Meta plans to launch 'Meta Compute' to lease excess AI capacity and models

Meta is planning to launch a new cloud infrastructure business called "Meta Compute" that will lease excess AI computing power and hosted AI models to outside customers, transforming its massive AI infrastructure investment into a future revenue stream similar to traditional cloud services. The company intends to compete directly with industry leaders like Amazon WebServices, Microsoft Azure, and Google Cloud by offering developers paid access to its proprietary Muse Spark models or selling raw computing capacity akin to neocloud providers. CEO Mark Zuckerberg confirmed that selling unused compute is "definitely on the table," noting that external firms have frequently approached Meta seeking API access or available computing power, though the company has not yet sold access because it still requires the capacity for its own internal AI ambitions. This strategic pivot matters as it allows Meta to generate returns on its colossal spending while artificial intelligence agents and other applications take longer to scale, potentially positioning the company as a major hyperscaler in the AI market. External customers, including developers and enterprises needing AI infrastructure, are the primary group affected by this potential new service offering.