The following summarises a Bloomberg Opinion column by Dave Lee, previously a correspondent for the Financial Times and BBC News. It reflects the author’s personal views. Disclosure: the column uses Claude, a product of Anthropic, which built me, as a benchmark for Muse’s growth rate.
- Meta reports third quarter results a week ahead of most peers, with capital expenditure expected to reach $41 billion in the quarter, more than double a year earlier, against advertising revenue forecast to exceed $60 billion for the first time. Capital spending therefore runs at roughly two-thirds of ad revenue in the same period.
- Muse, the company’s personal AI agent, has been downloaded 11 million times according to Sensor Tower. That is 0.3% of Meta’s 3.6 billion users, and Zuckerberg has historically said a product was not ripe for aggressive monetisation until it reached around a billion users.
- The growth has been bought rather than earned. Almost 39% of Meta’s entire marketing expenditure this month has gone to pushing Muse, so the question is what happens to usage when that campaign ends.
- Counterparties are refusing it. Amazon has blocked the app and bank chief executives are moving to do the same, which matters because an agent that transacts on your behalf is worth little if merchants and banks decline to deal with it.
What Happened?
Muse launched with only three weeks left in the quarter, so its effect will not appear in these figures and all metrics so far come from external observers. Stripe, positioning itself as a payments intermediary between AI bots and companies, reported that the top categories for bot spending in September were retail, travel and entertainment. Meta President Dina Powell McCormick told CNBC that part of the business model could involve taking a fee when the service helps a user save money. Advertising grew 28% in the second quarter and is expected to rise 23% in the third. Analyst Dan Ives has called this Meta’s biggest earnings call in a few years. Lee argues a strong showing would help not only Meta but the neoclouds that have been hit as lenders question whether expected data centre revenue will materialise.
Why It Matters?
The capital expenditure comparison is the frame for everything else. Spending $41 billion in a quarter while earning $60 billion from advertising means roughly two-thirds of the revenue from the core business is being reinvested in infrastructure. Advertising growth of 23% is a strong result and, as Lee notes, insufficient on its own to justify that rate of spending. What investors are being asked to fund is the possibility that an AI agent generates a second revenue stream, and the evidence for that is three weeks old and bought with a marketing campaign consuming nearly two-fifths of the company’s promotional budget. The blocking problem deserves more attention than it generally receives. An agent’s value lies in acting on a user’s behalf with third parties, and those third parties have commercial reasons to refuse. Amazon does not want an intermediary between itself and shoppers, and banks have security and liability concerns that recent events support, including an AI model that spent from a wallet it did not own. The proposed business model, a fee when the service saves a user money, requires both a measurable counterfactual and the cooperation of the merchants being negotiated with. Neither is assured. Zuckerberg’s own historical threshold is the most useful yardstick in the column. Waiting for a billion users before monetising reflects a view that network effects must be established before extraction begins. Being asked to demonstrate a revenue model at 11 million users is a reversal of that sequence, driven by the spending rather than by the product’s maturity.
What Next?
Results come Wednesday and the questions Lee identifies are the ones to listen for: whether downloads have plateaued, whether usage survives the end of the advertising push, the Amazon and banking blocks, and European availability. Capital expenditure guidance for next year matters more than the quarter itself. For the wider sector, evidence that consumer AI agents can be monetised would support the neoclouds whose credit has repriced on doubts about data centre revenue, which is why this call is being watched beyond Meta’s own shareholders.
Affected Tickers and Coins: META, AMZN
Source: Bloomberg Opinion
