AI generated image for representational purpose Google is reportedly working on a new AI chip, internally called “Frozen v2.” According to a report by The Information (via CNBC), the chip is designed to run its Gemini models far more efficiently. The move comes as the Search giant grapples with a serious internal shortage of AI computing power for which it reportedly announced a deal with Elon Musk’s SpaceX.
What is ‘Frozen v2’
The report says that unlike Google’s existing Tensor Processing Units (TPUs), which are built to handle a wide range of AI workloads, Frozen v2 takes a more specialised approach. The chip would permanently embed parts of Gemini’s architecture directly into the silicon itself, cutting down the number of calculations and the amount of data movement needed to answer a query. Google engineers reportedly believe this could let the chip serve six to ten times more tokens per unit of power compared to the company’s latest TPUs – essentially mitigating the high cost of processing. Notably, Frozen v2 will not replace TPUs and will become a more specialised branch within Google’s custom-chip lineup. The company is reportedly targeting 2028 for its deployment, and for now, sees the project partly as a trial run rather than something to be produced at TPU-scale.
Why Google needs new chip and what it said
The project comes at a time when Google (and other companies) is dealing with a major compute crunch internally. Last month, Google reportedly agreed to pay SpaceX close to $1 billion a month to help ease this shortage and meet its enterprise compute commitments.Responding to the report, Alphabet told CNBC that its teams are “constantly researching and experimenting with new innovations to deliver maximum performance and efficiency” for users and customers, adding that not every project makes it to production but that this kind of exploration is central to its “full stack approach.” The company also said that designing hardware and software together helps keep its systems “integrated and highly optimised for real-world workloads.”Beyond the chip shortage, Google’s AI ambitions face more immediate pressure too. The next release of Gemini Pro has been delayed, reports said, and the company has lost several senior AI researchers to rival firms like Anthropic and OpenAI. That competitive pressure intensified over the weekend, with new AI model releases from Moonshot AI and Alibaba narrowing the capability gap with US labs.