Muse Tops 10M Users; Meta Targets 10x Compute Efficiency

Meta said on its third quarter earnings call on 21 October 2026 that its AI agent app Muse has been adopted by more than 10 million users, TechCrunch reported. Mark Zuckerberg said about 25% of people who set up Muse now use it daily, and that over a quarter of engaged users shopping via the agent had an average purchase size of over $200, according to the report. He also said Meta expects to improve Muse’s compute cost efficiency by more than 10x in the next year, TechCrunch reported.

In plain terms, Muse is an AI agent: software that does tasks for a person, such as buying something or handling email, rather than only answering questions. Zuckerberg said Meta expects Muse to eventually bring in more revenue per user than its existing apps, TechCrunch reported.

Meta Q3 2026 explained · video

The quarter in about a minute and a half, from our published pieces (21 Oct 2026).

Business Pill · AGENTIC COMMERCE EXPLAINED

A short Business Engineer explainer on agentic commerce: what changes when the shopper is an AI agent. It teaches the general idea only and says nothing about the companies in this piece.

The key insight: As we read it, the number that changes the Muse story is the cost line: Zuckerberg was quoted by TechCrunch as saying Meta expects to improve Muse’s cost efficiency “by more than 10x in the next year,” which he tied to Muse becoming profitable at scale.

Update (23 Oct): Muse launches in Japan. Alexandr Wang, Meta’s chief AI officer, posted on X on 23 October: “in honor of muse japan 🇯🇵 launch, muse is doing a speedrun of japan’s most important cultural icons.” TestingCatalog, which tracks product rollouts, reported that the Muse agent “is rolling out in Japan,” listing the US, Canada and Japan. Muse was US-only when it launched in September, as we reported then. As we read it, Japan is the first market outside North America to appear in TestingCatalog’s rollout list.

What Meta Said on the Call

According to TechCrunch, Zuckerberg said about 25% of people who set up Muse now use it daily, which he said is helping retention. TechCrunch wrote that “over a quarter of Muse’s engaged users who were shopping via the AI agent, had an average purchase size of over $200.”

Muse uses the connector of Plaid, the company that links payment apps to customers’ bank accounts, for its agentic shopping, TechCrunch reported. Zuckerberg was quoted as saying: “Plaid told us that they hadn’t seen activity like they’re seeing with their connector on any other service before.”

Other popular uses he named were personal productivity tasks, such as connecting the agent to email or calendars, the report says. He pointed to a security feature called “Muse Secure VM” and said a forthcoming “Muse Confidential VM” will prevent even Meta from seeing users’ content, according to TechCrunch.

About 25% of people who set up Muse use it daily, and Meta expects to improve Muse’s compute cost effici
About 25% of people who set up Muse use it daily, and Meta expects to improve Muse’s compute cost efficiency by more than 10x in the next year, Mark Zuckerberg said on Meta’s third quarter 2026 earnings call on 21 October 2026, as reported by TechCrunch. Icons scaled at one per 5% of people who set up Muse (our scaling).

What Was Posted on X

Shay Boloor posted on X on 21 October that Mark Zuckerberg says Muse could eventually generate more revenue per user than Facebook and Instagram, with shopping already its most popular use and purchases averaging about $200, and that this could give Meta a new commerce business alongside its ad platform. The commerce read is his own. This piece takes the call remarks from TechCrunch’s report, which puts the $200 figure as the average purchase size of over a quarter of engaged users shopping via the agent.

Alexandr Wang, Meta’s chief AI officer, marking what he calls the Muse Japan launch by saying Muse is doing a “speedrun” of Japan’s most important cultural icons.

Bloomberg posted a Bloomberg Opinion column in which Dave Lee says Muse’s computing costs will need to fall at least tenfold before the AI assistant is scalable, calling that a significant challenge if Zuckerberg is to reach his stated goal of more than a billion users. This is an opinion column, not reporting. Why it matters: it puts a figure on the cost question behind the compute-efficiency target in this piece.

The Revenue Plan

TechCrunch reported that Zuckerberg said Muse has reached “product-market fit” and that Meta’s focus now shifts to scaling the agent, increasing revenue and optimizing its compute costs. Meta forecasts that Muse will eventually bring in more revenue per user than its “family of apps” users do today, a figure Zuckerberg put at “several hundreds of dollars annually,” the report says.

Meta has said it intends to keep Muse free for most users while keeping a percentage of the commerce on the platform, with power users paying for the most advanced models or more tokens, TechCrunch wrote. Meta’s own results release names the product in Zuckerberg’s statement: “We hit major milestones this quarter toward delivering personal superintelligence for everyone – including launching Muse and Muse Wonder, our first frontier model.”

The Cost Side

Zuckerberg was quoted by TechCrunch as saying: “On the compute side, we are building out the capacity to serve more than a billion people over the next few years.” He added: “In the near term, Muse is very compute-intensive to run, but we expect to improve its cost efficiency by more than 10x in the next year, which will enable it to be profitable once the revenue lines are at scale.”

The same quarter’s release reported capital expenditures, including principal payments on finance leases, of $34.91 billion against $32.79 billion of cash flow from operating activities, leaving free cash flow of negative $2.12 billion (our results piece).

