Meta Becomes a Major Microsoft AI Customer as Big Tech’s Circular Spending Draws Scrutiny

Meta has become one of Microsoft’s largest AI customers, spending hundreds of millions of dollars annually on Azure services, according to Bloomberg. The arrangement gives Meta access to external AI models and computing capacity even as the company develops its own models, chips and infrastructure.

The scale of the relationship highlights how AI demand remains concentrated among a small group of technology companies. Meta’s Azure usage includes model testing and evaluation, with developers drawing on outside systems, including OpenAI technology, to assess Meta’s own work. A Threads post summarizing the Bloomberg reporting said Meta processes trillions of tokens each week across the service, though that figure was not detailed in the Bloomberg excerpts provided.

Meta’s spending reflects a practical division between building AI systems and operating them at scale. Developing proprietary models and chips gives Meta control over core technology, while Azure provides immediate access to infrastructure and models that can support experiments, comparisons and product development.

Microsoft has positioned Azure as a broad AI platform rather than a service tied only to OpenAI. In 2023, Microsoft added Meta’s Llama 2 models to Azure and Windows, giving developers and organizations access to the model family for research and commercial applications. Meta and Microsoft described that partnership as part of a shared effort to expand access to generative AI, while CNBC emphasized Microsoft’s effort to move beyond an exclusive focus on OpenAI.

The current customer relationship follows an alliance that began before the latest spending surge. Meta selected Azure as a strategic cloud provider in 2022 and tied the agreement to broader collaboration around AI infrastructure and PyTorch, according to Microsoft’s Azure blog.

The partnership expanded publicly with Llama 2. Microsoft integrated the models into Azure’s model catalog and Windows ecosystem, while Meta made Llama 2 available for research and commercial use. The arrangement created benefits on both sides: Microsoft gained another major AI offering for Azure, and Meta gained distribution, cloud integration and a path for developers to run its models through Microsoft’s platform.

Meta’s purchases have renewed scrutiny of circular AI deals, in which major technology companies buy cloud capacity, models or services from one another while investing heavily in their own competing systems. Microsoft earns revenue from Meta’s AI demand, while Meta uses Microsoft’s platform to test systems that could eventually reduce reliance on external providers.

The pattern matters to investors because reported AI growth can reflect transactions inside the same technology group rather than broad adoption by ordinary businesses. Microsoft’s customer base has expanded beyond a single major spender: The Verge reported in 2024 that Adobe and Meta were among Azure OpenAI’s top customers, with several companies spending more than $1 million per month. Forbes’ August 2026 analysis framed the contrast in market terms, with investors rewarding Microsoft for demonstrated AI demand while penalizing Meta amid concerns about the cost and payoff of its AI push.

Meta’s Azure bill underscores the strength of Microsoft’s cloud position and the expense of competing in frontier AI. A company with its own models, chips and large-scale infrastructure still benefits from buying capacity and model access from a rival, especially when rapid testing matters more than owning every layer of the stack.

The arrangement also gives Microsoft a role in Meta’s development process while Meta remains a competitor in models, advertising technology and consumer products. The relationship reflects a wider industry structure in which cloud providers, model developers and platform companies compete in products while collaborating through infrastructure contracts. Meta is also exploring an AI cloud business of its own, according to the Threads summary, adding another layer to the competitive tension.

Microsoft’s immediate priority is turning large AI workloads into durable cloud revenue from more customers. Meta’s spending provides evidence of heavy usage, but the larger test is whether Azure can broaden demand across enterprises and reduce dependence on a handful of technology companies.

For Meta, the next stage is proving that external model access and Azure capacity translate into better products, stronger internal systems or a viable AI infrastructure business. Its partnership with Microsoft gives engineers flexibility today, while its own models and chips preserve strategic control over time. Investors will watch whether that combination produces measurable returns as AI costs continue to rise.