Google, in a move to become more self-reliant in terms of AI chips, is developing a server chip to fuel its Gemini AI model. Codenamed Frozen v2 chip, the chip is being slated for a 2028 launch — giving Google nearly two years to complete its development. The functionality of this chip will be centred on optimizing Gemini’s performance to help Google keep pace with AI-native rivals like OpenAI and Anthropic.
Google, unlike many AI companies, does not rely on Nvidia for its AI chip requirements. Instead, it uses its custom-built AI hardware 00 the Tensor Processing Units (TPUs). As per Google, the latest TPUs can process 121 quintillion calculations per second with double the bandwidth of previous generations.
The chip will reportedly be capable of responding to queries upto six to ten times faster than what the current chips allow.
For Google, the bigger game is to stitch together a native AI ecosystem that largely on home-made semiconductors to achieve an neatly integrated advanced tech fabric.
With this approach, the search engine giant will also be able to extract itself from the intensifying race to procure the latest AI infrastructure from companies like Nvidia, bringing costs up.
Reports circulating about Google’s plans with the Frozen v2 chips pushed its stock prices up by around 5.25 percent on Monday. At the time of writing, GOOG’s price on Nasdaq stood at $351.37.
Source: Yahoo Finance
As AI infux rapidly makes inroads into areas such as blockchain, finance and tech, established companies are feeling the burden on their pockets to be able to afford the spendings.
For 2026, Google has set its AI expenditure target between $175 billion and $185 billion. Microsoft, on the other hand, has projected its AI spending to range between $120 billion and $190 billion. To manage these skyrocketing costs, multiple companies have announced thousands of layoffs this year so far.
To better manage these costs and cut the dependency on third-party chip makers like Qualcomm, TSMC, AMD, and market titan Nvidia — Google, Meta, and OpenAI have started designing their own semiconductors.
Earlier this month, Meta said it would be manufacturing its own AI chip dubbed “Iris” starting this September. Last month, OpenAI launched its first “intelligence processor” called “Jalapeño”, marking its entry into the $13.5 trillion chip market.
While Anthropic is not making its own chips yet, it is exploring partnerships with Microsoft and Samsung to fulfil its requirements.
