Mark Zuckerberg has confirmed that Meta will sell off surplus computing capacity from its data centers as the company scales up its artificial intelligence infrastructure. The disclosure, made during an earnings call, signals a strategic shift in how the social media giant manages the enormous power demands of training and running advanced models. According to a report on Yahoo Finance , Zuckerberg framed the decision as a practical response to fluctuating internal needs and the high costs of maintaining idle hardware. Meta has poured tens of billions of dollars into building one of the largest clusters of graphics processing units in the industry. The company now operates hundreds of thousands of Nvidia chips across multiple facilities, with plans to expand that footprint dramatically in the coming years. Much of this investment supports the development of Llama, Meta’s open-source large language model family, along with the recommendation systems that power Facebook, Instagram, and Threads.…