Meta Muse: Meta Stock Rose Nearly 13% on a $28.5 Billion…

Meta stock did not reprice because shoppers needed another chatbot. The Meta Muse launch is a payments-and-loyalty event: an agent that finishes the purchase can move the toll from Amazon, a booking site or a broker toward whoever owns the agent and the card pipe. Reuters reported on 23 September 2026 that Muse had overtaken ChatGPT as the top free app on Apple’s App Store and Google Play in the United States and Canada, with Sensor Tower counting 2.8 million downloads within two weeks, and that Meta shares had soared nearly 13% that week. The chart is a weak trophy on its own. Sensor Tower told 9to5Mac that Meta’s house ads took as much as half of daily house impressions between 14 and 27 September, and that downloads jumped 73% in a day when that spend ramped on 16 September. A free-chart rank is not a profit pool.

The trade is a payments-and-loyalty rerating, and I would not book it as revenue. Agents that complete checkout shift the take-rate, the slice taken when money actually moves, away from the incumbent that used to own the session and toward the agent plus the pipe allowed to charge the card. On this tape those pipes are Meta, Shopify and PayPal. Reuters also named Stripe and Ticketmaster as private-company partners on similar deals. A closed catalog can refuse the agent, and Amazon did, on the same weekend the US free chart flipped. Truist Securities’ estimate of at least $28.5 billion in additional annual revenue by fiscal 2030 sits on top of that possibility. It is not Meta guidance, it is not a disclosed fee, and it is not cash. What repriced in a week was the option that loyalty schemes and sponsored slots lose their grip if a user’s agent is willing to switch, and that firms holding an approved checkout keep a toll the holdouts are trying to deny.

What Muse actually does

Meta Muse is built to finish a task, not to hold a chat. CNBC, on 21 September, dated the US launch to 8 September and described assistants that fill forms and organise email, running on a model family Meta calls Muse Spark. LiveMint added travel booking and negotiation, with a free plan and $20 or $100 monthly plans for heavier use. Reuters’ line is the one a desk should keep: users delegate shopping, travel booking and form-filling. This is not the Connect hardware slate. FinanceFeeds covered that separately, including a Muse Charm with no published price. The stock question is the checkout.

Where there is no direct plug-in, the agent uses a browser it controls. TechRepublic reported a cloud virtual machine that can keep working after the app is closed, which is why a retailer can cut Muse off without a new build. Drawing on Meta’s security notes, the same report said a review step checks actions before they reach the web, that purchases need a person’s approval, and that Stripe’s Link can issue a one-time card so the merchant does not see the real number. MIXED, citing the launch material, reported that Muse cannot see passwords or payment methods. Those are Meta’s claims, not an audit.

The download figures are a sequence. Sensor Tower, via CNBC on 21 September, had Muse as the top US iOS free app from Friday 18 September, with about 730,000 downloads in the first five days and more than 2.5 million by Monday, roughly 1.5 million on iOS and 1.1 million on Android. In each product’s own first 13 days, ChatGPT stood at 3.1 million downloads, Claude at about 400,000 and Grok at about 200,000. Muse led the live chart and trailed ChatGPT on that early count. Reuters on 23 September, again citing Sensor Tower, had 2.8 million downloads in two weeks, 55% average daily download growth over the first 10 days, and the top free slot on both stores in the US and Canada. On 30 September, 9to5Mac reported Kara Lee’s projection that downloads had crossed 5 million, and that only 23 apps have reached more than 5 million US downloads inside 22 days. Five million installs are not five million subscribers.

Bernstein’s Stacy Rasgon put the product shift on CNBC that Monday. “I think up to this point most of the agentic use cases have not really been for normal people,” he said. “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.”

Who opened the checkout, and who shut it

Shopify opted in. CNBC said Tobias Lütke announced on 21 September that Shopify would let Muse check out on its stores. Reuters said the company would use Muse to process checkouts and that the shares had jumped 11.4% on the week. Nasdaq shows $128.50 on 18 September and $142.25 on 25 September, a 10.7% gain. Buyers sent to that checkout put Shopify in the pipe, not around it.

PayPal made the same choice. Reuters said users could discover products and complete purchases through the agent, and that the stock was only marginally higher as of 23 September. Closes were $52.41 on 18 September, $52.51 on 23 September and $55.04 on 25 September, about 5% on the full week. FinanceFeeds has set out how PayPal sits beside Stripe’s Link. Reuters also named private Stripe and Ticketmaster. No revenue share was disclosed. Permission was.

