Meta’s latest earnings discussion puts Muse’s commercial model, rather than its launch, at the center of the story. TechCrunch covered the October 21 update.
For an AI agent, adoption and profitability answer different questions. Signing up shows willingness to try a service. Paying for the computing behind repeated tasks requires another set of economics. Meta’s new disclosures provide some evidence about the first question, while leaving important parts of the second unresolved.
Early activity is not a margin calculation
In the published full earnings-call transcript, Mark Zuckerberg says more than ten million people in the United States and Canada set up Muse in roughly six weeks. He says about a quarter of those who set it up use it on a given day, while more than a quarter of users engage in shopping, with average purchases above $200. These are management’s figures, not an independent usage audit.
The measures should remain separate. A sign-up count is not a paying-customer count. A daily-use share does not, by itself, describe how an identified cohort behaves after several months. An average shopping purchase is also not the amount Meta receives from that purchase.
Those distinctions are important when interpreting commercial promise. A large transaction can coexist with a small platform fee; frequent use can coexist with expensive service delivery. The useful next disclosure would connect activity to revenue and delivery cost on a comparable basis, rather than asking readers to combine unrelated denominators into an apparent return.
A large company can finance an unfinished model
Meta’s original quarterly results report $64.60 billion in revenue and $32.79 billion in operating cash flow. Capital expenditures, including finance-lease principal payments, were $34.91 billion, leaving negative $2.12 billion in the company’s non-GAAP free-cash-flow measure. Its full-year capital-expenditure outlook is $135 billion to $145 billion. These are company-wide figures, not Muse’s accounts.
That scope prevents two tempting shortcuts. It would be wrong to assign the entire investment bill to this one agent, and equally wrong to assume the agent is profitable because the parent generates substantial operating cash. The release does not provide a standalone Muse income statement.
The distinction also changes how scale should be discussed. Financing a rollout is a question about available resources and investment priorities. Making each additional customer economically attractive is a question about the relationship between incremental receipts and incremental expense. An organization can have the means to pursue the former before it has demonstrated the latter. A useful assessment would also separate temporary experimentation from repeat purchases: a burst of shopping immediately after sign-up need not establish a stable commercial habit.
The cost target is still a forecast
On the call, Zuckerberg describes planned transaction fees and subscriptions, and forecasts more than a tenfold improvement in Muse’s compute cost efficiency over the following year. Asked about current per-user costs, CFO Susan Li declines to give specific figures. The forecast therefore cannot be checked against a disclosed starting cost in that discussion. Source: original call transcript.
An efficiency target becomes much more informative when its denominator is clear. Cost per token, per task and per satisfied customer need not move together. A cheaper individual operation could be offset by an agent performing more operations, while a more capable but costlier operation could still be valuable if it completes a useful task. Neither scenario establishes what is happening inside Muse today.
The next meaningful financial milestone would be a disclosure that allows those alternatives to be distinguished: realized revenue, a defined serving-cost measure, and a consistent period of activity. Without that connection, the commercial story remains an investment proposition supported by early engagement claims, not a demonstrated standalone profit engine.