports.

Although the reported Frozen v2 chip is still several years away from commercial deployment, the project highlights how competition in artificial intelligence is increasingly being shaped by advances in hardware as well as software. As AI models become more sophisticated, companies are expected to place greater emphasis on developing specialised processors capable of delivering higher performance with improved efficiency at scale.