Meta is experiencing a powerful rebound driven by AI expectations, with the popularity of its personal AI assistant Muse serving as a key catalyst for this rally.
Earlier this year, Meta's stock price came under significant pressure, but it rebounded rapidly after entering September. As of September 24, the stock had risen approximately 36% for the month, marking its strongest monthly performance since 2013, with market capitalization once again approaching $2 trillion, just about 1% away from that threshold.
Behind this reversal, the AI commercialization potential brought by Muse has become an important support. Muse officially launched on September 8 and can autonomously execute multi-step tasks such as shopping, travel, and email management. It recorded 902,000 downloads in its first six days, surpassing Meta AI's 773,000 during the same period, and quickly topped the free charts on Apple's App Store in the United States and Canada.
However, optimism in the capital markets is also heating up rapidly. AI agents are beginning to enter real consumer scenarios such as shopping and bookings, potentially disrupting the search, account, and usage habits that traditional internet platforms rely on to retain users. At the same time, Muse's own security vulnerabilities, privacy controversies, and high computing costs also impose real constraints on this AI boom.
From year-to-date weakness to a strong rebound, Meta welcomes AI catalyst
Meta was previously not a popular AI trade in the market. Before August 18 this year, the company's stock had fallen 18% from the start of the year, ranking among the 50 worst performers in the S&P 500 Index year-to-date. Revenue guidance released at the end of July fell short of market expectations, which also weighed on investor sentiment at one point.
Subsequently, two factors drove a change in the situation. On one hand, Meta agreed to pay up to $18 billion to settle social media-related litigation, easing long-standing legal uncertainty. On the other hand, Muse quickly gained user attention after its launch, providing the market with a new AI growth narrative.
Since its August low, Meta's stock has rebounded approximately 43%. JPMorgan analyst Doug Anmuth upgraded Meta from Neutral to Overweight on September 10, believing the company is in the early stages of launching frontier models and AI products beyond advertising, with still significant growth potential.
This means that market focus is shifting further from Meta's traditional advertising business growth toward whether AI can contribute new users, revenue, and valuation space.
Muse moves from "chat" to "execution," AI agents begin entering consumer scenarios
Compared with traditional chatbots, Muse places more emphasis on directly completing tasks for users. Users can have it execute multi-step workflows such as shopping orders, travel bookings, and email drafting. It currently offers a free tier, with paid tiers at $20 and $100 per month respectively.
After launch, Meta continued to expand Muse's capabilities, including testing phone calls to U.S. businesses, opening developer access connectors, launching a Mac version of Muse, and integrating Shop Pay for agent checkout. During Meta Connect, the company also launched a handheld device for operating Muse called "Muse Charm," as well as a new smart glasses product without a camera.
The commercial ecosystem is also expanding simultaneously. Meta has reached a grocery sales partnership with Instacart's parent company Maplebear and established a partnership with Expedia. Muse is beginning to extend from a pure AI assistant to a platform connecting consumer services.
JPMorgan expects Muse to become one of the most widely used consumer AI applications after ChatGPT, with daily downloads currently exceeding the combined total of Meta's other applications. However, the bank expects Meta to prioritize user growth and usage depth in the short term, with formal commercialization potentially no earlier than 2027.
AI agents disrupt "consumer inertia," traditional platforms begin to feel pressure
Muse's popularity has also begun to trigger a reassessment of the traditional consumer internet model. Goldman Sachs' trading desk recently noted that the potential impact of agentic AI on "consumer inertia" stocks has become a hot market topic, with its basket index tracking high-switching-cost companies falling 2.58% in a single day.
The core market concern is that once AI agents can directly complete tasks such as search, shopping, and bookings for users, the usage habits and account stickiness that users have formed around specific platforms may be weakened.
Meta is also actively expanding this influence. By integrating services such as Maplebear and Expedia, Muse is entering more specific consumer chains. For traditional internet platforms, whether users will directly open apps in the future or let AI agents complete operations on their behalf may become a new competitive variable.
But Muse's own security and privacy issues are also rapidly being exposed.
Researchers disclosed that Muse's macOS version has a zero-day vulnerability, where local applications or terminal commands could obtain tokens used to verify user accounts and further control user accounts. On the privacy front, some users also claimed that Muse read Mac notification content without explicit authorization and pushed related suggestions based on that.
In addition, code analysis found that Muse's automated browser has mechanisms to evade bot detection, modifying certain browser characteristics to reduce the probability of being identified. Amazon subsequently blocked Muse from accessing its retail website, and the related automated behavior has also become a focus of external attention.
Another barrier for AI agents: computing costs
In addition to security issues, Muse's business model also faces the test of infrastructure costs.
According to the product design, each user theoretically has a dedicated cloud virtual machine, configured with approximately 2 virtual CPUs, 8GB of memory, and 100GB of solid-state storage, and needs to run around the clock.
Rough calculations show that if Muse reaches 100 million users, Meta may need to deploy approximately 1.6 million AMD EPYC processors, 100PB of memory, and 10,000PB of solid-state storage, corresponding to a power demand of approximately 1.6 gigawatts. Meta's overall power plan this year is approximately 7 gigawatts, which also needs to simultaneously support model training and advertising business.
The revenue side is temporarily unable to cover such massive investment. Assuming 10 million users subscribe to the $20 per month paid tier, annualized revenue would be approximately $2.4 billion, while Meta's capital expenditure budget this year is as high as $130 billion to $145 billion. If the dedicated virtual machine model is maintained, Muse's early infrastructure investment will still be considerable.
This is also where the market currently needs further verification: how many users AI agents can bring does not equal how much profit they can create.
Valuation is not aggressive, but AI investment is testing cash flow
Meta's current valuation remains relatively moderate. According to analyst consensus compiled by Bloomberg, the company's forward price-to-earnings ratio for the next 12 months is approximately 21 times, slightly below the Nasdaq 100 Index's 22 times.
Analysts expect Meta's revenue to grow 26% to $254 billion in 2026, with net profit growing 33% to $80.6 billion. But by 2027, revenue and profit growth are expected to slow to 20% and 9% respectively.
At the same time, AI infrastructure investment is significantly squeezing cash flow. Meta's free cash flow was $46 billion last year, but due to the substantial increase in capital expenditure, it is expected to turn to -$6.4 billion in 2026 and further decline to -$29.2 billion in 2027.
Currently, more than 90% of analysts tracked by Bloomberg maintain a Buy rating on Meta, but the stock price has already approached the analysts' average target price. After this round of rapid gains, the market's AI expectations for Meta have clearly heated up. Whether Muse can further move from user growth to stable monetization will become the key to whether valuation can continue to expand in the next stage.