Muse in Numbers, From Three Sources

Three sets of Muse figures have been published in two days. Each measures something different, so they are listed side by side, not combined.

SourceWhat it measuresFigure
Sensor Tower, via TechCrunch (20 Oct)Downloads since the September debut10.9 million, 10.2 million in the US
Sensor Tower, via TechCrunch (20 Oct)Share of users still active after 7 days30%, against a 19% rival average
Chart credited to Morgan Stanley, posted by @intelfabs (20 Oct)Installs active on Day 1 and Day 1473% and 23%
Meta call, via TechCrunch (21 Oct)UsersMore than 10 million
Meta call, via TechCrunch (21 Oct)People who set up Muse and use it dailyAbout 25%
Meta call, via TechCrunch (21 Oct)Engaged users shopping via the agent with an average purchase size of over $200Over a quarter

Connecting the Coverage

On 20 October we reported Sensor Tower’s count of 10.2 million US downloads in Muse’s first 40 days, against 3.7 million for ChatGPT over the same span (our Sensor Tower piece). Earlier on 21 October we covered the retention chart credited to Morgan Stanley, which showed 73% of installs active on Day 1 and 23% by Day 14 (our retention piece).

Meta’s results release the same evening reported negative free cash flow of $2.12 billion (our results piece), and our quarter analysis puts every figure in one table (our analysis piece). Ad prices rose 14% while ad volume grew 10% (our ad pricing piece).

Agents that buy things were also the subject of two other pieces on 21 October. Anthropic’s Claude Duet can pay through Stripe’s Link after the user approves the spend (our Claude Duet piece), and Walmart opened its stores to AI shopping agents with spending limits, while TechCrunch reported that Amazon blocks them (our Walmart piece).

The Structural Read

As we read it, the call moved the Muse discussion from installs to use: TechCrunch reports Zuckerberg said about 25% of people who set up Muse now use it daily, after earlier outside figures that tracked downloads and retention.

As we read it, Meta is describing a second business model alongside ads: TechCrunch reports Meta intends to keep Muse free for most users while keeping a percentage of the commerce on the platform.

As we read it, the cost and the capacity are the same question: Zuckerberg was quoted as saying Meta is building out the capacity to serve more than a billion people, in a quarter when its capex was larger than its operating cash.

Mark Zuckerberg, Meta Q3 2026 earnings call, as quoted by TechCrunch

“In the near term, Muse is very compute-intensive to run, but we expect to improve its cost efficiency by more than 10x in the next year, which will enable it to be profitable once the revenue lines are at scale.”

Three Implications

FOR ADVERTISERS TechCrunch reports Meta intends to take a percentage of the commerce on Muse, a revenue line separate from the ad pricing in its release.

FOR PAYMENT AND DATA COMPANIES TechCrunch reports Muse uses Plaid’s connector for agentic shopping, and Zuckerberg was quoted as saying Plaid had not seen activity like it on any other service.

FOR RIVAL AGENT MAKERS Anthropic’s Claude Duet and Walmart’s Agent Gateway, both covered in our 21 October pieces, also set rules for agents that buy.

The Business Engineer Lens

This story maps onto the Business Engineer framework AI Agents & The New Payment Infrastructure.

The framework puts it this way: “Every payment system ever built rests on the same assumption: a human is present, hesitating, and approving.”

As we read it, Muse’s shopping runs through Plaid’s bank-account connector, per TechCrunch, so the question the framework raises applies to Meta too: who approves when the agent buys. The framework was published in February 2026 and does not discuss Muse.

What Is Not Established

The call remarks here come from TechCrunch’s report: Meta had not posted the call transcript when this was published. TechCrunch does not say how Meta defines an engaged user or a Muse user, or what share of all users shop through the agent.

Meta’s release gives no separate revenue figure for Muse. The three sources in the table count different things over different periods and are not comparable one-to-one. We did not contact Meta.

Business Engineer Framework

AI Agents & The New Payment Infrastructure

A Business Engineer read on why payments built for a human who hesitates and approves have to be rebuilt for software that buys.

Read the Map of AI →

The Bottom Line

As we read it, Meta told analysts on 21 October 2026 that Muse has more than 10 million users, about 25% of people who set it up use it daily, and over a quarter of engaged users shopping through it had an average purchase size of over $200, according to TechCrunch. Zuckerberg said Meta expects to improve Muse’s compute cost efficiency by more than 10x in the next year, TechCrunch reported.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. The earnings-call remarks come from TechCrunch’s report of 21 October 2026, because Meta had not posted the call transcript at publication; the CEO statement is from Meta’s third quarter 2026 results release. Sensor Tower’s figures are as reported by TechCrunch on 20 October, and the retention chart is the one posted by @intelfabs and credited to Morgan Stanley, as covered in our earlier pieces. We did not contact Meta. Nothing here is financial or investment advice.

Sources: TechCrunch: Meta says Muse has over 10M users, with shopping a top activity (21 Oct 2026) · Meta: Third Quarter 2026 Results (Exhibit 99.1, 21 Oct 2026) · TechCrunch: Meta’s Muse hits 10M US downloads in 40 days (20 Oct 2026)

Scroll to Top

Discover more from FourWeekMBA

Subscribe now to keep reading and get access to the full archive.

Continue reading

FourWeekMBA