A consumer AI agent has finally produced a download curve that beats ChatGPT's, and Wall Street repriced Meta by roughly $192 billion in a day on the strength of it. Whether that enthusiasm survives depends on two unresolved tests: getting merchants to let the agent transact, and finding enough subscribers to cover the free compute it hands out.
Meta Platforms' Muse agent held the top spot on Apple's U.S. free app chart for three straight days through Sept. 18, and the download record lifted Meta shares 11.4% on Sept. 21, adding roughly $192.3 billion in market value in a single session. The question now is whether the app's 1.43 million first-12-day iOS downloads convert into paying users before the novelty fades.
"Up to this point most of the agentic use cases have not really been for normal people," Stacy Rasgon, an analyst at Bernstein, said on CNBC's "Squawk on the Street." "Now you've got Meta's Muse and other agents out there that are starting to maybe get more potential for more broad-based adoption."
Sensor Tower data cited by analysts shows Muse drew 1.43 million iOS downloads in its first 12 days, edging OpenAI's ChatGPT at 1.37 million over a comparable launch window, while Apptopia puts total iOS and Android installs at 2.8 million. The app carries a 4.9-star rating across more than 25,000 App Store reviews. Muse runs on Meta's Muse Spark 1.3 model family and executes multi-step tasks — drafting email, booking travel, managing calendars, and completing transactions — rather than only answering prompts.
The re-rating spread beyond Meta. AMD rose 9.95% to close at $615.52 on Sept. 21, crossing a $1 trillion market capitalization for the first time, as investors bet that browser automation, code execution, and task scheduling will drive CPU and DRAM load. JonesTrading chief market strategist Michael O'Rourke said the market read Muse as a boost for an already bottlenecked CPU market. Meta's own capital spending is set at $130 billion to $145 billion for 2026, up from $72.22 billion in 2025.
Amazon blocked Muse from browsing and purchasing on its behalf, saying Meta did not notify it in advance and that the agent did not identify itself as an AI agent while on the site. Amazon called the unauthorized access a violation of its terms of use and asked Meta to remove the site from Muse's experience. The restriction protects a business where advertising services revenue rose 26% year over year to $19.809 billion in the second quarter of 2026, and it shields Amazon's own Alexa for Shopping assistant.
Shopify took the opposite position, agreeing to support Muse checkout through Shop Pay, though no launch timeline has been announced. Muse currently settles payments through Stripe Link, which issues a one-time virtual card number. Meta has said Muse cannot view user passwords or payment details, with credentials held in secure storage and called by the system rather than displayed to the model.
The split matters because agent commerce only scales if merchants opt in. Amazon's block shows that a user's authorization does not substitute for a platform's permission, and that the gatekeeper can decide whether an agent reaches its catalog at all.
Truist projects Muse could add $28.5 billion in incremental revenue by 2030. Oppenheimer's Jason Helfstein kept a Neutral rating, estimating Meta would need roughly 115 million paying users at $20 per month and an 80% incremental operating margin to move earnings meaningfully. Muse offers a free tier alongside $20 and $100 monthly subscriptions, and Bank of America does not expect subscriptions, advertising, or commerce commissions to become financially material before 2028.
Evercore ISI's Mark Mahaney framed the upside differently, calculating that 1% of Meta's 3.6 billion users at $20 per month would generate about $8 billion in annual revenue. Jefferies raised its Meta price target to $875 from $710, modeling one billion users by the end of 2027 with 3% converting to paid plans for roughly $10.8 billion in annualized revenue. JPMorgan's Doug Anmuth reiterated an Overweight rating and $820 target, calling Muse potentially the most widely used consumer AI application since ChatGPT.
The free tier cuts both ways. Meta absorbs inference and cloud compute costs for every non-paying user, so the more generous the free allowance, the harder Helfstein's margin assumption becomes to reach. Goldman Sachs research head Eric Sheridan has flagged that a consumer agent may ultimately lean on advertising and subscriptions, and that if the assistant takes promotion fees, users need to know whether recommendations were influenced.
Internal testing surfaced problems before launch. Muse exposed private data in some tests and stalled on a ticket-monitoring task after about 15 minutes, and Meta confirmed it delayed the original April release to address security. Rasgon put the tension plainly: "They care about security, I don't know how much they care about privacy."
Meta's architecture assigns each user a Muse Secure VM, a separate cloud virtual machine that stores personal data, with execution programs confined to restricted containers and account credentials managed apart from task state. A permission-review agent called Sentinel vets outbound access and connector actions. The design limits what a misled browsing process can reach, but Meta has not published independent test results validating it.
Retention is the open variable. OpenAI's Sora and Meta's own Vibes AI video tool both spiked on release and failed to hold their download curves, a pattern that makes the 4.9-star rating a weak predictor of month-three engagement.
Meta closed down 0.63% on Sept. 22 and rose 0.6% after hours, the first sign that the launch trade is cooling. Meta Connect runs Sept. 23-24 in Menlo Park, where CEO Mark Zuckerberg delivers the keynote and analysts expect details on the next-generation model code-named Watermelon. Jefferies analyst Jeffrey Favuzza warned the event could trigger a sell-the-news reaction after the 11% single-day gain. Meta trades near 21 times its 2027 earnings estimate, up from about 16 times several months ago, and the $810 Bank of America target implies only about 8% upside — leaving little room for a retention miss. Of 64 brokerages covering the stock, 57 rate it buy or higher, with a median target of $760, according to LSEG data.
This article is for informational purposes only and does not constitute investment advice.
