{“content”: “
Google is reportedly developing a new chip codenamed \”Frozen v2\” optimized for its Gemini series of artificial intelligence models. The chip is expected to deliver between six and 10 times better performance per watt than Google’s existing AI chips, according to multiple sources including SiliconANGLE and TechCrunch. Shares of Alphabet Inc. rose 1.5% following the report about the Frozen v2 chip.
The Frozen v2 chip aims to enhance efficiency by embedding elements of the Gemini AI model directly into the hardware. This design intends to reduce energy and latency costs of inference by hardwiring model parameters into silicon. The chip will complement Google’s existing Tensor Processing Units (TPUs) rather than replace them.
Google plans to roll out the Frozen v2 chip to its data centers in 2028. The company is reportedly addressing internal AI computing capacity shortages with this initiative. Alphabet has committed $75 billion in capital expenditures for 2025 to expand AI computing capacity.
In Q1 2026, Alphabet’s cloud division reported over $20 billion in revenue for the first time, driven largely by AI demand. However, executives acknowledged that cloud growth was \”capacity constrained\” during the earnings call.
To mitigate these constraints, Google has signed a deal worth up to $30 billion with SpaceX to lease approximately 110,000 Nvidia GPUs through mid-2029. The company also unveiled its eighth-generation TPUs, the TPU 8t for training and TPU 8i for inference, at its Cloud Next conference in April 2026.
Google has not officially confirmed details about the Frozen v2 chip but stated that its teams are continuously researching new innovations. The project reflects the company’s ongoing efforts to optimize AI performance amid growing demand and competitive pressures in the AI landscape.
“}
