Competition among AI model companies is shifting from “who’s more powerful” to “who can keep their customers.” According to reports, Microsoft and Meta, two of Anthropic’s biggest customers, are both cutting back on employee use of Claude and turning more to their own AI tools.

Specifically, Microsoft has cut its planned spending on Claude internally by more than a third, from an initial estimate of over $1 billion a year. The average monthly budget per employee in its cloud division has been slashed from $100,000 to around $10,000. Meta, meanwhile, has reduced its Claude Code seats from about 60,000 to 30,000. Previously, it spent more than $105 million on the coding tool over a 28-day period.

Why is this happening? On the one hand, AI coding tools are extremely “expensive to run,” and usage-based pricing quickly drives up bills for large companies. On the other hand, Microsoft and Meta have their own models and tools, and promoting their own products internally can save money while also helping them refine those products. For them, it’s a natural choice.

But the timing is awkward for Anthropic. The company is raising tens of billions of dollars in debt financing through partners to buy chips, while also preparing to go public. When two major customers cut back on usage at the same time, it’s bound to raise concerns about the sustainability of its revenue growth. In addition, the U.S. Department of Defense has said it has stopped using Anthropic’s products.

My take: this is a wake-up call for all AI application companies—large customers can be both their biggest source of revenue and their competitors. As model capabilities become increasingly similar, pricing, cost control, and ecosystem lock-in will matter more and more. For investors, rather than betting on a single model company, it may be better to focus on the “picks and shovels” providers that can benefit no matter who wins, such as companies in computing power, cloud services, and electricity.

The content above is for informational purposes only and does not constitute investment advice.

#AI #微软 #Meta