财联社·2025-11-25 12:07Core Insights - Google is intensifying its competition in the AI chip market to challenge Nvidia's dominance, with Meta being a potential major customer for Google's TPU (Tensor Processing Unit) [2][3] - The TPU, designed specifically for AI tasks, has undergone seven iterations since its internal deployment in 2015 and is now being marketed for customer data center integration [2][3] - If the deal with Meta is finalized, it would significantly affirm Google's hardware ambitions and could help capture 10% of Nvidia's annual revenue share [6][7] Group 1 - Google is promoting its self-developed TPU to clients for deployment in their data centers, marking a strategic shift [3] - Meta is reportedly considering investing billions to integrate Google's TPU into its data centers starting in 2027 and plans to rent TPU computing power from Google Cloud as early as next year [3][4] - Following the news, Alphabet's stock rose by 4% in after-hours trading, while Nvidia's stock fell by nearly 4% [5] Group 2 - Currently, Meta's AI infrastructure primarily relies on Nvidia's GPUs [4] - Google has informed potential clients, including high-frequency trading firms and large financial institutions, that deploying TPU locally can help meet strict security and compliance requirements for sensitive data [6] - The push for TPU deployment is seen as a response to the surging demand for AI computing power and signifies a more intense phase in the AI chip competition [7]