Meta’s stock climbs in premarket as AI agent Muse addresses investor concerns

Meta just gave Wall Street something it’s been asking for: a reason to believe the company’s massive AI spending will actually pay off. After launching Muse, its new personal AI agent, on September 8, Meta’s stock climbed roughly 3-5% in premarket trading the following day, reversing a slight dip on launch day itself.

The timing matters. Meta has earmarked $130 to $145 billion in AI capital spending for 2026, a figure large enough to make even the most bullish investors squint. Muse is the company’s clearest attempt yet to show that spending can produce something beyond better ad targeting.

What Muse actually does

The agent can send emails, book travel, and manage long-term goals across a range of integrated services including Google Workspace, Spotify, OpenTable, and Apple Health.

Under the hood, Muse runs on the Muse Spark model family, developed under Chief AI Officer Alexandr Wang. Each instance operates inside what Meta calls a “Muse Secure VM,” a dedicated virtual machine designed to keep user data compartmentalized and require explicit approval before the agent takes sensitive actions.

The product launched in the US through a dedicated app and a web portal at muse.ai. Meta is also teasing future integration with its AI glasses, which would bring Muse into the wearable space.

Pricing follows the now-standard AI subscription playbook. There’s a free tier for casual users, a $20 per month plan, and a $100 per month premium option for heavier usage. Meta hasn’t disclosed any revenue targets or subscriber projections for the service.

Why investors care

For years, Meta’s revenue story has been essentially one sentence long: advertising. Muse introduces a subscription revenue stream that, if it scales, could meaningfully diversify the business.

The initial stock dip on launch day was classic “sell the news” behavior. But the premarket bounce the next morning suggests a more considered take was forming: this is a real product with a real monetization model.

The competitive landscape adds urgency. Meta isn’t entering a vacuum. OpenAI has been building out its own agent capabilities, Google has deep integration advantages through Android and its productivity suite, and newer entrants like OpenClaw and Instinct are carving out niches in the AI agent space.

The privacy problem that won’t go away

For Muse to work as advertised, it needs extensive access to user data: emails, calendars, health information, browsing habits, spending patterns.

The Secure VM architecture is Meta’s answer to this concern. By isolating the AI agent in its own sandboxed environment and requiring user sign-off for sensitive operations, the company is trying to build trust through technical guardrails.

Regulatory scrutiny is another variable. European regulators in particular have shown little patience for expansive data collection by Big Tech, and Muse’s US-only launch may reflect Meta’s awareness that clearing GDPR hurdles will take additional time and engineering.

Muse gives the company a tangible consumer product to point to when analysts ask where all the money is going. But Meta hasn’t offered any timeline for when Muse might contribute materially to the bottom line.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.