Meta is preparing to launch a cloud computing business that will sell access to its enormous inventory of AI training and inference chips, multiple sources confirmed Tuesday. The move would transform the social media giant into a direct competitor to Amazon Web Services, Google Cloud, and Microsoft Azure in the rapidly consolidating AI infrastructure market.
Meta has spent over $30 billion building its AI infrastructure, including hundreds of thousands of NVIDIA H100 and next-generation GPUs. But with much of that capacity sitting idle between major training runs — and with CEO Mark Zuckerberg signaling the company may slow its breakneck CapEx pace — selling excess compute to enterprise customers represents a logical path to monetization.
Analysts note that Meta's entry into cloud computing could reshape pricing dynamics. The company's existing data center footprint, combined with its custom silicon efforts (including the MTIA inference chips), gives it a cost structure that pure cloud providers may struggle to match. However, Meta has no enterprise sales organization, no cloud management console, and no track record of serving external customers — gaps that will take years to fill.
The news sent shares of major cloud providers slightly lower in after-hours trading, though several analysts cautioned that Meta's cloud ambitions are unlikely to threaten incumbents in the near term. 'This is a 2028 story, not a 2026 one,' wrote a Morgan Stanley analyst. 'But it absolutely changes the long-term competitive landscape for AI compute.'
Source context: Quartz
