Meta Platforms Inc. has emerged as one of Microsoft Corp.'s biggest AI customers, demonstrating how demand for artificial intelligence technology remains concentrated within the tech industry. According to reports from Bloomberg, the social media company is spending hundreds of millions of dollars annually to access artificial intelligence models through Microsoft's Azure cloud service. Each week, Meta uses trillions of tokens - the unit of consumption for AI computing - via the platform, as reported by people familiar with the matter. The arrangement underscores a persistent reality in the AI boom: demand for the technology remains overwhelmingly concentrated within the tech industry itself.
A major component of Microsoft's AI strategy involves offering models from various providers through a marketplace called Foundry, which had 100,000 customers as of July. As reported by Bloomberg, Microsoft's AI strategy focuses on providing models from multiple vendors rather than developing its own proprietary technology. The company also offers an AI assistant called Copilot and rents out AI-optimized computing resources to customers. For Meta, the same figure represents roughly 0.4% of the midpoint of its 2026 capital expenditure guidance. The relationship is particularly interesting because Meta is already spending significantly on its own models and data centers, but still rents third-party AI tools when it makes sense from an availability or economics standpoint.
While Microsoft's promotional materials often feature clients from traditional industries such as manufacturing or transportation, the largest AI customers remain primarily tech companies. According to people familiar with the matter cited by Bloomberg, ByteDance, the Chinese maker of social media app TikTok, has generally been the biggest spender on Foundry. Other major customers include Adobe Inc., AI search company Perplexity, and Sierra, the customer service AI startup co-founded by OpenAI Chairman Bret Taylor. The company uses the service primarily to support its software development efforts, with developers tapping external models to build applications and benchmark the performance of Meta's own in-house AI systems.
Meta is investing heavily in AI to assist its software development efforts, purchasing access to models through multiple platforms based on availability and cost. As reported by Bloomberg, developers at Meta have used OpenAI technology through Foundry to help assess the output of their own models. Chief Technology Officer Andrew Bosworth explained during a July podcast appearance that Meta rents other leading models as part of its development process. The company is also building its own business to sell access to various AI models as an API service, which could potentially reduce spending on external services like those from Microsoft. Meta developers have also used OpenAI models through Foundry to evaluate the output quality of Meta's proprietary models, according to one person familiar with the arrangement.
Meta's reliance on Microsoft has historical precedent, as in the late 2000s, Facebook used Microsoft's Bing search engine to power web search functionality on the social network before fully discontinuing it by the end of 2014. Similarly, Microsoft provided Meta with computational infrastructure for AI research in the years before ChatGPT's launch. Since then, Meta has transformed into one of the largest builders of AI data centers globally, with the company expecting to spend between $130 billion and $145 billion on capital expenditures in 2026, largely on AI and data center capacity. The potential downsides of revenue concentration and circular business dealings have been concerns for the tech industry, as AI must be widely adopted across the economy rather than just by sophisticated tech firms. Microsoft remains especially reliant on a handful of tech sector customers, with OpenAI providing about 70% of Microsoft's overall AI revenue in its most recent fiscal year.