Amazon refused. TechRepublic said the block appeared on Sunday 20 September after Amazon told Meta to remove the store. MIXED quoted the popup: continued access by an unauthorised AI agent violates Amazon’s Conditions of Use. “We think it’s fairly straightforward that third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions about whether or not to participate,” an Amazon spokesperson told CNBC. No lawsuit against Meta over Muse turned up in these sources. Meta stock still rose about 11.3% the next session.

OpenAI did not, in the pages read here, open a comparable checkout. Apple and Google hosted the No. 1 ranking and did not, in these sources, copy Amazon’s block. Their leverage is the install, and a commission if the $20 or $100 plan is sold in the app. Nothing here shows that rule changing. If the agent finishes the purchase, a sponsored feed unit is no longer the only toll. FinanceFeeds has framed that auction risk beside Alphabet’s spending. It is not a measured loss of ad share.

On 29 September TechCrunch reported small-business Muse on Shopify, Dropbox, Slack and Meta ad accounts, free up to a limit. Meta’s post said: “Small businesses have been growing on our apps for nearly two decades. They told us they’re short on hours, not ideas. So we built Muse for Small Business to help get work done with the tools they already use.” The Next Web carried Zuckerberg from the day before: “Today we are starting the next major pillar of our business, Meta Enterprise Platform.” CJ Desai left MongoDB after about 11 months to run it. MongoDB closed at $334.68 on 28 September, down 18.5% from $410.44, with a $300.00 low.

What Meta stock actually paid for

Meta closed on 8 September at $613.48, under $616.77 the session before. From $665.75 on 18 September it finished 21 September at $741.25, up about 11.3% on 48.7 million shares. CNBC said Wells Fargo’s target went to $796. That figure is a target, not a print. The high trade in this window was $779.82 on 24 September. The close that day was $777.59.

Reuters’ “nearly 13%” is the Friday arithmetic. The 25 September close was $751.66, 12.9% above $665.75, in what the wire called a fifth straight up week. From $572.34 on 31 August to $725.18 on 30 September, the month was 26.7%. The 2 October close was $728.08, 23.4% above 3 August’s $590.24 and about 6.4% under the 24 September close. Saturday 3 October had no session.

These markdowns are Reuters’ tally as of 23 September, not 2 October prices. JPMorgan and Wells Fargo were each down more than 3%, Schwab nearly 5%, Marsh 2.4%, Arthur J. Gallagher 4.6%, TripAdvisor and Booking Holdings more than 6%, Planet Fitness 17% and the New York Times 10%. The Philadelphia Semiconductor Index was up 7% from the launch. FinanceFeeds has a separate note on the broker and travel markdown. Michael O’Rourke of JonesTrading told Reuters the partners were the clearer side. “Meta has really found lightning in a bottle here and someone who partners with Muse is likely to be a clear beneficiary off the bat to me,” he said. “As far as net losers, it’s just going to take more time to find out.”

Art Hogan of B. Riley Wealth told Reuters those disruption concerns were far-fetched. The Meta stock question is who collects that toll.

Bull case for Meta stock	Bear case for Meta stock
Shopify, PayPal, Stripe and Ticketmaster opened checkout. Muse can charge a card without Amazon.	Amazon revoked access on 20 September. Other closed catalogs can do the same.
Truist’s at least $28.5 billion by fiscal 2030 is the published sketch. September rose 26.7% before any fee.	Five million downloads are not payers. At $20 a month, 5 million subscribers would be about $1.2 billion a year, and that conversion is unreported.
A former MongoDB chief now runs an enterprise unit, and small-business Muse already touches ad accounts.	From $777.59 to $728.08 is about a 6.4% giveback. House ads drove a 73% one-day download spike.

Nasdaq daily closes for Meta (META) from 3 August to 2 October 2026. Dashed line: the 8 September Muse launch. Shaded band: 18–25 September, the week Reuters called nearly 13%. High close $777.59 on 24 September. Last close $728.08 on 2 October. No session on 3 October.

The rule that will decide the toll

No report cited here describes an FTC case, a European Commission case, or a lawsuit by Amazon against Meta over Muse. The rule that already binds is contractual. MIXED, reading Amazon’s Conditions of Use, said they were updated on 14 August 2026. As MIXED summarised them, an agent must send “Agent/” plus its name, must not mimic a person’s typing, must admit it is software if asked, and must stop when told. The popup shoppers saw says an unauthorised agent violates conditions they already accepted. Any other catalog can copy that house rule without a new statute.

The hacking route looks weaker, and the appellate order on file is not a Muse case. Amazon sued Perplexity over an agent in the Comet browser. TechRepublic reported that on 4 August 2026 the Ninth Circuit vacated a preliminary injunction. The court thought Amazon was unlikely, at that stage, to prove Perplexity itself had “accessed” the computers, because the user was reaching the site with the agent’s help. Private terms stayed available. MIXED noted that no court has put Muse under that fight. Do not price a court-ordered shutdown from that opinion. Price the path Amazon already used.

