Meta reports 10 million Muse users in the US and Canada within six weeks — figures from the Q3 call
When Meta presented its third-quarter figures on 21 October 2026, the results were not the main focus — the agent was. Muse, launched on 8 September, has according to the company reached 10 million users in the US and Canada in roughly six weeks, with shopping as the most common use. On the same call, leadership outlined a unified revenue model: commerce fees, subscriptions and enterprise offerings. At the same time, they warned that serving costs will rise before promised efficiency gains take effect — in a quarter where a legal charge of nearly $10 billion had already weighed on results.
The milestone on the call
The figures are company-provided and came via third-party transcripts of the earnings call, not independent verification. With that caveat, the picture is as follows: Muse is reported to have reached 10 million users in the US and Canada in about six weeks, with 25 percent daily use — a level leadership, according to BigGo Finance's summary of the call, described as more than double Threads' early retention (BigGo Finance).
Mark Zuckerberg was quoted on the call: "In just a month and a half, more than 10 million people in the US and Canada have created Muse personal agents," and further: "Growth continues to be strong, and we've been number one in the App Store almost every day" (BigGo Finance, The Motley Fool).
CNBC's Investing Club summarized the quarter under the headline "Meta put out an ambitious roadmap to Muse monetization after strong quarter and guide" — that is, Meta laid out an ambitious roadmap for Muse monetization after a strong quarter (CNBC).
Shopping as the primary use
The most concrete element in the new usage picture is what people are actually doing with the agent. According to leadership, shopping is the most common use case: over a quarter of weekly active users conduct commerce through Muse, with an average purchase size above $200 (company-provided, BigGo Finance).
This ties directly back to the launch event at Meta Connect in September, where a range of major players were presented as integration partners: Walmart, Best Buy, Gap, Sephora, Wayfair, Dick's Sporting Goods, Ulta Beauty and Fanatics for commerce, Expedia for travel planning, and Instacart for grocery orders. Meta also stated that the connector platform — which lets developers link their own services to the agent — has received over 1,500 developer applications (Benzinga via Yahoo Finance).
Three revenue streams
The monetization plan outlined on the call has three legs (BigGo Finance):
- A free consumer tier, funded by commerce fees paid by businesses. Zuckerberg has previously said Meta expects to make money by taking a small fee on the transactions the agent completes.
- Paid subscriptions for advanced models and extra tokens — tiers at $20 and $100 per month, launched alongside the free base version on 8 September.
- Enterprise offerings, including an API, Muse Code, and the new Meta Enterprise platform, led by CJ Desai.
The logic is that the compute-intensive free product should sustain itself through agentic commerce: the more users shop through the agent, the more fee revenue is generated. What the actual fee levels are, the sources did not specify.
The counterpoint: costs come first
The same leadership that promised revenue streams warned about the cost side. Muse is highly compute-intensive, and serving costs are expected to rise before efficiency gains of over 10x — a target leadership set within one year — materialize (BigGo Finance). The efficiency improvement is thus a target or expectation from leadership, not something that has been demonstrated.
The cost pressure is already visible in guidance figures. Meta raised its full-year expense guidance to $175–179 billion, and guided fourth quarter to $73–76 billion in revenue, assuming a 1 percent currency headwind (The Motley Fool).
The quarter around the agent
The financial context sets the frame for how much Meta can invest in Muse. The company reported $64.6 billion in revenue in the third quarter, up 26 percent from a year earlier measured in both reported currency and constant currency.
Results, however, were heavily weighed down by a legal charge of $9.89 billion tied to an August settlement with US state attorneys general. The charge pulled GAAP operating income down 37 percent, to $13 billion, and produced an operating margin of 20 percent — 35 percent measured excluding the charge. GAAP EPS came in at $3.83, versus $7.40 adjusted (a non-GAAP measure that excludes the legal charge) (The Motley Fool, BigGo Finance).
On other fronts, Meta announced a new non-voting share class, distributed through a 2:1 stock split (stock dividend), with trading expected to begin in late January 2027. The company also said it expects Reality Labs losses to narrow from $18–19 billion in 2026 to around $15 billion in 2027, as investments shift toward AI (BigGo Finance).
What we still don't know
There is good reason to hold the figures loosely:
- The entire Muse picture is company figures. User numbers, engagement, commerce share and average purchase size come from leadership's own statements on the earnings call, carried through third-party transcripts. No independent verification exists. The two BigGo documents are likely built on the same call and do not constitute two independent sources.
- Geographic scope. The sources explicitly say US and Canada. It is unclear whether the figures reflect only the launch markets, or what they would look like globally — but they do not cover the world as a whole.
- The enterprise tier is thinly described. The API, Muse Code and Meta Enterprise under CJ Desai are mentioned, but pricing and details are missing from the source material.
- The efficiency gain of over 10x is a promise of future improvements, not a result. Whether commerce fee revenue and subscriptions can cover rising serving costs in the meantime remains unresolved.
What stands firm is the sequence: adoption and usage patterns came first, the revenue model is built around them, and the costs are real now. Whether the sequence holds — whether commerce volume and fees eventually rise above the compute bill — none of the sources can confirm today.