Meta Reports 10M Muse Users as Agentic Shopping and Productivity Drive Growth
Meta announced that its AI app Muse has surpassed 10 million users, highlighting strong engagement in agentic e-commerce transactions and personal productivity tasks.
Meta has officially confirmed that its AI assistant, Muse, has reached more than 10 million users, aligning with recent estimates from market intelligence firm Sensor Tower. Speaking on Meta's third-quarter earnings call, CEO Mark Zuckerberg shared that roughly 25% of individuals who set up the AI agent use it on a daily basis. Beyond strong adoption, the executive highlighted surprising user engagement patterns, particularly around agentic shopping and high-value consumer transactions.
High-value shopping and security features
A primary driver of engagement on Muse is financial activity. Zuckerberg revealed that over 25% of engaged users leveraging the AI agent for shopping recorded an average purchase size exceeding $200. The application relies on a integration provided by financial technology company Plaid, which links payment applications directly to bank accounts. Plaid reportedly informed Meta that transaction activity on the Muse connector surpassed levels observed on any other service.
In addition to e-commerce, personal productivity tasks—such as linking the assistant to email accounts and calendar tools—have emerged as popular use cases. Zuckerberg attributed this adoption to user confidence in the underlying system architecture. Muse currently operates using a privacy feature called "Muse Secure VM" to protect user data. Meta also plans to deploy "Muse Confidential VM" in an upcoming update, an enhancement designed to prevent even Meta from accessing user content processed by the agent.
Monetization, compute scaling, and the "Muse network"
With Muse reaching what Zuckerberg described as "product-market fit," Meta is shifting its strategic priorities toward scaling the agent, generating revenue, and optimizing compute expenses. Meta projects that Muse will eventually generate higher average revenue per user than its core social platforms—Facebook, Instagram, WhatsApp, and Messenger—which currently generate several hundreds of dollars per user annually.
The business model for Muse will keep basic access free for most individuals while capturing a percentage of commercial transactions executed on the platform. Power users seeking additional tokens or access to the most advanced AI models will be required to pay.
To support this expansion, Meta is building out infrastructure capacity to serve over a billion people over the next few years. Although the AI agent is compute-intensive today, Meta expects to improve its cost efficiency by more than 10 times over the next year to enable profitability at scale. Furthermore, Muse will play a central role in Meta’s hardware strategy via Reality Labs, the division whose losses are expected to improve from $19.1 billion in 2025 to $15 billion this year.
Looking ahead, Zuckerberg teased plans for the "Muse network." This initiative aims to evolve Muse from a single-user system into an interconnected framework where individual Muses collaborate with agents belonging to friends, family, or commercial businesses.
What it means for developers
Meta’s progress with Muse highlights several critical trends for software engineers building next-generation AI applications. First, the success of agentic shopping demonstrates growing consumer readiness to delegate financial actions directly to AI systems, provided secure connection layers like Plaid and VM-level isolation tools are present. Developers designing autonomous agents should prioritize robust privacy primitives similar to Meta's Secure VM framework to establish user trust.
Second, Meta's vision for a multi-agent network suggests an industry shift toward cooperative AI architectures where consumer-facing agents interact directly with merchant AI endpoints. Developers will need to build systems capable of standardized agent-to-agent negotiations rather than rely solely on standard web interfaces.
Finally, managing compute costs remains a central obstacle when operating continuous AI workflows at scale. As engineering teams build complex multi-agent logic and LLM integrations, cost management is essential. Developers looking to experiment with top AI models like Claude, GPT, Gemini, DeepSeek, and Grok can access affordable pay-per-token API access through a single key at https://apixoai.online.
Meta’s target of a 10x compute efficiency improvement underscores that as agentic adoption expands, optimizing token usage and backend execution will be just as critical as model intelligence.
Source: Meta says Muse has over 10M users, with shopping a top activity — TechCrunch AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
