Zuckerberg bets Meta’s future on a network of Muse AI agents

On Meta’s earnings call, Mark Zuckerberg described a future where people’s Muse agents shop, work and deal with each other, but gave few details on how it will work.

1 min read

The Verge

At a glance

US and Canada users who set up Muse
10M+
average spend per Muse purchase
$200+
Meta’s planned 2026 spending
$135–145B
What’s new
Zuckerberg told investors he expects billions of people to use Muse, with agents transacting with one another in a large network, and predicted Muse will be highly profitable within a few years.
Who it affects
Meta investors, shoppers who might let an agent buy for them, and businesses expected to pay transaction fees.

Muse launched to the public last month. Zuckerberg said a quarter of people who downloaded it use it daily and a quarter shop through it. Meta earns a fee on each purchase and says Muse will likely stay free for consumers, with businesses covering the fees. He said it has not launched worldwide because it is costly to run, and that Meta aims to make its computing 10 times more cost-efficient next year. Privacy worries persist: agents have reportedly leaked content from private chats, and a Canadian bank executive said it might block Muse from customers’ financial data.

Why it matters

Meta is spending heavily on data centres and needs Muse to show those bets can pay off. The plan rests on two open questions: whether people will trust an agent with their money and data, and whether Meta can cut computing costs enough to expand. A $25 billion New Mexico jury verdict adds to the pressure. Without a clear rollout plan, investors are being asked to trust the vision.

In Meta’s version of the future, I’ll have my Muse call your Muse

The Verge

theverge.com

Summary by Second Order AI. Full reporting is at the original source.