Meta becomes one of Microsoft’s largest AI customers

Meta and Microsoft are locked in one of tech’s most fascinating relationships: fierce AI competitors by day, major business partners by the numbers. The arrangement traces back to May 2022, when Meta selected Microsoft Azure as its strategic cloud provider. That deal gave Meta access to Azure NDm A100 v4 virtual machines, powered by roughly 5,400 GPUs, for large-scale AI research workloads.

The spending arms race

Both companies reported quarterly results on July 29, and the contrast was striking. Microsoft posted a 70% increase in quarterly capital expenditure to $41 billion, largely to meet surging demand for cloud and AI services. The market rewarded that spending: Microsoft’s stock climbed on evidence that Azure was successfully converting AI hype into actual revenue.

Meta told a different story. The company reported Q2 revenue of $60.8 billion, a solid 28% year-over-year increase. But profit fell 14%, and free cash flow collapsed 91% to just $784 million. Meta raised its 2026 capex guidance to between $130 billion and $145 billion, a figure that dwarfs previous estimates.

Frenemies with benefits

Meta uses Azure for AI research workloads, which means it’s effectively paying a competitor to help build the technology that powers its own AI ambitions, including the Llama family of models. Meta has invested heavily in its own data centers and GPU clusters, training Llama models primarily on its own hardware. Azure serves as supplementary capacity for research, not as the backbone of Meta’s AI operations.

In early July, reports surfaced that Meta is developing its own cloud infrastructure business, one designed to sell excess AI computing power and models to third parties. If realized, this would put Meta in direct competition with Azure, Amazon Web Services, and Google Cloud.

What the numbers mean for investors

Microsoft has built a flywheel where AI demand flows through Azure, generating revenue that justifies the massive infrastructure spending. Meta’s primary monetization path runs through its advertising business rather than direct cloud services. AI makes Meta’s ad targeting better, which theoretically increases revenue per user, but that’s a longer and less direct path to return on investment than simply selling cloud compute by the hour.

The 91% decline in free cash flow is the number that should keep Meta investors up at night. Revenue growth of 28% is healthy by almost any standard, but when your capital spending grows so fast that it nearly eliminates your free cash flow, you’re essentially asking shareholders to trust that the investment will pay off later.

Meta’s planned cloud business could change the calculus significantly. If the company can monetize its excess AI capacity by selling it to external customers, it would create a second revenue stream that more directly offsets infrastructure costs. But building a credible cloud business from scratch means competing against AWS, Azure, and Google Cloud, three incumbents with years of enterprise relationships and battle-tested platforms.