Meta's Muse Is a Breakout Success. But There's 1 Glaring Problem With It.

Meta Platforms' (NASDAQ: META) Muse has set the bar for personal AI agents. Just a few weeks after its release, the agentic AI app has more than 10 million users, and 25% of them use it nearly every day.

But the biggest problem with Muse has nothing to do with how the agent works. The issue is that it's simply expensive and unprofitable.

While that type of problem is nothing new in the tech world, it could be a serious one for Meta because there's virtually no way for it to recoup the nearly $145 billion it's spending this year on AI-related capital expenditures by selling Muse subscriptions.

Image source: Getty Images.

Meta is already spending heavily, and Muse will make it worse

Lots of AI companies are troubled by rising AI price tags these days, but Meta's business model is different from those of many of its peers -- and I think that puts the company at a distinct disadvantage.

For example, Microsoft's management recently said the company will spend up to $190 billion on capital expenditures this year, with most of that going to AI infrastructure. That's even higher than Meta's AI capex spending, but there's a big difference between the two: Microsoft is generating significant sales directly related to AI.

In its fiscal 2026 (which ended June 30), the company's Azure cloud computing unit benefited from selling AI services to customers, resulting in Azure's annualized revenue run rate topping $100 billion for the first time ever.

Those sales go a long way toward making Microsoft's hefty AI spending make financial sense. And because it rents AI cloud computing capacity to many customers -- it's the world's second-largest cloud provider -- it is uniquely positioned to continue generating AI-related cloud sales for years to come.

Meta is in a completely different position. It generates nearly 98% of its revenue from advertising. Its social media apps are designed to keep users scrolling, and Meta makes its profits from in-app display ads.

That has worked well for the company so far, but AI is an entirely different type of platform. Users want agents to perform tasks and answer questions. It's much harder to insert ads into a conversation with an AI agent when the entire point of it is to work for you in the background.

It's not that Meta won't figure out a way to put some ads into Muse, but rather that it likely won't be as easy to do as it is on Facebook or Instagram, and that those ads probably won't be as lucrative.

This isn't helping either

In addition to AI agents being more difficult to monetize with ads, they are also expensive to run. Muse has several paid tiers right now, including $20 and $100 monthly subscription options, but most people using it are sticking to the free version, which gives them a limited number of requests.

All AI models spend "tokens" when completing tasks, and higher-end models generally burn through more tokens when they work than older models.

Analysts at Deutsche Bank estimate that Muse's free tier -- which gives each user 100 million AI tokens per month -- probably costs Meta about $40 per user per month. And that's the free version. Higher tiers offering more tokens could cost Meta even more, even though the users are paying for their subscriptions. For example, OpenAI's ChatGPT Pro plan costs users $200 per month, but it can cost the company up to $14,000 in actual token usage, according to SemiAnalysis.

Meta CEO Mark Zuckerberg said on the company's third-quarter earnings call that he expects to eventually start charging businesses fees on the transactions Muse makes. The company isn't doing that yet, and Zuckerberg also said that Meta intends to keep the free version of Muse free to users.

So, for now, Muse costs Meta money, doesn't have ads, and when Muse makes purchases on behalf of its users, Meta isn't getting any fees from the sellers for those transactions. All of that may be great for Muse's users and encourage rapid adoption, but it's bad for Meta when it comes to converting its massive AI spending into profitable income streams.

What's more, there's a lot of competition among AI agents now, especially with OpenAI's dots and Anthropic's Duet. Muse may have raced ahead with users initially, but it's costing Meta financially, and the agentic AI field is about to get very crowded. That all adds up to a big problem for Meta and its shareholders.

Chris Neiger has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Meta Platforms and Microsoft. The Motley Fool has a disclosure policy.