Google to Make Gemini Up to 10x More Efficient With New Chips

Alphabet is reportedly designing a new server chip to make Google’s Gemini AI models run more efficiently.

The chip is internally called “Frozen v2” and is expected to launch sometime in 2028, The Information reported, citing anonymous sources.

According to the report, the chip could be six to 10 times more efficient than Google’s current AI chips when measured by the number of tokens generated per unit of power.

Google Did Not Deny It

Google did not directly confirm the report in a statement to TechCrunch. However, it also did not deny it.

The company said its teams constantly research new ways to improve performance and efficiency for users and customers. It added that not every project reaches production, but this work is part of Google’s full-stack approach to building hardware and software together.

Google said this approach helps it optimize systems for real-world workloads.

AI Firms Want Their Own Chips

The reported chip comes as major AI companies try to reduce their reliance on Nvidia.

Nvidia still dominates the AI chip market, making companies heavily dependent on its hardware. At the same time, global demand for AI computing power remains high, and supply shortages have pushed companies to develop their own processors.

Custom chips can also help AI companies lower power use and improve efficiency, which has become more important as investors question the rising cost of AI infrastructure.

OpenAI and Anthropic Moving Into Chips

Google is not the only AI company working on custom hardware.

In June, OpenAI announced its first custom chip, an inference processor called Jalapeño. Earlier this month, reports also said Anthropic was discussing a possible chipmaking partnership with Samsung.

These moves show how AI companies are trying to control more of their hardware stack as model training and inference become more expensive.

Investors Are Watching AI Spending

Alphabet has already faced investor concerns over the cost of its AI strategy.

Earlier this year, Google said it planned to spend between $180 billion and $190 billion. With that level of investment, the company needs to show that its AI spending can produce long-term returns.

A more efficient server chip could help Google reduce the cost of running Gemini models and improve margins over time.

Alphabet Stock Rises

The report appeared to reassure investors ahead of Alphabet’s upcoming earnings report.

After The Information published the report, Alphabet’s stock rose about 3 percent on Monday morning.

For now, Frozen v2 remains an unconfirmed project. However, the report suggests Google is preparing for the next stage of AI competition, where power efficiency and custom hardware may become just as important as model performance.