What a regulator will test is disclosure. Meta’s notes, via TechRepublic, say the agent cannot see passwords, that a purchase needs approval, and that a one-time card hides the real number. Amazon’s account, via CNBC, is that Muse did not identify itself and could open account pages. Both cannot describe the same session in full. Stacy Rasgon’s second line on CNBC fits the gap. “They care about security, I don’t know how much they care about privacy,” he said. Apple and Google put Muse on top of the free charts. They can also slow an update or bill a commission on an in-app plan. Nothing in these sources says they have done either. On 29 September FinanceFeeds reported that the European Central Bank had invited firms to study AI agents alongside a digital euro. That is a design brief, not a case against Meta. A take-rate, if Meta ever charges one, still has to say who consented when software, not a person, hit pay.

Three paths from here, none of them a promise

Through the 2026 holiday quarter, the working case is a split catalog. Shopify, PayPal and Link keep the traffic they invited, and Amazon stays shut unless it sells access. The cause is already public. The block hit on 20 September under terms posted in August, and Meta stock rose the next session anyway. The tell by year-end is whether Amazon announces a deal or Muse keeps routing around the store. A truce with no fee would not confirm the bull case.

By the full-year 2026 report, usually late January or early February 2027, the $20 and $100 plans are likely a footnote unless management prints a commerce fee or an enterprise contract. Sensor Tower’s 5 million is downloads. Five million people paying $20 a month would be about $1.2 billion a year, and LiveMint reported that Alexandr Wang expects most tasks to stay on the free tier. The bull case needs a line on take-rate or signed checkout volume. If that line is missing, Truist’s $28.5 billion stays unanchored.

Across 2027 the rule that spreads is more likely Amazon’s identification demand than a hacking verdict. The 4 August order left private terms standing when the human user is the one logged in. Other retailers can copy the “Agent/name” string and the right to switch the agent off. A consumer authority or an app store could write the same duty down later. The ECB work is the wholesale version of that question, not a caption with Meta’s name on it. If disclosure arrives before a disclosed fee, September’s multiple was an option on the pipe, and options expire. None of this is a forecast of the $728.08 close.

FAQ

When did Meta Muse launch, and what does it cost?

CNBC and LiveMint date the US launch to 8 September 2026. CNBC reported a free tier and monthly plans at $20 or $100, depending on usage. LiveMint said the paid tiers add capacity and that Alexandr Wang expects most tasks to stay free. Meta stock closed that day at $613.48. The nearly 13% week was 18–25 September, not launch day. Sensor Tower’s figures are downloads, not a count of paying subscribers.

Did Muse beat ChatGPT, or only the chart?

CNBC, citing Sensor Tower on 21 September, said Muse took the top US iOS free app on 18 September, while ChatGPT had 3.1 million downloads in its own first 13 days against more than 2.5 million for Muse. Reuters on 23 September said Muse led both free stores in the US and Canada, with 2.8 million downloads in two weeks. By 30 September Sensor Tower, via 9to5Mac, had Muse past 5 million downloads faster than ChatGPT’s 56 days.

Why did Meta stock rise nearly 13%?

Reuters used “nearly 13%” for the week of 23 September. Nasdaq puts it at 12.9%, from $665.75 on 18 September to $751.66 on 25 September. Monday alone was about 11.3%, to $741.25, on 48.7 million shares, with the chart lead and Amazon’s block both public. The tape was paying for who might keep checkout, not for the 8 September close of $613.48. Truist’s $28.5 billion is a 2030 sketch, not Meta’s forecast.

What did Shopify, PayPal and Amazon do?

Tobias Lütke said on 21 September that Shopify would allow agent checkout, CNBC reported. Reuters put the week’s share gain at 11.4%; Nasdaq Friday closes show 10.7%, from $128.50 to $142.25. PayPal said customers could discover and buy in the agent. The stock was $52.51 on 23 September and $55.04 on 25 September. Amazon blocked shopping on 20 September. These sources show no lawsuit against Meta over that block. Reuters also named Stripe and Ticketmaster.

Is the $28.5 billion figure something Meta has guided?

No. Reuters attributed at least $28.5 billion of extra annual revenue by fiscal 2030 to Truist, not to Meta. Nothing here shows Meta adopting that sum or disclosing a Muse take-rate. Five million subscribers at $20 a month would be about $1.2 billion a year, and that conversion from downloads is not reported. The bull case is a cut of goods through an approved checkout. Until Meta prints the cut, the Truist number is a scenario, not a booking.