Meta Spends Millions on Microsoft Azure AI Services

Meta has emerged as one of the largest customers for Microsoft's artificial intelligence services, spending hundreds of millions of dollars annually to access models via the Azure cloud platform. According to reports, the social media giant processes trillions of tokens every single week through Microsoft's infrastructure. Meta's engineering teams specifically leverage OpenAI models hosted on Microsoft's Foundry marketplace to benchmark and evaluate the performance of Meta's own proprietary AI models.

This heavy reliance on a direct competitor's infrastructure highlights a delicate balancing act for AI practitioners. While Meta is currently paying Microsoft to evaluate its technology, the company is simultaneously developing its own API service. This upcoming platform is expected to compete directly with Microsoft's Foundry. Industry observers note this mirrors Meta's historical relationship with Microsoft's Bing search engine, which Meta initially relied upon before replacing it entirely with in-house technology.

The arrangement also sheds light on Microsoft's broader AI revenue streams, where major technology firms make up the bulk of its client base. ByteDance currently leads the list of customers on the Foundry marketplace, followed by other prominent tech firms including Adobe, Perplexity, and Sierra. However, the Foundry marketplace itself represents only a fraction of Microsoft's overall AI business. Approximately 70 percent of Microsoft's AI-related revenue is generated directly by OpenAI, which purchases massive volumes of cloud computing power through Azure.