On Sept 21, 2026, Meta's stock jumped — about in a single day — on early downloads of Muse, its new AI agent that shops, books and switches on your behalf. The next day, investors sold the businesses that profit when customers never switch. We checked the prices, made three analysts argue, and looked at what happened the last times switching got easier.
Muse isn't a chatbot that answers — it's an agent that acts. Meta launched it on Sept 8, 2026 as "the world's first personal agent built for everyone".
Sends emails, books travel and appointments, fills in forms, lowers bills and makes purchases (checkout via Link by Stripe).
Runs tasks on an isolated cloud computer with its own browser ("Muse Secure VM") and keeps working after you close the app. Lives inside WhatsApp and its own app; runs on Meta's Muse Spark models.
Free (capped) · $20 Power · $100 Maximum. US from Sept 8, Canada from Sept 18 — still labelled "limited testing".
Indexed to 100 at the Sept 4 close, the last before launch. The "habit basket" is the 12 stocks Bloomberg and Goldman Sachs named as exposed — brokers, insurers, banks, telecoms, streaming and travel booking — equally weighted. Hover or tap for daily values.
Each stock's move on Sept 22, next to the S&P 500. Not every "habit" business fell equally — some barely moved. Switch to see the move since launch.
Switching has been made easier before — by comparison sites, regulators and number portability. Those natural experiments are the best guide to how exposed each industry really is.
| Industry | Tuesday | Evidence says | Precedent |
|---|---|---|---|
| Brokerage cash | Exposed | Schwab's own "client cash sorting": bank deposits fell ~$50B (32%) from Aug 2022 to Apr 2023 as clients moved to higher yield, and net interest revenue fell 24%. Muse would speed up a trend that is already real. | |
| Insurance | Mixed | The UK regulator banned charging loyal renewers more than new customers from Jan 2022 — proof the lever can be taken away. But renewal prices still rose afterwards: the profit partly shifted rather than vanished. | |
| Banks | Stickier | UK open banking (from 2018) raised switching measurably but did not break the big banks' dominance. | |
| Telecom | Stickier | Mobile number portability cut EU prices by roughly 4–8% (15 countries, 1999–2006) — a one-time step, not a collapse. | |
| Streaming | Unclear | No clean natural experiment found. Paramount even rose on Tuesday. | |
| Travel booking | Unclear | Both stocks were already falling before Muse launched, and Expedia was flat on Tuesday — the drop can't be pinned on Muse. |
Meta added in one day. How big would a Muse subscription business have to be to match it? Move the sliders — or try the presets.
Illustrative arithmetic, not a valuation: the market also prices ads, commerce fees and Meta's AI infrastructure, and revenue is not profit. Muse's 2.8M US/Canada downloads in 12 days are third-party app-tracker estimates — and downloads are not paying users.
Three research agents with opposite briefs — each required to cite every claim and to name the weakest point in its own case.
The neutral economist tested the "habit" thesis industry by industry — its findings are the table above.
Daily adjusted closes from Yahoo Finance. Yahoo was missing Sept 22 for six stocks (ALL, EXPE, MS, PGR, TMUS, WFC); those days are rebuilt from Nasdaq's historical quotes, which match Yahoo on the day before. Indexed to 100 at the Sept 4 close — the last before the Sept 8 launch (Sept 7 was a US holiday). The habit basket is equally weighted. Meta's value added uses 2.55B shares outstanding from its Q2 2026 10-Q. Many reports dated Meta's jump Sept 22; closing prices show it on Monday, Sept 21.
Muse download figures are third-party estimates. This page explains how the market reasoned; it is not investment advice. Data through .
Market analysis, dashboards and interactive tools — from raw data to something people actually use.
Start a project →