Meta’s Muse launch added nearly $200 billion in market value

A product investors can price

Before Muse, Meta’s AI investments mostly improved advertising, recommendations and engagement inside existing products. Those benefits mattered, but they did not look like a distinct source of revenue. Muse changed that framing: it can send emails, plan trips, buy products online, fill out forms and call companies on a user’s behalf.

Meta has about 3.6 billion daily users across Facebook, Instagram, WhatsApp and Messenger. The company is reportedly considering a monthly Muse subscription priced from $20 to $100. Even a modest share of paying users could mean billions in recurring revenue.

That audience gives Meta a distribution advantage most AI companies would need years and billions of dollars to build. It could bring Muse into WhatsApp or promote it through Instagram and Facebook without paying separately to acquire every user. But reach says little about the cost of fulfilling those users’ requests.

The launch also landed while Meta’s core business was strong. Second-quarter revenue rose 28% to $60.8 billion, while operating margin fell to 31%. Meta spent $31.1 billion on capital expenditures that quarter and expects $130–145 billion in 2026, with most of the spending going to servers, data centers, networks and AI development.

Muse was not the only reason the stock rose: the technology sector was recovering, and Meta’s existing business remained strong. But the app gave investors a new way to imagine revenue alongside those expenses. Analysts raised price targets, and the shares jumped before Muse had been out for two weeks.

The work behind the automation

Reuters reported that Meta is testing a “human assistant” feature in which contractors handle some calls made by Muse. The test was offered to half of Meta’s employees, who can opt out. Meta said it is studying safety and privacy issues ahead of a public launch.

Using people to train or improve an unfinished product is not inherently troubling. The optimistic case is that human involvement falls as the system learns. That may happen, but investors do not yet have enough evidence to treat it as a given.

Some tasks should be easier to automate: writing an email, finding a restaurant or comparing simple options. The economics get harder when a user wants to dispute a hotel charge, rebook a missed trip or cancel a service a provider would rather keep. These cases require negotiation and knowledge the system may not have. And companies often have little incentive to help the customer.

There is also the cost of mistakes. A single task may require several model calls, access to outside services and a person to handle an unusual case. The $100 subscribers may use Muse most often, making the most valuable customers potentially the most expensive to serve.

Merchant access is another constraint. Shopify welcomed Muse, gave it access to Shop Pay and sees it as an additional channel for its merchants. Amazon, by contrast, has barred Muse from making purchases on its site, citing concerns about unlawful activity, privacy and user experience.

Meta cannot make commerce sites accept its agent. Some will want the extra sales; others will protect their customer relationships and data. Agreements, technical integrations and potential revenue sharing will all shape the service’s economics. Download charts cannot tell investors how those arrangements will work.

The metric that matters

I think the most revealing number will not be subscriptions or downloads, but the cost of a successfully completed task. It would bring together computing costs, the share of tasks completed without human help and the cost of fixing errors.

If that cost falls, it would suggest Muse is learning and that subscription revenue can grow with the service. If it stays high, Meta may have built a useful service whose economics look more like a personal-assistant business than a software platform.

Human involvement could help Meta get there. Contractors can identify requests that stump the system, moments when users feel uncomfortable and safeguards needed before an agent takes action. If that work reduces the need for people over time, today’s expense could become an investment in automation.

Muse could also matter beyond subscription revenue. If consumers use it to choose products, book trips or contact companies, Meta gains another way to understand their commercial intent. That could strengthen its advertising business and give it influence over transactions that now begin on Google, Amazon or a merchant’s own app. The strategic value may eventually exceed the subscription income.

My guess is that this broader possibility helps explain Wall Street’s quick response. Meta has shown before that a huge audience can help products that were not first to market. It does not need to invent every important technology if it can put a useful version in front of more people than competitors can reach.

But Meta gained nearly $200 billion in market value before investors had seen Muse’s economics. That increase depends on more than getting people to try the app: automation must rise, human help must fall, and each completed task must become profitable. Until then, the value of Muse is easier to imagine than to measure.