Alphabet shares rose 1.51% Monday after reports emerged that Google is developing a specialized server chip called "Frozen v2," projected to deliver six to ten times the power efficiency of its current TPUs when running Gemini AI models.
Alphabet shares closed 1.51% higher on Monday after reports emerged that the company is developing a new server chip internally dubbed "Frozen v2," designed to run its Gemini AI models with significantly greater power efficiency than its current hardware.
The chip would permanently embed portions of Gemini's architecture directly into the silicon, reducing the volume of calculations and data movement required to process queries, according to The Information, which first reported the project.
Google engineers project the chip could serve between six and ten times more tokens per unit of power compared with the company's newest AI chips, known as TPUs, or tensor processing units.
Alphabet told CNBC in a statement that its teams are "constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers" and that "while not every project moves into production, this rigorous exploration is central to 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 company added.
The chip is not intended to replace Google's general-purpose TPUs. Instead, it would represent a more specialized branch of Google's custom-chip portfolio, targeting 2028 for deployment.
The project is aimed in part at addressing a significant internal compute shortage that has reportedly forced Google Cloud to turn away outside business. Just last month, Google agreed to pay SpaceX nearly $1 billion a month to help bridge that gap and meet enterprise compute commitments.
The approach carries trade-offs. Frozen v2 would be compatible with future Gemini models only if Google maintains the same underlying architecture, limiting its flexibility relative to general-purpose chips. Google reportedly views the project partly as a trial run and does not plan to produce it at the same scale as its TPUs.
The chip development comes as Alphabet faces mounting competitive pressure on multiple fronts. The company's next Gemini Pro release is reportedly delayed, and Google has lost several senior researchers to rivals. Chinese AI models now account for 45% of U.S. company token use, according to the report, with recent releases from Moonshot AI and Alibaba further narrowing the capability gap.
On Capitol Hill this week, Google DeepMind chief Demis Hassabis is pitching lawmakers on a FINRA-style AI watchdog — a federally overseen, largely industry-funded body that would test advanced models for national-security risks before release.
Monday's broader market backdrop added context to the stock's movement. Semiconductor and AI-related equities have faced pressure in recent weeks, with the iShares MSCI USA Momentum Factor ETF — whose top holdings include Micron Technology, AMD, Broadcom, Intel, and Caterpillar — down nearly 12% so far in July. Against that backdrop, Alphabet's single-day gain reflected investor appetite for concrete steps toward cost efficiency in AI infrastructure.
Whether Frozen v2 advances from internal trial to large-scale production will hinge on whether Google maintains architectural continuity in its Gemini line — a strategic bet that could shape the company's compute economics well into the second half of the decade.
Marc Sabatini is a staff writer at TechEchelon covering enterprise software, cybersecurity, and the regulatory beats that shape both. He focuses on the deal flow and policy decisions that move markets.
More from Marc →Chris Fall has resigned as director of the Center for AI Standards and Innovation just three months after the Trump administration appointed him, adding to growing uncertainty over U.S. AI policy leadership as Chinese AI models gain ground domestically.
Moonshot AI and Alibaba unveiled rival frontier AI models this weekend, claiming performance comparable to OpenAI and Anthropic's best systems while making their models publicly available — a move that sharpens the U.S.-China AI rivalry.
Researchers and business school professors say workers who use AI to outsource critical thinking risk professional stagnation, eroded workplace trust, and a weakened sense of professional identity, even as adoption rates surpass 43% in the U.S.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.