Key Highlights
- Alphabet shares advanced 1.51% Monday following news of a specialized server processor dubbed “Frozen v2” under development
- The processor would hardwire portions of Gemini’s framework directly into hardware to reduce energy consumption
- Internal projections suggest efficiency gains of 6-10x compared to existing tensor processing units in token delivery per power unit
- Target launch timeline points to 2028, though specifications remain under refinement
- Development aims to address internal computing capacity constraints that have led to rejected cloud customer requests
Shares of Alphabet (GOOGL) climbed 1.51% during Monday’s trading session after The Information disclosed that Google is creating a specialized server processor codenamed “Frozen v2.”
The processor is engineered to execute Gemini AI operations with greater efficiency by hardcoding specific elements of the model’s framework permanently into the chip itself. This design philosophy minimizes computational overhead and data transfer requirements when processing user queries.
Internal engineering estimates indicate the specialized chip could process six to ten times more tokens for each unit of electrical power consumed versus Google’s current tensor processing unit technology, based on reporting from The Information.
In a statement to CNBC, Google emphasized that its engineering teams are “constantly researching and experimenting with new innovations” and noted the company’s practice of co-developing hardware and software “from the ground up” to maximize performance for actual deployment scenarios.
Addressing Critical Computing Bottlenecks
The initiative directly responds to substantial internal processing capacity limitations. These constraints have sparked internal friction at Google and allegedly resulted in Google Cloud rejecting potential enterprise clients due to resource unavailability.
As recently as last month, Google entered into an agreement to compensate SpaceX nearly $1 billion monthly to secure additional computing resources and fulfill corporate customer obligations.
Frozen v2 is intended to complement Google’s current tensor processing unit portfolio rather than serve as a replacement ā functioning as a purpose-built solution for particular workloads.
The approach involves compromises. The chip would only support subsequent Gemini iterations if Google maintains architectural consistency. Company leadership currently regards the project partially as an experimental endeavor and doesn’t intend to manufacture it at the volume levels of standard TPUs.
The rollout timeframe targets 2028, although engineering teams continue to refine the technical specifications.
Mounting Competitive Challenges
This chip development emerges amid significant artificial intelligence headwinds for Google. The upcoming Gemini Pro version has experienced postponement after failing to achieve internal performance benchmarks, Bloomberg has reported.
Google has additionally experienced the departure of multiple senior research personnel to competing organizations in recent months.
Chinese artificial intelligence platforms are expanding their footprint among American enterprises, now representing 45% of token consumption by U.S. companies. Recent product launches from Moonshot AI and Alibaba over the weekend are further closing the technological performance differential.
Google DeepMind CEO Demis Hassabis appeared before congressional leaders this week advocating for a FINRA-modeled federal oversight body for artificial intelligence, primarily industry-funded, that would evaluate advanced systems for national security vulnerabilities prior to public deployment.
Alphabet shares extended gains by an additional 0.97% during extended trading hours Monday.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
