Alphabet, Google’s parent company, is developing a new server chip that would incorporate elements of its Gemini AI models directly into the hardware to improve efficiency, according to a report by Reuters.
The chip, informally known internally as ‘Frozen v2,’ could be deployed as early as 2028, although engineers are still finalising its design and determining how much model information will be hardwired into it.
The chip could be six to 10 times more efficient than Google’s latest custom AI chips, based on the number of AI tokens generated per unit of power. The project is also aimed at addressing a shortage of AI computing capacity that has created internal tensions and led Google Cloud to decline some deals with outside customers.
The Frozen project is intended to create a new line of Google-developed chips separate from its tensor processing units, or TPUs, rather than replace them.
“Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers,” Google told TechCrunch. “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.”
Technology companies have increasingly developed their own AI chips to improve the efficiency of their models and reduce dependence on external suppliers. AI chip efficiency has become a key focus as investors scrutinise the high cost of building and operating AI infrastructure.
Companies are also seeking alternatives to Nvidia, which has historically dominated the AI chip market and supplies hardware used by many major AI companies.
OpenAI announced its first custom chip, an inference processor called Jalapeño, in June. Anthropic was also recently reported to be discussing a chipmaking partnership with Samsung.
Alphabet has faced investor scrutiny over its planned AI spending. The company said earlier this year that it plans to spend between $180 billion and $190 billion as it expands its AI infrastructure and strategy.
Alphabet’s shares rose 3.3% in early trading after the report about the potential new chip, according to Reuters.
The development comes as Google works to improve its Gemini models. Bloomberg News reported last week that Google had delayed the launch of its latest Gemini model after it fell short of internal targets, particularly in coding capabilities.