Google plans new AI Chip to improve Gemini AI performance: All details
Google's in-house AI chip is said to generate 6 to 10 times more AI tokens for the same amount of electricity.
- Jul 21, 2026,
- Updated Jul 21, 2026 11:13 AM IST
Alphabet, Google’s parent company, is reportedly developing a new in-house artificial intelligence (AI) chip to boost Gemini AI models. This chip may help Gemini AI models run more efficiently, which means it could offer faster performance, lower power consumption, and reduced operating costs.
According to The Information report, the new server AI chip is codenamed Frozen V2, and it is planned to start manufacturing the chip in 2028. Frozen V2 is said to be much more efficient than Google's current AI chips. It is expected to operate as a dedicated AI accelerator, designed to improve efficiency across real-world AI tasks.
Must read: Google’s Gemini 3.5 Pro delayed; Performance falls short of expectations
Reportedly, the new chip is said to generate 6 to 10 times more AI tokens for the same amount of electricity. Therefore, the chip is expected to deliver improved AI performance while using significantly less power. Lastly, the new AI chip may come alongside Google’s existing Tensor Processing Units (TPUs).
TechCrunch also quoted a Google spokesperson saying, “Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers. While not every project moves into production, this rigorous exploration is central to our full stack approach.”
Must read: ‘We're entering a new age’: Google DeepMind CEO Demis Hassabis on emergence of AGI
The company further added that “By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimised for real-world workloads.
While the AI chip may help improve Google’s AI models, it will also give a competitive edge over rivals like OpenAI, Microsoft, Meta, and Amazon. However, rivals are planning similar moves; in June, OpenAI launched its first custom inference processor called Jalapeño. Anthropic is also in talks with Samsung for a chipmaking partnership. Therefore, businesses have started to secure the future of AI innovations.
The report also comes amid Google’s delayed launch of its flagship AI model, the Gemini 3.5 Pro. The company is reportedly facing performance issues, and it is still going through training processes before the public release.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Alphabet, Google’s parent company, is reportedly developing a new in-house artificial intelligence (AI) chip to boost Gemini AI models. This chip may help Gemini AI models run more efficiently, which means it could offer faster performance, lower power consumption, and reduced operating costs.
According to The Information report, the new server AI chip is codenamed Frozen V2, and it is planned to start manufacturing the chip in 2028. Frozen V2 is said to be much more efficient than Google's current AI chips. It is expected to operate as a dedicated AI accelerator, designed to improve efficiency across real-world AI tasks.
Must read: Google’s Gemini 3.5 Pro delayed; Performance falls short of expectations
Reportedly, the new chip is said to generate 6 to 10 times more AI tokens for the same amount of electricity. Therefore, the chip is expected to deliver improved AI performance while using significantly less power. Lastly, the new AI chip may come alongside Google’s existing Tensor Processing Units (TPUs).
TechCrunch also quoted a Google spokesperson saying, “Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers. While not every project moves into production, this rigorous exploration is central to our full stack approach.”
Must read: ‘We're entering a new age’: Google DeepMind CEO Demis Hassabis on emergence of AGI
The company further added that “By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimised for real-world workloads.
While the AI chip may help improve Google’s AI models, it will also give a competitive edge over rivals like OpenAI, Microsoft, Meta, and Amazon. However, rivals are planning similar moves; in June, OpenAI launched its first custom inference processor called Jalapeño. Anthropic is also in talks with Samsung for a chipmaking partnership. Therefore, businesses have started to secure the future of AI innovations.
The report also comes amid Google’s delayed launch of its flagship AI model, the Gemini 3.5 Pro. The company is reportedly facing performance issues, and it is still going through training processes before the public release.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine