Meta’s AI bet is colliding with a $25 billion privacy penalty

Meta’s earnings are no longer simply a referendum on advertising demand. They are becoming a stress test for how much financial pain investors will tolerate while the company builds an AI business—and while privacy litigation threatens to make that business more expensive to operate.

The contradiction was clear on October 21, 2026. Meta reported third-quarter revenue of $64.6 billion, above the $63.3 billion analysts expected, and forecast fourth-quarter revenue of $73 billion to $76 billion. Yet free cash flow fell to negative $2.1 billion, while a New Mexico judge ordered the company to pay $25 billion over claims that it misled consumers about data privacy. Fortune reports that Meta plans to appeal the penalty.

That is the investment argument in miniature: the core machine is still producing growth, but the costs attached to its next act are arriving faster than the returns.

Meta is buying an AI future with yesterday’s advertising profits

The bullish case is not hard to understand. Meta’s family of apps remains enormous, with 3.64 billion daily active people in the third quarter. Its advertising business is funding infrastructure that could eventually support agents, foundation models and new commerce products. The company says more than 10 million people in the U.S. and Canada had set up Muse personal agents during the month and a half after launch, and roughly 25% of those users were using the service on any given day.

Those are encouraging early signals, but they are not yet an economic model. A user opening an AI assistant is not the same as a user generating profitable revenue. Meta says shopping is the biggest AI-agent use case and that more than a quarter of Muse users are engaging with it around purchases, with an average purchase size of over $200. That suggests a plausible route to monetization: recommendations, transactions and eventually services for businesses.

We think this is where Meta has a stronger story than many AI ventures. It already has distribution, advertisers and a large consumer network. But distribution is only the starting advantage. The company still has to show that the revenue created by AI exceeds the cost of serving it, especially when the service is compute-intensive enough that Meta has not launched Muse globally.

The spending scale makes that hurdle impossible to ignore. Meta raised the lower end of its 2026 data-center capital-expenditure range, narrowing it to $135 billion-$145 billion from $130 billion-$145 billion. Third-quarter capital expenditures reached $34.9 billion, compared with $19.4 billion a year earlier. Fortune reports that free cash flow turned negative as this infrastructure push accelerated.

The privacy penalty is more dangerous than a single bad quarter

A $25 billion judgment is obviously material. But the larger concern is what it says about the durability of Meta’s economics. Privacy enforcement is not merely a fine that can be placed in a spreadsheet and forgotten. It can affect how data is collected, how products are designed and how confidently the company can use its most valuable asset: behavioral information across its platforms.

We should be careful about treating the New Mexico order as a final cash outflow. Meta plans to appeal, and the documentation does not establish the ultimate amount or timing of any payment. Still, the legal event arrives alongside other costs. Meta’s third-quarter operating income was $13 billion, down from $20.5 billion in the second quarter of 2025, with a $10 billion settlement over youth safety contributing to the decline. The terms described in the reporting could require about $18 billion in payments to states over the next decade, depending on cooperation from TikTok and Google’s YouTube.

The point is not that every legal charge will recur at the same scale. It is that Meta’s financial model now has two expensive transformation projects: AI infrastructure and the continuing repair of trust and compliance. Investors may be underestimating the possibility that both demand capital at the same time.

Investors are rewarding the promise before demanding proof

The market’s response was revealing. Meta shares rose in extended trading after the results, with CNBC reporting a gain of about 2.5%, while early signs of Muse traction helped support investor enthusiasm. That reaction says investors still view Meta’s advertising engine as strong enough to finance the experiment.

Perhaps they are right. Meta’s fourth-quarter revenue midpoint of $74.5 billion was above the $73.73 billion analyst expectation cited by CNBC. But the stock response also reflects a willingness to look through present cash-flow pressure toward a hoped-for AI payoff.

That tolerance has limits. Meta’s daily active people figure came in slightly below analysts’ expectation, and Reality Labs produced $443 million in third-quarter sales while recording a $4.13 billion loss. The company says it will reallocate investment from that money-losing unit toward AI, which is sensible in principle. Yet moving money between speculative projects does not make the overall bet less speculative.

The planned Class C non-voting stock, effective in January 2027, is another signal. Meta says the new class will help finance growth, employee stock compensation and acquisitions while preserving its founder-led structure. In our view, it also highlights the governance question investors should not ignore: the company wants more financial flexibility for an expensive strategy without changing the founder’s central control.

The real verdict will come from cash flow, not downloads

Meta has earned the right to invest aggressively. Its core business is growing, its revenue outlook is strong and Muse appears to have gained an initial audience. But none of those facts proves that AI will become a durable source of cash rather than a permanent demand on it.

The company’s own executives acknowledge the issue. Zuckerberg said Meta would likely pursue monetization sooner for Muse than its normal product playbook because of the cost of running AI agents and their fit with retail transactions. That is a reasonable response, but it also means the company cannot wait indefinitely for scale before explaining the economics.

Our view is that Meta’s AI opportunity is real, but the stock’s most optimistic narrative is ahead of the evidence. The privacy penalty makes that gap more consequential: every dollar directed toward data centers is a dollar investors will increasingly compare with legal liabilities, compliance costs and negative free cash flow.

Meta does not need to abandon its AI buildout. It does need to prove that Muse and its foundation models can become businesses, not just products. Until cash flow turns convincingly positive and the legal exposure becomes more predictable, Meta is best understood as a profitable advertising company financing an unusually costly—and increasingly regulated—future.