1 hour ago
18
Mark Zuckerberg has spent the better part of the last two years convincing Wall Street that pouring tens of billions into AI infrastructure would eventually pay off. Now he’s pitching the receipts, or at least the plan for generating them: a dedicated AI cloud business that would sell Meta’s excess computing power to outside companies.
The company’s AI capital expenditure guidance for 2026 sits between $125 billion and $145 billion. And if you’re spending that kind of money on compute, you’d better find more than one way to use it.
From social network to cloud contender
Zuckerberg first floated the idea publicly at Meta’s shareholder meeting on May 27, saying that entering the cloud market was “definitely on the table.” By July 9, the language had firmed up considerably, with the CEO confirming that the company was actively exploring an AI cloud business and arguing that selling excess computing power simply “makes sense.”
The company is reportedly in early discussions for a potential two-year, $10 billion deal to lease computing capacity to Anthropic, the AI safety startup behind the Claude family of models. If that deal materializes, it would represent one of the largest single cloud computing agreements in the industry and serve as a major proof of concept for Meta’s cloud ambitions.
Taking on the incumbents
Entering the cloud business means going toe-to-toe with Amazon Web Services, Microsoft Azure, and Google Cloud. These three control the vast majority of the global cloud infrastructure market and have spent decades building the enterprise sales teams, compliance frameworks, and service-level agreements that large customers expect.
Meta’s pitch is different, though. Rather than competing as a general-purpose cloud provider, the company appears to be positioning itself specifically as an AI compute platform. Its advanced data centers and proprietary AI models, including successors to the open-source Llama family and the Muse Spark series, could give it a differentiated offering that appeals to AI-native companies hungry for raw computing power.
During Meta’s Q2 2026 earnings call on July 29, Zuckerberg pointed to increasing external demand for Meta’s compute resources and noted that discussions with potential customers were happening at premium rates.
What this means for investors
The $125 billion to $145 billion capex range for 2026 alone is roughly equivalent to the entire GDP of Kuwait. Every earnings call has featured some version of the same question: when does this spending start generating returns?
The potential Anthropic deal is particularly significant as a signal. Anthropic is one of the most well-funded AI companies in the world and has existing relationships with both Amazon and Google. If it’s willing to commit $10 billion over two years to Meta’s infrastructure, that validates the quality of what Meta has built and could open the door for similar agreements with other AI firms.
That said, execution risk is real. Meta’s core advertising business, which still generates the overwhelming majority of its revenue, faces its own set of pressures. Building out an enterprise cloud operation requires hiring sales teams, establishing customer support infrastructure, meeting security and compliance requirements, and managing service reliability at a level Meta has never had to deliver for external clients.
There’s also the strategic tension inherent in the model. Some of the companies that might want to lease Meta’s compute are also competitors in the AI space. Leasing capacity to Anthropic, for example, means powering a company whose products compete with Meta’s own AI offerings.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·