Meta’s $145B AI Spend Could Crack the Cloud Oligopoly for SaaS Firms

Meta, the social media giant behind Facebook and Instagram, is in early-stage talks with AI startup Anthropic about leasing a significant portion of its AI-focused data center capacity. The potential two-year deal, first reported by the New York Times and confirmed by a source to CNN, could be worth as much as $10 billion—though CNN's source cautioned that any specific numbers are speculative. If consummated, the arrangement would mark Meta's dramatic entry into the cloud computing market, putting it in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud. The news lands amid Meta's unparalleled infrastructure buildout. The company plans to spend between $125 billion and $145 billion in capital expenditures this year, nearly double the prior year's outlay, largely to support AI training and inference. That historically high investment has come with pain: in April 2026, Meta laid off 10% of its workforce, roughly 8,000 employees, partly to offset the infrastructure cost. CEO Mark Zuckerberg had foreshadowed the possibility of leasing excess capacity, telling shareholders in May 2026 that outside companies "almost every week" inquire about buying compute at a premium. Yet he emphasized that Meta would only sell capacity if it found itself overbuilt, saying, "We haven't done that yet because we think that we have a use for the compute." The Anthropic discussions suggest that calculus may be shifting. Anthropic, the maker of the Claude family of AI models, already has multibillion-dollar cloud deals with Amazon and Google. Diversifying its compute supply to include Meta could reduce its dependency on any single provider, potentially improving negotiating leverage and cost stability. For Meta, the move would transform a massive cost center into a possible revenue stream, with industry analysts noting that even a fraction of its AI infrastructure capacity rented out could generate billions annually. The cloud computing implications are substantial. AWS, Azure, and Google dominate the market, but Meta's entry—if it scales beyond Anthropic—could introduce new pricing dynamics. Meta's infrastructure, built around its custom AI chips and state-of-the-art data centers, might offer a compelling alternative, particularly for the largest AI workloads. However, the talks remain preliminary, and the deal is far from certain. Any agreement would require Meta to be confident it does not need the capacity for its own AI ambitions, while Anthropic would need to ensure performance, security, and interoperability across a heterogeneous compute environment. The optics are also delicate: leasing to a competitor could raise questions about data privacy and competitive moats. Beyond the immediate parties, the development underscores the staggering capital intensity of leading-edge AI. Companies are spending hundreds of billions to build infrastructure that may, at times, be surplus to their own needs—a dynamic that could reshape how AI labs fund and access the raw fuel of modern machine learning. If Meta successfully monetizes its AI infrastructure, other large tech firms may follow suit, blurring the line between consumer platforms and cloud providers. For now, the market watches: a working deal would validate the idea that AI compute is a liquid, tradable asset, not just a proprietary cost sink.