Alphabet, the parent company of Google, is working on a new server chip designed specifically to improve the efficiency of its proprietary Gemini artificial intelligence models. The development signals a broader push within the company to optimize its AI infrastructure from the ground up rather than relying solely on third-party hardware suppliers.
The chip, known internally as "Frozen v2," is reportedly targeted for release in 2028, according to a report by The Information, which cited people familiar with the matter who were not authorized to speak publicly. Based on those accounts, the chip could deliver efficiency gains of roughly six to ten times over Google's current generation of AI chips, with efficiency measured by the number of tokens the chip can generate per unit of energy consumed.
Google did not offer a direct confirmation when approached for comment, nor did the company dispute the claims. In a statement, Google described a broad commitment to hardware and software development without addressing the specific project by name.
"Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers. 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 Industry Push Toward Custom AI Silicon
The move is part of a wider trend across the technology industry, where major AI-focused companies have been investing heavily in designing their own chips rather than depending exclusively on commercial suppliers. The motivations are twofold: custom chips can be fine-tuned to run proprietary AI models with greater efficiency, and they help companies navigate persistent constraints in the global supply of AI computing hardware.
Efficiency has become a particularly important metric as enthusiasm in financial markets around AI spending has given way to more measured scrutiny. Investors and analysts have begun pressing companies to demonstrate tangible returns on the enormous capital being directed toward AI infrastructure. Against that backdrop, the ability to do more with less power has become a meaningful competitive differentiator.
There is also a strategic dimension to this trend. For years, Nvidia has held a commanding position in the market for AI accelerator chips, and that dominance has meant that the largest AI developers have been heavily reliant on a single supplier for a critical component of their operations. Building proprietary chips offers a path toward reducing that dependency, something that has become a priority for several of the industry's biggest players.
Competitors Also Moving Into Custom Chip Development
Google is not alone in this effort. In June, OpenAI unveiled its first custom-designed chip, an inference processor referred to internally as Jalapeño, marking the company's first significant step into proprietary silicon. Meanwhile, Anthropic has reportedly been in discussions with Samsung about establishing a new chipmaking partnership, according to a report published earlier this month. These moves collectively suggest that the leading AI laboratories are treating custom hardware development as a long-term strategic necessity rather than an optional enhancement.
Investor Scrutiny Over Alphabet's AI Spending
The news about Frozen v2 arrives at a moment when Alphabet is under considerable pressure to justify its AI-related capital commitments. Earlier this year, the company indicated it plans to spend between $180 billion and $190 billion as part of its broader effort to build out the infrastructure required to support its AI ambitions. That level of expenditure has drawn concern from investors who want assurance that the spending will eventually translate into proportionate returns.
The prospect of a chip that dramatically improves efficiency appears to have offered some reassurance to the market. Following publication of the report by The Information, Alphabet's stock rose approximately 3 percent on Monday morning, providing a positive signal ahead of the company's scheduled earnings report later in the week. The market reaction suggests that investors view advances in hardware efficiency as a meaningful indicator of Alphabet's ability to manage costs and sustain its AI strategy over the long term.
If Frozen v2 reaches production as described, it would represent a significant step forward in Google's effort to build a tightly integrated hardware and software stack - one where its chips, systems, and AI models are developed in concert rather than assembled from independently sourced components. Whether the chip ultimately makes it from internal experimentation into large-scale deployment remains to be seen, but the reported efficiency figures, if accurate, would mark a notable leap in the economics of running large-scale AI workloads.
