Meta is spending $145B on AI infrastructure — and selling the leftovers

meta is following spacex's playbook by establishing a cloud business to rent out its spare ai compute capacity to outside customers, according to a report by bloomberg. this strategy indicates that meta's internal ai models, including recent efforts like muse spark, are not consuming the company's massive gpu inventory, which required $145 billion in infrastructure spending this year. by offering both compute and access to models, meta joins spacex in monetizing hardware originally acquired for proprietary training. the news boosted meta's stock by 10%, signaling investor approval of the move to generate revenue from excess capacity. this shift highlights the growing trend of hyper-scalers turning into cloud providers to support capital-heavy ai investments.