MENLO PARK, UNITED STATES — Meta's Muse AI assistant has reached 2.8 million downloads in 12 days, fueling a sharp rally in its shares as investors assess the potential for new subscription revenue.
Meta Platforms has attracted millions of downloads for its new artificial intelligence assistant, Muse, prompting renewed investor interest in whether AI agents can develop into a meaningful source of consumer revenue.
The assistant, introduced on September 8, recorded 2.8 million downloads during its first 12 days, according to figures from analytics company Apptopia cited by Reuters on September 22.
Meta's shares climbed more than 11% on Monday, September 21, as investors responded to the product's early reception. The stock has gained more than 20% since Muse launched, adding over $200 billion to the company's market capitalization over that period.
The rally reflects growing expectations that consumer AI could create commercial opportunities beyond Meta's established advertising operations. However, the number of people downloading an application does not establish how frequently they use it or whether they will eventually become paying customers.
What Meta's Muse AI assistant can do
Muse belongs to an emerging category of artificial intelligence products known as AI agents. Unlike conventional chatbots, which generally respond to individual prompts, agents are designed to carry out actions across applications and websites.
Meta says Muse can send emails, arrange travel, make purchases and complete other multistep tasks on a user's behalf. It can also assist with longer-term projects, using information from connected services to organize activities and suggest next steps.
Users interact with Muse through a dedicated application or WhatsApp. The company launched the product in the United States on iOS, Android and the web, with integration into its AI glasses planned for a later stage.
The assistant operates using Muse Spark, a model developed by Meta Superintelligence Labs. Meta introduced the first Muse Spark model in April 2026, before releasing the consumer-facing agent in September.
The distinction between the model and the assistant is important. Muse Spark provides underlying AI capabilities, while the Muse agent combines those capabilities with software that can access services and execute tasks.
Meta's product announcement describes a system intended to work across everyday digital activities, rather than requiring users to open separate applications and complete every step manually.
For example, arranging a trip may involve searching for transport, comparing options, entering information and completing a reservation. An AI agent is designed to coordinate several such steps, subject to the permissions and confirmations required.
That approach makes the technology potentially useful for routine administration, but it also introduces questions about reliability, privacy and the consequences of an incorrect action.
Why 2.8 million downloads matter to investors
Muse's reported 2.8 million downloads in 12 days provide an early indication of consumer interest. They do not, by themselves, demonstrate a sustainable business.
Downloads measure installations rather than active users. A person who installs the application and stops using it is counted in the initial figure, but does not necessarily contribute to recurring revenue.
Investors therefore have several different measurements to consider: the number of installations, the proportion of users who remain active, the frequency of completed tasks and the percentage who purchase subscriptions.
The initial market reaction has nevertheless been substantial.
Reuters reported that Meta's shares had risen more than 20% since the launch, increasing the company's market value by over $200 billion. Monday's gain exceeded 11%, while enthusiasm for AI-related businesses also supported other technology shares.
The Financial Times reported that Meta's rally contributed to a 2.8% advance in the Nasdaq 100. Semiconductor companies also benefited from expectations that increasingly capable AI software could support demand for computing infrastructure.
These movements indicate that investors are considering both the direct commercial potential of AI agents and their possible effects on the wider technology industry.
However, market capitalization is not revenue. A $200 billion increase in Meta's market value reflects changes in the price investors are willing to pay for its shares, not money received by the company.
The distinction matters when evaluating a product whose long-term financial performance remains unproven.
How Muse's $20 and $100 subscriptions could generate revenue
Meta offers Muse through a free tier and two paid subscription options priced at $20 and $100 per month.
The model allows consumers to experiment with the assistant before deciding whether its additional capabilities justify a recurring payment.
For Meta, subscription revenue would represent a different commercial relationship from its established advertising business.
Advertising revenue depends heavily on the company's ability to attract audiences and deliver advertisements. Subscription revenue, by contrast, depends on persuading individual customers to pay regularly for access to a service.
The commercial opportunity can be illustrated through a simple calculation.
If Muse reached 10 million active users and 3% subscribed to a $20 monthly plan, it would have 300,000 paying subscribers. At that price, the resulting annualized gross subscription revenue would be $72 million.
That figure is an illustrative calculation, not a forecast. It assumes every paying customer uses the same plan and remains subscribed throughout the year. It also excludes payment fees, operating costs, discounts and other factors.
