Alphabet Stock Gains On Report Of Google’s New ‘Frozen’ Chip To Boost Gemini AI Efficiency

The project is reportedly part of Google's broader effort to reduce AI infrastructure costs and lessen its reliance on Nvidia's GPUs.

• Google’s new Frozen v2 could be six to ten times more power efficient than its latest TPU chips by embedding parts of Gemini directly into the hardware, according to a report by The Information.
• Unlike Google's general-purpose TPUs, Frozen v2 is reportedly being built exclusively for Gemini, reducing computation and data movement.
• Google views Frozen v2 as a complementary chip family rather than a replacement for its existing TPU lineup.

Alphabet Inc. (GOOG, GOOGL) shares climbed in midday trade on Monday after The Information reported the company is developing a new server chip, internally dubbed “Frozen v2,” designed to run Gemini models more efficiently.

According to the report, the chip would hardwire parts of Gemini's architecture directly into the silicon. By embedding some of the model's decision-making into the chip itself, Frozen v2 would reduce the amount of computation and data movement needed to generate AI responses, making it significantly more power efficient.

The report said Google engineers estimate Frozen v2 could be six to 10 times more efficient than the latest generation of the company's Tensor Processing Units (TPUs) when measured by the number of AI tokens processed per unit of power.

GOOGL’s stock gained more than 3% in midday trade, with retail sentiment on Stocktwits trending in ‘bullish’ territory over the past day, accompanied by chatter at ‘normal’ levels.

How Is Google’s Frozen v2 Different From Its Existing AI Chips?

Unlike Google's TPUs, which are built to run a wide range of AI models, Frozen v2 is reportedly being designed specifically for Gemini.

General-purpose AI chips must constantly adapt to different workloads, requiring additional processing and data transfers. Frozen v2 would eliminate many of those steps by permanently embedding parts of Gemini into the hardware, allowing the chip to process requests more efficiently while consuming less power.

The Information’s report said Google viewed Frozen v2 as a new family of AI chips rather than a replacement for TPUs. The company currently plans a much smaller production run, with the project serving partly as a test bed for building increasingly specialized AI hardware as foundation models mature.

Google Continues Push To Reduce AI Costs

The report comes as Google continues expanding its in-house AI hardware efforts to reduce reliance on Nvidia (NVDA) and lower the cost of running Gemini.

Earlier this year, Google unveiled its eighth-generation TPU and has increasingly marketed the chips to external cloud customers, including signing a multibillion-dollar agreement to supply TPUs to Meta Platforms (META). The company has also been approaching other cloud providers that have traditionally relied on Nvidia's GPUs.

NVDA’s stock gained as much as 1.5% in midday trade on Monday. Retail sentiment on Stocktwits around the AI bellwether fell to ‘bullish’ from ‘extremely bullish’ over the past day, accompanied by chatter at ‘normal’ levels. 

How Is Retail Feeling About Alphabet’s Stock?

Some retail traders on Stockwits questioned whether another custom AI chip would materially affect Alphabet's financial performance, while others said the company's earnings call would provide a clearer picture of how Gemini investments are translating into revenue growth.

View this Stocktwits post

With Alphabet scheduled to report quarterly earnings on Wednesday, Wall Street is expecting to see earnings per share (EPS) of $2.90 on revenue of around $117 billion as per consensus estimates on Koyfin.

GOOGL’s stock has gained over 12% this year and more than 180% in the last 12 months. 

For updates and corrections, email newsroom[at]stocktwits[dot]com<