Meta plans cloud business to sell excess AI compute capacity

Meta plans to launch a cloud infrastructure business called Meta Compute, selling access to AI compute power and models, according to Bloomberg. The move pits Meta against AWS, Google Cloud, and Azure. Meta has committed $182.9B to AI infrastructure, with massive data centers under construction in Louisiana and Ohio — the latter described by Zuckerberg as the size of Manhattan. The initiative is led by Santosh Janardhan, Daniel Gross, and Dina Powell McCormick. Meta may offer both raw compute capacity (like CoreWeave) and hosted AI models (like AWS), including its closed-weight Muse Spark model.

This strategy mirrors SpaceX/xAI, which recently sold all compute capacity at its Colossus 1 data center to Anthropic. Meta's decision signals that the winners of the AI race may be those owning the infrastructure, not necessarily the best models. However, skeptics warn that the AI infrastructure boom could create a bubble — Nvidia's chips depreciate rapidly, and end-user demand may not justify trillion-dollar bets. Meta hasn't yet seen significant standalone revenue from its AI products like Llama or Meta AI, making this cloud pivot a crucial way to monetize its enormous spend.