Meta (META.US) becomes one of Microsoft’s largest AI clients, but demand has yet to break out of the tech industry, highlighting concerns about the future profitability of AI.

According to Zhitong Finance APP, Meta Platforms (META.US) has become one of Microsoft’s (MSFT.US) largest artificial intelligence (AI) clients, underscoring that demand for emerging AI technology remains highly concentrated within the tech sector. According to a source familiar with the matter, Meta spends hundreds of millions of dollars annually to access AI models via Microsoft’s Azure cloud service. The person said that Meta uses trillions of tokens through this platform each week—a unit used to measure AI computing consumption.

An important component of Microsoft’s AI strategy is providing models from different vendors through a marketplace platform called Foundry. As of July, Foundry had 100,000 clients. Microsoft’s promotional materials often feature customers from traditional industries such as manufacturing and transportation.

However, sources say that at present, Microsoft’s largest AI clients are still mostly other tech companies. According to one source, other major clients include Adobe (ADBE.US), AI company Perplexity, and Sierra, a customer service AI startup co-founded by OpenAI chairman Bret Taylor.

Microsoft is particularly reliant on a handful of clients in the technology sector. Its long-term partner OpenAI contributed about 70% of Microsoft’s total AI revenue in the most recent fiscal year. Apart from providing model access services, Microsoft also offers the AI assistant Copilot and rents out AI-oriented computing power.

People familiar with the matter revealed that Meta is making significant investments in AI to support software development. The company purchases model access rights across multiple platforms depending on the availability and cost of different models. One of the insiders said Meta’s developers had accessed OpenAI’s technology through Foundry to help them evaluate the output of self-developed models.

Meta CTO Andrew Bosworth explained on the "Big Technology" podcast in July that, in addition to developing its own AI models, Meta also rents other leading models and incorporates them as part of its development process.

At the same time, Meta is building its own business providing various AI models for clients’ access—a type of API service which may eventually compete with Foundry. This could help Meta cut expenditures on external services such as those from Microsoft.

In the past, Meta also relied on Microsoft’s technology before eventually replacing it with in-house solutions. From the late 2000s, Microsoft’s Bing search engine supported Facebook’s web search. By the end of 2014, Meta had stopped using Bing. In the years before the rise of ChatGPT and the current AI boom, Microsoft also provided Meta with computing power for AI development. Since then, Meta has become one of the world’s largest builders of AI data centers to support its AI model development.

It is worth noting that the fact Microsoft’s largest AI clients are still other technology companies is not necessarily good news for some investors. For AI to truly meet market expectations, the technology must be widely adopted throughout the economy, not just by technologically advanced companies. This concern is especially relevant as the tech sector continues to face challenges from concentrated revenues and cyclical business exchanges among companies in recent years.