A larger user base or a greater proportion of customers choosing the $100 plan would change the calculation considerably.
Jefferies has outlined a more ambitious scenario in which Muse reaches one billion users by 2027, with a 3% paid conversion rate. Reuters reported that the firm's scenario envisages approximately $10.8 billion in annual revenue.
Such projections depend on assumptions that have not yet been demonstrated through sustained consumer behavior.
Meta must also consider the cost of operating AI agents. Completing complex tasks can require substantial computing resources, particularly when an assistant performs repeated searches, interacts with external services or runs several processes.
The relevant commercial question is therefore not simply how many subscriptions Muse can sell, but how much revenue remains after the costs of delivering the service.
Privacy and security become central to AI agents
Muse's ability to perform tasks requires access to information and services that users may regard as sensitive.
An assistant that can send emails, manage appointments or make purchases needs permissions extending beyond those required by a basic question-and-answer chatbot.
Meta says it has designed Muse around a dedicated cloud-based environment called Muse Secure VM.
According to the company's technical explanation, each agent operates within its own virtual machine, with a browser and storage for information associated with connected services.
Meta also describes a separate safety component, called Sentinel, which reviews actions before they reach the internet and requests user permission when necessary.
The company says Muse cannot directly view stored passwords or payment credentials and that users can control which services the agent accesses.
It also says the assistant requests confirmation before sensitive actions, including sending emails and making purchases, and provides a record of its activities.
These are company-described safeguards rather than a guarantee that every action will be correct or secure.
Reuters reported at the September launch that internal testing had identified reliability and security concerns, including problems involving access to personal information. Meta delayed the product's original launch while addressing safety issues.
The distinction between a system's intended safeguards and its performance in real-world conditions remains significant.
A mistaken chatbot response may provide inaccurate information. An agent acting on that information could potentially send an incorrect message, purchase an unwanted item or disclose information through an external service.
For consumers, reviewing permissions and understanding which actions require confirmation will be important considerations when deciding how extensively to use the assistant.
For Meta, maintaining trust could affect whether initial downloads translate into continued engagement.
Competition and access to external services
Muse enters a market in which technology companies are developing assistants capable of acting across multiple digital services.
The competition is not limited to generating accurate responses. It also concerns the ability to complete tasks reliably, integrate with applications and provide a convenient interface.
Meta has several potential distribution channels through its existing products, including WhatsApp, which already provides a familiar messaging environment.
However, access to external services is not entirely under Meta's control.
Business Insider reported that Amazon had blocked Muse from placing orders through its platform, illustrating the complications that can arise when an AI agent attempts to interact with another company's commercial systems.
The development highlights an important limitation of agent-based services: a product may be technically capable of completing a transaction but still face restrictions imposed by the service involved.
At the same time, partnerships could expand the range of supported activities.
The Wall Street Journal reported on September 22 that Shopify and Meta had announced plans to integrate Muse with Shopify's commerce infrastructure and Shop Pay checkout.
Such arrangements could help establish more structured methods for AI agents to interact with merchants.
For consumers, the practical value of Muse will depend partly on which services it can use and whether transactions can be completed with appropriate authorization.
For businesses, the emergence of AI agents raises questions about how customers discover products, compare offers and complete purchases.
What Meta needs to demonstrate after its launch
Muse's early performance has given investors a reason to reconsider the commercial potential of consumer AI.
The application has attracted substantial initial interest, and Meta has introduced a pricing structure that could generate recurring revenue if enough users decide to subscribe.
The company has also outlined a technical approach intended to support complex tasks while giving users control over connected services.
Nevertheless, several important questions remain unanswered.
Meta has not established through the reported download figures how many people will continue using Muse after the initial launch period, how frequently they will complete tasks or how many will choose a paid plan.
Its long-term economics will also depend on computing costs, reliability, competition and the ability to maintain access to external services.
The share-price rally demonstrates that investors are assigning greater value to Meta's AI opportunities. It does not establish the eventual profitability of Muse.
The next meaningful evidence will come from sustained user activity, subscription adoption and any financial disclosures that allow the assistant's commercial contribution to be assessed.
For now, Muse represents a significant consumer launch in the development of AI agents, with early demand established but its longer-term financial results still to be determined.