Meta Muse Hits 10 Million Users as Walmart Opens Agent Gateway and Amazon Closes the Door

Meta's AI agent Muse crossed 10 million users in the third quarter of 2026, and the company is now telling investors it expects the product to eventually generate more revenue per user than Facebook, Instagram, WhatsApp and Messenger combined. The same week, Walmart announced a purpose-built Agent Gateway that lets approved AI agents transact on its site, while Amazon continued its policy of blocking AI agents from making purchases entirely. The split reveals an industry at a crossroads: consumer AI agents are no longer a demo, but the rules of who may transact on whose behalf are still being written.

Muse: From Side Project to Product-Market Fit

Meta disclosed the 10 million user milestone during its Q3 2026 earnings call on Wednesday, October 21, confirming estimates released earlier in the week by market intelligence firm Sensor Tower. CEO Mark Zuckerberg said roughly 25 percent of people who set up Muse now use it daily, a retention signal that has exceeded internal expectations.

The more surprising figure was commercial. Over a quarter of Muse's engaged users who shopped through the agent recorded an average purchase size of more than 200 dollars, and Plaid — the fintech connector that links payment apps to bank accounts — told Meta it had never seen activity like this from any other service. Muse uses Plaid's connector to enable agentic shopping, letting the agent browse, compare and complete purchases on a user's behalf.

Beyond shopping, Zuckerberg highlighted personal productivity uses: connecting the agent to email and calendars. He stressed that this indicates users trusted Muse's security setup, which includes a feature called "Muse Secure VM" designed to protect data and privacy, with a forthcoming "Muse Confidential VM" intended to prevent even Meta from seeing users' content.

Now that Muse has achieved what Zuckerberg called "product-market fit," Meta's focus shifts to scaling, increasing revenue and optimizing compute costs. The company is forecasting that Muse will eventually bring in more revenue per user than its "family of apps" users do today — a figure currently around "several hundreds of dollars annually." Monetization, Zuckerberg said, will keep the agent free for most users while taking a percentage of commerce taking place on the platform; power users wanting access to the most advanced models or more tokens will pay.

Meta believes Muse will be profitable after it scales. "On the compute side, we are building out the capacity to serve more than a billion people over the next few years," Zuckerberg told analysts. "In the near term, Muse is very compute-intensive to run, but we expect to improve its cost efficiency by more than 10x in the next year, which will enable it to be profitable once the revenue lines are at scale."

Muse will also play a central role in Meta's smart glasses strategy. Reality Labs, the division responsible for smart glasses, VR and the metaverse, is still losing billions — losses are expected to improve from 19.1 billion dollars in 2025 to 15 billion dollars this year — but Muse is the software layer Meta intends to put in front of users' eyes.

Photo: Meta Platforms headquarters complex in Menlo Park, California. Credit: LPS.1 / Wikimedia Commons (CC0).

The Trust Question Meta Has Not Answered

Zuckerberg's claim that users trust Muse's security setup arrived the same week that researchers at security firm Claroc disclosed a serious flaw in the agent's sharing feature. Claroc found that a single passive DNS query could enumerate more than 6,800 Muse share links, exposing executive-search notes for a frontier AI lab, a revenue review, a loan application, a bank fraud alert to-do list, a real-estate performance report and a staff vacation calendar. The share links embedded the conversation title in the subdomain, making them trivially enumerable.

Meta paused the sharing feature following the disclosure. The episode is the latest in a string of security questions around Muse: earlier in the year, a YouTuber reported that his home address was shared with a stranger through the agent, and researcher Patrick Wardle identified a zero-day vulnerability in Muse that Meta downplayed before patching. Taken together, the incidents illustrate the gap between the security architecture Meta describes — Secure VMs, confidential compute — and the everyday features users actually touch.

The tension matters commercially. Meta is asking users to hand an agent access to email, calendars and bank accounts, and is forecasting revenue per user above the family of apps. If a single DNS query can surface the content of thousands of shared conversations, the premise that users "trust Muse's security setup" becomes a marketing claim rather than a measured fact.

Photo: A person uses their index finger to scroll through a list of news articles on a smartphone. Credit: Japanexperterna.se / Wikimedia Commons (CC BY-SA 2.0).

Walmart Opens the Door, Amazon Slams It

The competitive backdrop makes Muse's timing consequential. Walmart announced on Wednesday its own Agent Gateway, an MCP server — built on Anthropic's Model Context Protocol — that lets approved AI agents search Walmart's inventory and make purchases on a user's behalf. The retailer also introduced Agent Profiles, rolling out next month, which give shoppers a dedicated profile per agent with separate purchase tracking and spending limits, similar to adding an authorized user to a credit card.

Walmart's embrace stands in contrast to Amazon, which began blocking AI agents from making purchases on behalf of online shoppers. Travel and booking sites have followed a similar lockout strategy. Walmart, by contrast, already partners with Meta's Muse and offers integrations with ChatGPT, Gemini and other AI apps.

The divergence is a bet on who controls the customer relationship. Amazon's position protects its own recommendation and advertising funnel; Walmart's position treats AI agents as a new distribution channel. For Meta, Walmart's openness is a validation of Muse's commerce ambitions — and a reminder that the agent's value depends on retailers willing to let it in.

Photo: A person using a stylus to sign for a package on a smartphone. Credit: Meanwell Packaging / Wikimedia Commons (CC BY 2.0).

Financing the AI Bet

Meta's Q3 2026 financial results show the cost of the strategy. Revenue was 64.60 billion dollars, up 26 percent year-over-year, while income from operations fell 37 percent to 12.99 billion dollars, driven by a 9.89 billion dollar charge related to a settlement with 52 state attorneys general over teen mental health and social media. Capital expenditures, including principal payments on finance leases, were 34.91 billion dollars in the quarter, and the company narrowed its full-year 2026 capex guidance to 135-145 billion dollars.

To fund the build-out, Meta's board approved a new publicly listed, non-voting Class C common stock, effective January 19, 2027, with a one-time stock dividend of one Class C share per Class A and Class B share. The board said the new class will provide Meta the ability to finance its growth and help the company remain a leader amid fierce domestic and international competition. The stock dividend will not change the voting power of any shareholder, though the market price of Class A shares is expected to reflect a two-for-one split effect after payment.

The structure is a signal: Meta is asking capital markets to fund an AI infrastructure push measured in hundreds of billions of dollars, while simultaneously telling consumers its agent is secure enough to handle their money. Whether those two messages can coexist — growth at scale and security at the edge — is the question the next year will answer.

This article synthesizes Meta's Q3 2026 earnings release (investor.atmeta.com), TechCrunch reporting on Muse's 10 million users and Walmart's Agent Gateway, and Meta's investor relations filings, all dated October 21, 2026. For more AI coverage, see our AI category.