Meta Platforms Plans to Sell Excess AI Compute Capacity

What happened

Meta Platforms revealed it will sell surplus AI computing capacity to external customers through a new offering called Meta Compute. The initiative comes as the company dramatically scales its infrastructure investments, with 2026 capital expenditures projected at $125 billion to $145 billion, nearly double the $72.2 billion spent in 2025. Meta has signed $107 billion in new contractual commitments for multiyear cloud deals and infrastructure purchase agreements. The company is deploying more than one gigawatt of custom silicon developed with Broadcom and committed $6.5 billion to Samsung for its third-generation MTIA accelerator. This strategy mirrors SpaceX's approach, which generates a $26 billion annual run rate by renting compute capacity to customers like Anthropic ($1.25 billion per month) and Google ($920 million per month).

Why it matters

The move addresses investor concerns about Meta's ability to earn meaningful returns on unprecedented AI spending. By selling excess capacity in a supply-constrained market, Meta can generate revenue from its infrastructure investments sooner while maintaining its profitable core business. First-quarter advertising revenue jumped 33% year over year to $56.3 billion, with ad impressions up 19% and pricing up 12%, demonstrating that the company's traditional operations remain strong at a 41% operating margin. The external compute sales model could transform Meta's capital-intensive AI buildout from a financial burden into a revenue stream, particularly as demand for AI compute resources significantly exceeds available supply across the industry.

Bigger picture

Meta's entry into cloud computing reflects broader industry dynamics where AI infrastructure demand far outpaces supply. Major tech companies are paying premium rates for compute capacity-Google's $920 million monthly commitment to SpaceX exemplifies the severity of the shortage. While Meta will compete against established cloud providers like Amazon Web Services, Azure, and Google Cloud, the tight supply environment creates opportunity for new entrants with substantial infrastructure. The company's $201 billion in 2025 advertising revenue at a 41% margin provides financial stability to fund this expansion. Meta's approach of monetizing surplus capacity rather than relying solely on internal use cases represents a strategic shift in how tech giants manage massive AI investments.

What to watch

Investors should monitor Meta's ability to secure external compute customers and pricing terms compared to established cloud providers. The company's execution on its over $107 billion in infrastructure commitments and the deployment timeline for custom silicon will be critical. Watch for updates on Meta Compute's customer pipeline and whether revenue materializes at levels comparable to SpaceX's compute rental model. The sustainability of Meta's core advertising growth-particularly ad impression volume and pricing trends-will determine whether the business can continue supporting the elevated capital expenditure cycle. Any signs of AI compute demand softening or increased competition in the cloud market could affect the viability of this strategy.