Meta Platforms to Launch Cloud Infrastructure Business to Monetize Massive AI Spending

Meta Platforms is preparing to enter the cloud computing market by selling excess AI computing power and access to its AI models to external customers. This strategic pivot aims to monetize the company's massive capital expenditures, which are projected to reach $135 billion this year, by diversifying revenue beyond its core advertising business. The move positions Meta as a direct competitor to hyperscale titans such as Amazon Web Services, Microsoft Azure, and Google Cloud.

Meta's decision to launch a cloud business follows growing investor concern over the company's escalating infrastructure costs, which rose from $37.2 billion in 2024 to a projected $135 billion in 2026. While hyperscalers like Microsoft and Amazon have long justified similar spending through their established cloud divisions, Meta previously relied almost exclusively on its advertising business to recoup these investments. The new venture, confirmed by reports and internal discussions, allows Meta to offer its massive compute capacity to third parties, potentially transforming the company into a diversified enterprise services provider and easing the stock's primary overhang.

The proposed business model may take two forms: a 'bare metal' infrastructure rental service or a full-service platform similar to AWS Bedrock. Analysts suggest that Meta is evaluating both renting out raw computing capacity—akin to 'neocloud' providers like CoreWeave—and hosting its own AI models for developer use. This move places Meta in direct competition with established cloud giants and emerging players like SpaceX, which recently secured a $920 million monthly deal with Google for computing power. CEO Mark Zuckerberg previously indicated that such a move was 'on the table' to ensure the company's balance sheet remains robust despite heavy AI investments.

Despite the positive market reaction, which saw Meta's stock jump 9% to $617 per share, industry experts highlight significant hurdles for the new division. Building a cloud service for external workloads requires complex software layers and developer tools that established providers have spent years perfecting. Furthermore, Paul Meeks of Freedom Capital Markets noted that AI labs like OpenAI or Anthropic might be hesitant to host sensitive data on infrastructure owned by a direct competitor. However, with the current insatiable demand for AI compute, analysts like Ben Bajarin believe the sheer need for hardware may outweigh these competitive concerns in the short term.