Google is reportedly developing a next-generation custom AI server chip, internally codenamed “Frozen v2,” as part of its long-term strategy to improve the efficiency and performance of its Gemini artificial intelligence models. According to media reports, the chip is expected to enter production around 2028 and could deliver between six and ten times greater power efficiency for AI inference compared to Google's current generation of AI processors.
The proposed chip is expected to complement Google's existing Tensor Processing Units (TPUs) rather than replace them. It is being designed specifically to accelerate AI inference—the stage where trained AI models generate responses—while reducing energy consumption and improving the overall cost efficiency of AI workloads across Google's cloud infrastructure.
Focus on AI Inference and Power Efficiency
According to reports, Frozen v2 is being optimized to generate significantly more AI tokens per unit of power consumed, making it substantially more energy-efficient than Google's existing AI chips. Improving inference efficiency has become a strategic priority for AI companies as the cost of running increasingly sophisticated generative AI models continues to rise.
By combining custom-designed hardware with its proprietary AI software stack, Google aims to optimize performance for real-world AI applications while supporting the rapid expansion of Gemini across consumer and enterprise services.
Google Highlights Ongoing Hardware Innovation
While Google did not officially confirm the development of Frozen v2, the company also did not deny the reports.
In a statement shared with media outlets, Google said its engineering teams continuously research and experiment with new technologies to maximize performance and efficiency for users and customers.
The company added that not every internal project reaches production, but emphasized that its full-stack approach—co-designing hardware and software together—helps create highly optimized AI systems capable of handling demanding workloads.
AI Chip Race Intensifies
The reported development comes as major technology companies increasingly invest in custom AI silicon to reduce dependence on third-party chip suppliers, particularly Nvidia, which currently dominates the AI accelerator market.
Building proprietary AI chips allows companies to optimize performance, improve energy efficiency, reduce operating costs, and secure access to critical computing resources amid ongoing global demand for AI infrastructure.
Several AI companies have accelerated their semiconductor strategies in recent months. OpenAI recently introduced its first custom inference processor, while Anthropic has reportedly been exploring new chip development partnerships. Similar initiatives are also underway across other hyperscale cloud providers seeking greater control over their AI infrastructure.
Part of Google's Broader AI Investment Strategy
Frozen v2 is expected to become another important component of Alphabet's broader AI roadmap as the company continues investing heavily in artificial intelligence, cloud computing, and custom silicon.
Google has committed substantial capital toward expanding its AI infrastructure, including next-generation data centres, advanced networking, and proprietary hardware. As AI adoption accelerates globally, improving computing efficiency is becoming increasingly important for controlling operational costs while delivering faster and more scalable AI services.
If developed as reported, Frozen v2 could further strengthen Google's position in the increasingly competitive AI infrastructure landscape by enabling more efficient deployment of Gemini models across its ecosystem.
𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲
𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬 👉 Facebook, LinkedIn, Twitter, Instagram