Alphabet Shares Rise on Report of New 'Model-Baked' AI Chip

Shares of Alphabet Inc. (NASDAQ: GOOGL) climbed on Monday morning following a report that the company is developing a novel type of custom hardware that integrates its flagship AI model directly onto the chip.

The experimental project, dubbed "Frozen v2", aims to dramatically increase processing efficiency by hard-coding the architecture of its Gemini AI model into the silicon itself, according to a report from *The Information*.

A New Approach to AI Hardware

Instead of running AI models on general-purpose processors like GPUs or Google's own Tensor Processing Units (TPUs), the "model-baked" approach bypasses significant processing overhead. By etching the model's design onto the chip, Google aims to reduce data movement and processing steps, leading to substantial efficiency gains.

Key details from the report include:

• Efficiency Target: The new chips are projected to be 6 to 10 times more power-efficient per token than Google's current-generation TPUs.
• Flexibility: While the core architecture is fixed, engineers can still update the model's parameters with new weights, allowing for continued learning and refinement.
• Timeline: The project is reportedly planned for potential deployment as early as 2028 as a parallel product line to its existing TPUs.

Ad

Strategic Implications

This initiative appears designed to address the significant compute shortages that have become a bottleneck for both internal development and Google Cloud's ability to serve external clients. By creating highly specialized and efficient hardware, Google could alleviate these constraints and strengthen its competitive position in the AI infrastructure market.

In response to the report, a Google spokesperson acknowledged that its teams are actively experimenting with high-efficiency innovations. However, the spokesperson cautioned that not every experimental project is guaranteed to reach commercial production.

Market Reaction

Following the news, shares of Alphabet (GOOGL) rose 1.15% in Monday morning trading. The investor response suggests optimism about the potential for Google to develop a proprietary hardware advantage in the capital-intensive AI arms race.