Meta Platforms has quietly emerged as one of Microsoft’s largest AI customers, highlighting how even companies building their own artificial-intelligence infrastructure continue to rely on external cloud and model providers. According to Bloomberg, Meta is spending hundreds of millions of dollars a year on AI services through Microsoft Azure and consuming trillions of tokens each week, underscoring the scale of computing demand across the technology sector.

Meta’s Azure Spending Shows the Scale of AI Demand

Meta’s reported spending illustrates an important feature of the current AI market: owning large data centers and developing proprietary models does not necessarily eliminate the need for third-party infrastructure. Meta has invested heavily in its own AI systems and computing capacity, yet its developers are using external models and cloud services when availability, performance or economics make them attractive.

Through Microsoft’s Azure Foundry, Meta developers are reportedly accessing multiple AI models and using OpenAI technology to evaluate the output of Meta’s own systems. Bloomberg reported that Foundry had approximately 100,000 customers as of July, but the largest customers remain concentrated among major technology companies rather than traditional enterprises.

Microsoft Gains Another Major AI Revenue Stream

For Microsoft, the Meta relationship reinforces the strategic value of Azure as a neutral infrastructure and model-access platform. Rather than relying exclusively on one model provider, Microsoft offers customers access to a range of AI models through Foundry, allowing developers to select different technologies according to cost, availability and technical requirements.

That model potentially broadens Azure’s addressable market. Microsoft has previously disclosed rapid growth in its Azure AI customer base and has emphasized the availability of third-party models, including Meta’s models, through its cloud platform. The company reported more than 60,000 Azure AI customers in fiscal 2024, up nearly 60% from the prior year.

However, the latest Meta relationship also highlights a structural issue for Microsoft: AI revenue remains concentrated among a relatively small group of sophisticated technology companies. Bloomberg reported that other major customers include ByteDance, Adobe, Perplexity and Sierra, while OpenAI remains particularly important to Microsoft’s AI economics.

Meta Could Eventually Become a Competitor

The relationship has a strategic complication. Meta is reportedly building its own multi-model API service, which could eventually compete directly with Microsoft Foundry. Such a platform would allow Meta to offer access to multiple AI models itself and potentially reduce its dependence on external providers.

For Meta, using third-party models today can accelerate software development and testing while its own AI infrastructure expands. For Microsoft, however, the long-term question is whether major customers such as Meta remain primarily customers or eventually become competitors within the AI-services ecosystem. Meta has previously shifted from third-party technology toward internally developed systems, demonstrating that large technology companies can change suppliers when scale and economics justify the transition.

Going forward, the market will be watching whether Meta’s Azure spending continues to expand alongside its internal AI investments, and whether Microsoft can diversify AI revenue beyond a small group of technology customers. The development of Meta’s own API platform is also important, as it could reshape competitive dynamics between cloud providers, model developers and AI infrastructure companies. More broadly, the scale of Meta’s reported token consumption suggests that AI computing demand remains substantial, but the durability of that demand will increasingly depend on whether AI spending spreads beyond the technology sector into a broader range of industries.