Google Plans New Chip to Run Gemini Models More Efficiently, the Information Reports
- On Monday, Alphabet is developing a new server chip internally dubbed 'Frozen v2,' designed to embed Gemini model architecture directly into hardware to improve computational efficiency, The Information reported.
- Internal compute capacity shortages fueled tensions and forced Google Cloud to decline deals with outside customers, prompting the development of this specialized hardware solution.
- Engineers project the chip could serve six to 10 times more tokens per unit of power than current TPUs, with deployment targeted as early as 2028.
- Alphabet shares climbed 3% on Monday following the report, adding a combined $15 billion to the net worths of cofounders Larry Page and Sergey Brin.
- While the company views this as a trial run rather than a replacement for general-purpose TPUs, the move reflects a broader industry shift toward fusing AI models with specialized silicon.
37 Articles
37 Articles
Google is working on a new artificial intelligence chip, which focuses on significantly increasing efficiency when running Gemini series models.
It has been revealed that Google is developing custom semiconductors optimized for its artificial intelligence (AI) model, 'Gemini'. On the 20th (local time), The Information, a U.S. IT specialized media outlet, reported, citing multiple sources, that Google is developing an AI chip codenamed 'Frozen v2' with the goal of deploying it in servers by 2028. Frozen v2 is a Tensor Processing Unit (TPU) developed and released by Google
Google just bet its inference future on a chip built for one model
The race to make AI inference cheaper is pushing chip design beyond general-purpose accelerators. We’re now moving toward silicon that is tailored to individual models. And Google’s reported “Frozen v2” project suggests the Gemini AI model is part of the movement. First reported by The Information, the unannounced chip would reportedly hardwire parts of Gemini’s architecture while leaving its weights updatable. That compromise could give Google …
Coverage Details
Bias Distribution
- 47% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
