Alphabet stock pops on report it’s developing a more efficient AI chip

Alphabet shares rose 3 percent on Monday Information reported that the company has developed a new server chip internally dubbed “Frozen v2” designed to run Gemini models more efficiently.

According to the news outlet, the chip will permanently embed parts of the Gemini architecture directly into the silicon, thereby reducing the number of calculations and amount of data movement required to answer queries.

In a statement to CNBC, Alphabet said its teams are “constantly researching and testing new innovations to deliver maximum performance and efficiency to our users and customers” and that “while not every project makes it to production, this rigorous research remains at the core of our full-stack approach.”

“By co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads,” the statement continued.

The Information predicts that Google engineers could deliver six to ten times more tokens per unit of power than the company’s latest AI chips, called TPUs, or tensor processing units. Rather than replacing general-purpose TPUs, Frozen will become a more specialized branch of Google’s custom chip portfolio.

According to the report, the company targets 2028 for distribution. The project aims to address a major internal computing shortage that has raised tensions and reportedly forced Google Cloud to turn away external work.

Just last month Google accepted payment SpaceX Approximately $1 billion per month is spent to help bridge the gap and meet enterprise computing commitments.

The trade-off is flexibility. According to The Information, the chip will only work with future Gemini models if Google uses the same underlying architecture. Google is currently reportedly treating Frozen v2 partly as a test run and has no plans to produce it at the same scale as its TPUs.

Google’s AI efforts face a more immediate challenge.

The next Gemini Pro release was delayed, and Google lost several top researchers to rivals as Chinese models gained ground in American businesses. These models now account for 45% of US company token usage.

Competitive pressure is increasing with Moonshot AI’s recent releases and Alibaba’s narrowing the talent gap this weekend.

Google DeepMind chief Demis Hassabis is on Capitol Hill this week to pitch to lawmakers on a FINRA-style watchdog for artificial intelligence that would be federally regulated, largely financed by industry and built to test the most advanced models for national security risks before they are released.

Read the full story at The Information Here.