Meta is spending heavily on Microsoft's AI infrastructure, even as it develops technology that could one day reduce that dependence. According to Bloomberg, Meta has become one of Microsoft's biggest AI customers through its use of Azure and AI models available through Microsoft's Foundry marketplace.
The arrangement shows how intertwined the AI industry has become. Companies building major AI systems may still need rival platforms for access, testing, benchmarking, and compute-heavy workflows.
Meta is a major Azure AI customer
Meta spends hundreds of millions of dollars a year accessing AI models through Microsoft's cloud service Azure. The scale of usage is large: Meta burns through trillions of tokens on the platform every week.
That spending makes Meta one of Microsoft's most important AI customers. It also makes the relationship more complex than a standard cloud contract. Meta is not just buying ordinary software capacity; its engineers are using Microsoft-linked AI infrastructure as part of the work surrounding Meta's own models.
The source describes OpenAI models available through Microsoft's Foundry marketplace as one of the tools Meta's engineers use. Their purpose is to benchmark the performance of Meta's own models.
Benchmarking matters because it gives engineers a way to compare systems under similar conditions. In this case, Meta can use models accessed through Microsoft to evaluate how its own AI models perform. The source does not describe the exact tests, but it makes clear that Microsoft is part of the toolchain Meta uses in that process.
Foundry sits at the center of the relationship
Microsoft's Foundry marketplace gives customers access to AI models, including OpenAI models. For Meta, Foundry is one route to the external systems its engineers use while measuring the performance of internal work.
That creates a notable dynamic. Microsoft is selling AI services to a company that is also building its own AI capabilities. Meta benefits from access to the marketplace, while Microsoft benefits from Meta's spending and usage.
At the same time, the source says Meta is building its own API service that could end up competing with Foundry. That means today's customer relationship could become more directly competitive if Meta's API service develops into an alternative for developers or companies seeking AI model access.
This is not presented as a sudden break. It is a pattern the source compares to what happened with Bing. Meta once relied on Microsoft's search engine, then eventually replaced it with its own tech.
A familiar pattern for Meta and Microsoft
The Bing comparison is important because it frames the current Azure AI relationship as potentially transitional. Meta can rely on Microsoft services while it builds its own systems. Later, if Meta's internal technology matures enough, the company may shift more work onto its own stack.
That does not mean the current Microsoft relationship is unimportant. The opposite is true. Spending hundreds of millions of dollars a year and using trillions of tokens every week suggests that Microsoft's AI services are deeply useful to Meta's engineering work right now.
It also shows why Microsoft's AI business is not limited to one type of customer. Some customers may be startups, some may be software companies, and some may be major tech companies that also compete in AI. The source says most of Microsoft's biggest AI customers are other tech companies.
For Microsoft, this makes Azure and Foundry strategically valuable. Even companies with serious AI ambitions may still need Microsoft infrastructure or model access. For Meta, it keeps external tools available while internal services continue to develop.
Microsoft's AI revenue remains concentrated
Meta is large, but it is not the only major name in Microsoft's AI customer base. ByteDance tops the list of Foundry customers, followed by Adobe, Perplexity, and Sierra.
Still, Foundry is only one part of Microsoft's AI business. The source says roughly 70 percent of Microsoft's AI revenue comes from OpenAI alone. OpenAI mainly buys massive amounts of compute through Azure.
That distinction matters. Foundry brings in notable customers, including Meta, ByteDance, Adobe, Perplexity, and Sierra. But the source makes clear that Foundry accounts for only a slice of that business, while OpenAI's compute spending through Azure dominates Microsoft's AI revenue picture.
So Meta's spending is significant, but it sits inside a broader Microsoft AI strategy. Azure provides the cloud foundation. Foundry provides a marketplace for model access. OpenAI contributes the largest share of AI revenue described in the source.
What this says about the AI market
The clearest takeaway is that AI competition does not prevent AI dependence. Meta can be a major Microsoft customer while developing an API service that could compete with Foundry. Microsoft can sell AI access to companies that may later reduce reliance on its services.
That kind of overlap is likely to shape how large AI companies work with each other. The source does not suggest a simple divide between partners and competitors. Instead, it shows companies using each other's platforms where useful, even when their long-term products may collide.
For readers watching Meta, Microsoft, Azure, Foundry, and OpenAI, the relationship is a useful signal. The largest AI efforts still require access to powerful services, model marketplaces, and compute. The companies building the future of AI are also some of the biggest customers of the infrastructure behind it.