Alphabet shares rise on reports of custom 'Frozen v2' AI chip
Google is developing a new server chip to improve the performance of its large language models, according to a report by The Information.
Alphabet is designing a new server chip, internally known as 'Frozen v2', to enhance the efficiency of its Gemini large language models. According to a report by The Information, the chip is projected for release in 2028 and could offer six to ten times greater efficiency than current hardware, measured by tokens generated per unit of power.
Google has neither confirmed nor denied the existence of the project. In a statement to TechCrunch, the company emphasised its "rigorous exploration" and "full stack approach" without directly addressing the Frozen v2 project. The firm stated that its teams are constantly researching innovations to deliver maximum performance, noting that co-designing hardware and software ensures systems are integrated and optimised for real-world workloads.
The development occurs within a broader industry trend where major tech firms are pursuing custom silicon to mitigate reliance on Nvidia and manage escalating AI infrastructure costs. OpenAI announced its first custom chip, "Jalapeño," in June, while Anthropic was reported to be discussing a new chipmaking partnership with Samsung.
Efficiency has become a key selling point for tech companies as concerns about AI spending have dampened market euphoria. Alphabet plans to spend between $180 billion and $190 billion on its AI strategy, leading to investor scrutiny regarding the return on these investments. The news contributed to a 3% rise in Alphabet's stock on Monday morning, ahead of its upcoming earnings report.
The timeline for release and the specific efficiency metrics remain unverified claims from the report. However, the market reaction suggests that investors are viewing the potential for improved hardware efficiency as a positive indicator for the company's capital allocation strategy.
