For about twenty years, Meta cost you nothing. Facebook, Instagram, WhatsApp: you paid with your attention, advertisers paid with cash, and that was the whole deal. You, the user, were what Meta sold to advertisers.
That arrangement is starting to wobble. Meta has now poured more than $14 billion into building its own artificial intelligence, and the bill for that has to land somewhere. According to reporting from CNBC and The Wall Street Journal, the place it might land is on you.
Mark Zuckerberg, by one account, wants to charge Meta's roughly 3.5 billion daily users a monthly fee for extra AI features. The free internet you grew up with is quietly asking for your credit card.
How Meta ended up here
A year ago, Meta spent over $14 billion to bring in Alexandr Wang and a group of his top engineers from a company called Scale AI. The goal was to fix Meta's AI effort, which had fallen behind OpenAI (the maker of ChatGPT), Anthropic (the maker of Claude), and Google. Wang now runs a division called Meta Superintelligence Labs.
The trouble started earlier. Meta's first big bet was a family of AI models called Llama, which it gave away for free so outside developers could build on it while competitors charged for the same thing. In April of last year, the newest version, Llama 4, landed with a thud. Developers were not impressed, and Zuckerberg started rethinking the whole plan. Two months later came the $14.3 billion deal for roughly half of Scale AI, and with it, Wang and his lieutenants.
The payoff arrived this past April: a model called Muse Spark. It was Meta's first proper homegrown AI of this kind, and a step away from the give-it-away approach. The catch is what Muse Spark was built for. Instead of courting outside developers, it was designed to live inside the apps you already use, Facebook and Instagram, plus the Ray-Ban Meta glasses and Meta's standalone AI app.
The part that costs you money
Building an expensive AI is one thing. Getting people to pay for it is another, and that is the problem Zuckerberg now owns.
Meta has spent its entire life as an advertising company. Even today, ads bring in 98 percent of its money. The AI it built so far has mostly made those ads work better, which helps the business but does not prove that anyone will actually pay for an AI product on its own. Wall Street has noticed. Meta's stock is down 18 percent over the past year, one of the worst showings among the giant tech companies, even after the company grew its revenue 33 percent in a single quarter, its fastest pace since 2021.
So Meta is hunting for a second income stream. It has rolled out new AI subscription plans, and per The Wall Street Journal, Zuckerberg wants a monthly fee from regular users for extra features, plus an AI assistant it can sell to companies. The history here is not encouraging. Meta has tried to make money outside of ads before and mostly failed. One analyst, Ralph Schackart at William Blair, put it plainly: investors want proof that people will adopt and pay for a Meta AI product, and they have not seen it yet.
There is also a trust problem that money cannot fix quickly. When Meta gave Llama away, developers liked it precisely because it was free and open. Now that crowd has cooled on the company. One executive, Rob May, called Muse Spark a "yawn" because hardly anyone can get their hands on it, and said Meta stopped returning his messages. Not everyone is writing Meta off. Andrew Moore, a former Google Cloud AI chief, points out that Meta has worked on making its models cheaper to run, which could appeal to developers worried about rising costs.
If you use Facebook, Instagram, or WhatsApp, the thing to watch is a small one: the day an AI feature you have been using free suddenly sits behind a "subscribe" button. That is the shift this whole story is building toward, and it would be a real change in a relationship that has been free since you signed up.
It probably will not happen all at once, and it may not happen to you at all. Plenty of the new AI tricks will stay free because they make Meta's ads more valuable, which is still where almost all its money comes from. The paid tier, if it arrives, is likely to be the fancier stuff layered on top.
The wider point is worth holding onto. The "free" internet was never really free; advertisers covered the cost in exchange for your attention. Now that AI is so expensive to build and run, more of the companies you use every day are going to ask you to chip in directly. It is a fair time to start deciding which AI features are actually worth paying for, because the era of assuming they all come at no charge is ending.