Meta Bets Muse Will Outperform Facebook and Instagram in Per-User Revenue

A Revenue Thesis Built on Commerce

Six weeks after launch, Meta's Muse personal agent is no longer an experimental feature. It has become the centrepiece of the company's plan to extract returns from years of AI infrastructure spending. Mark Zuckerberg told investors during Meta's third-quarter earnings that Muse adoption has already crossed 10 million users in the United States and Canada, and that the company is shifting from product-market fit to operational scale.

The more striking claim came next. Zuckerberg said he expects Muse to eventually generate more revenue per user than Facebook and Instagram, the two platforms that together serve billions and underpin Meta's $117 billion in annual advertising revenue. That expectation rests on a single behaviour the company has observed early: shopping. Users are spending an average of $200 per purchase through Muse, and commerce has emerged as the most common use case for the agent.

Meta plans to build what Zuckerberg called "commerce tools" that will let the company charge businesses transaction fees when Muse facilitates a sale. The model mirrors payment processing or marketplace take-rates, a departure from Meta's traditional advertising business. For context, Meta's average revenue per user in the US and Canada was $69.26 in the third quarter of 2024, or roughly $277 annually. Zuckerberg is signalling that Muse, through a combination of transaction fees and subscriptions, can exceed that figure per active user.

The Cost Structure Behind the Bet

Muse is also significantly more expensive to run than Meta's existing services. Inference costs for large language models remain high, and personal agents that handle multi-turn conversations, memory, and task execution consume more compute per interaction than a feed refresh or a single search query. Zuckerberg acknowledged this gap but framed it as a temporary constraint, one that will narrow as Meta scales usage and improves model efficiency.

The company is layering in AI subscription revenue as a second monetisation lever. Meta already offers a paid tier for access to advanced models and features across its platforms. Zuckerberg indicated that heavier Muse users, both individuals and business owners, will be funnelled into subscription plans. The exact pricing and feature gates have not been disclosed, but the structure suggests Meta is testing willingness to pay before locking in a final model.

This dual approach, transaction fees plus subscriptions, is designed to de-risk the revenue model. If commerce adoption plateaus, subscriptions provide a floor. If subscriptions remain niche, transaction volume can carry the load. The company is hedging against the uncertainty that still surrounds consumer AI products, most of which have struggled to convert usage into sustainable revenue.

Muse as Hardware Catalyst

Zuckerberg also positioned Muse as central to Meta's hardware ambitions, specifically its Ray-Ban smart glasses developed with EssilorLuxottica. The company does not make significant profit on the glasses themselves, he said, but Muse integration is expected to drive sales and, more importantly, attach users to Meta's AI subscription ecosystem.

Meta previewed Muse on its smart glasses last month at its Connect conference. Users will be able to speak to their agent, check task status, and add items to Muse's to-do list using voice commands, much like the existing Meta AI assistant embedded in the glasses. The agent is scheduled to roll out on the hardware next month, though Zuckerberg's comments suggest at least some Muse capabilities on glasses will require a paid subscription.

This marks a shift in how Meta thinks about hardware. The glasses are no longer standalone products with their own business case. They are distribution channels for AI services, and the margin comes from software, not device sales. Meta has previously considered charging for other premium features on the glasses, including an upcoming hearing enhancement mode, signalling a broader move towards feature-based paywalls.

Scaling to Billions

Zuckerberg said Meta is now entering a "more operational phase" focused on scaling Muse to serve billions of people. That language, operational rather than experimental, indicates internal resource allocation is shifting. Engineering, infrastructure, and go-to-market teams are likely being reoriented around Muse as a core product, not a research project.

The company has not disclosed retention metrics, daily active user figures, or geographic expansion timelines beyond the US and Canada. Those gaps matter. Reaching 10 million sign-ups is distinct from retaining 10 million active users, and early adopters of new AI products often churn quickly once novelty fades. Meta will need to demonstrate that Muse solves recurring problems, not just curiosity-driven experimentation.

The commerce behaviour Meta is observing could be self-selecting. Users who shop through Muse in the first six weeks may be a subset with high intent and purchasing power, not representative of the broader user base. As the agent scales, average transaction values and conversion rates may regress. Meta has not shared enough data to assess whether the $200 average purchase is sustainable or an artefact of early adoption.

The Competitive Context

Meta is far from alone in racing towards monetisable AI agents. OpenAI is building agent capabilities into ChatGPT, Google has embedded assistant features across its product suite, and Anthropic is positioning Claude for enterprise workflows. The difference is Meta's distribution advantage. Facebook, Instagram, WhatsApp, and Messenger together reach more than 3 billion daily active users. Muse can be surfaced inside those apps, reducing friction for discovery and onboarding.

That distribution also creates expectations. If Muse becomes a revenue driver on par with or exceeding Meta's app ecosystem, the company will need to integrate the agent deeply into user workflows without degrading the experience of its core products. Balancing that tension, monetising aggressively while maintaining product quality, will define whether Muse becomes a multi-billion-dollar business or another over-hyped AI experiment.

For now, Meta is betting that personal agents represent a new computing platform, one where the interface is conversational and the business model is transactional. Whether users will pay, directly or indirectly, for that convenience remains an open question. But Zuckerberg's comments make clear that inside Meta, the answer is already assumed.