Meta Platforms Building Cloud Business to Sell Excess AI Capacity

What happened

Meta Platforms is developing a cloud computing business, internally called Meta Compute, to sell excess AI capacity to third-party customers. The service would offer two components: hosted access to AI models (including its unreleased Muse Spark model) where developers pay for computing power used, and raw AI computing capacity similar to neocloud providers like CoreWeave. At Meta's shareholder meeting in May, CEO Mark Zuckerberg indicated that entering cloud computing was on the table, noting that companies approach Meta almost every week to buy access to its AI models or spare computing power. The plans remain in development and could change. Meta is projected to spend as much as $145 billion on AI infrastructure in 2026, representing a significant portion of Big Tech's over $700 billion outlay on the technology.

Why it matters

This strategy could open a new revenue stream for Meta beyond its core advertising business while helping offset massive AI infrastructure costs. The company would be entering a market dominated by Amazon Web Services, Microsoft Azure, and Google Cloud, potentially tapping into booming demand for AI services from businesses. For investors, the initiative reframes Meta's heavy capital expenditure programme as a potential revenue generator rather than purely a cost centre. However, the move also deepens questions about Meta's ability to catch up with leading AI labs like Anthropic, an effort CEO Mark Zuckerberg has backed with billions in investment. Meta unveiled Muse Spark in April, its first AI model from a costly team it assembled, but has yet to release it to developers with no scheduled launch date.

Bigger picture

Meta's entry into cloud computing would intensify competition in an already crowded market led by hyperscalers with established customer bases and infrastructure. The move particularly threatens neocloud providers like CoreWeave and Nebius, which rely on Meta as a key customer while now facing it as a competitor. CoreWeave shares fell 10.8% and Nebius dropped 12.4% on the news, reflecting concerns that Meta could reduce spending on their services. CoreWeave has a $21 billion multi-year agreement with Meta through December 2032, but Meta's pivot from buyer to seller could reshape that relationship. The broader implications extend to the economics of AI infrastructure investment, with some analysts questioning whether U.S.-based data centre capacity will remain competitive against lower-cost alternatives emerging in regions like the Gulf. The initiative mirrors similar moves by SpaceX's xAI unit, which recently struck deals to rent out access to its Memphis data centres to Anthropic and Google.

What to watch

Key developments to monitor include whether Meta formally launches the cloud service and on what timeline, how existing neocloud customers like CoreWeave respond strategically, and whether Meta releases its Muse Spark AI model to developers as part of the offering. Investors should watch for details on pricing, service structure, and whether Meta can successfully balance building its own AI capabilities while competing with established cloud providers. The company's capital expenditure guidance updates will be important, as will any changes to its relationships with existing infrastructure partners. Additionally, watch for competitive responses from Amazon, Microsoft, and Google, and whether other hyperscalers with excess AI capacity follow Meta's lead.