Written by: Rita
Meta's stock price has risen 20% since its recent low but is down 1% year-to-date, whereas the S&P 500 has increased by 12% during the same period. In a report released on September 10, 2026, JPMorgan upgraded Meta's rating from "neutral" to "overweight," raising the target price from $640 to $820. The core logic is that frontier models and AI agent products are already in place, and monetization paths beyond advertising are beginning to open up.
JPMorgan analyst Doug Anumuth believes that when Meta rebuilds its MSL team in the summer of 2025, the goal is to deliver frontier models within a year. Muse Spark 1.1 was released in July, and version 1.3 is already competing with Claude and GPT models. Watermelon is about to launch, featuring more advanced pre-training. These models are at the core of Meta's years of product and monetization pipeline.
Muse AI agent ranked third in the U.S. App Store the day after its launch, with early usage being ten times that of the test cohort. Enterprise agents and model APIs provide direct monetization paths beyond advertising. Capital expenditure is expected to reach $243 billion in 2027 and $284 billion in 2028, with free cash flow under pressure; JPMorgan believes there is still room for AI-driven improvements in core advertising.
Frontier models are here, Muse agents rank third
The MSL team delivered Muse Spark 1.3, which is at the forefront in areas such as agents, programming, instruction-following, and long context benchmarks. Watermelon is about to launch, and the next-generation model after Watermelon has been expanded on the Prometheus GW cluster in Ohio. Meta has a distribution advantage with approximately 4 billion users and hundreds of millions of businesses.
The Muse AI agent was launched on the U.S. iOS, Android, and web platforms. It browses websites and navigates UIs through its own virtual computer, completing online purchases, restaurant reservations, form filling, and mail and text communications. Muse reached number three in the U.S. App Store free apps the next day, with early usage being ten times that of the test cohort. The free version offers 100 million Muse tokens per week, while the paid version at $20/month offers 500 million tokens, and $100/month provides 3 billion tokens.
Enterprise agents and APIs open new monetization
The Meta Business Agent is a plug-and-play tool accessible by subscription, capable of answering questions, recommending products, making appointments, and filtering leads. As of the second quarter of 2026, over 1 million businesses are using it weekly on WhatsApp and Messenger. Pricing is 20 to 25k tokens per message, costing 4 to 5 cents. In the future, it may evolve into a results-based auction system similar to advertising businesses.
The Meta Model API provides developers and enterprises with self-service access to Muse Spark. Muse Spark 1.3 is priced at $1.25/$4.25 per million input/output tokens, with contributor versions at $0.10/$0.20. Muse Code subscriptions are available for $5/month, $15/month, and $50/month. Early data from OpenCode shows that Muse Spark 1.3 has the largest market share among contributors, processing 31T tokens and covering 212k unique users.
Capital expenditures far exceed consensus, FCF under pressure
Meta plans to maximize computing capacity in 2026 and 2027, reaching 7GW in 2026 and doubling to 14GW in 2027. JPMorgan expects capital expenditures of $243 billion in 2027, a year-on-year increase of 70%, and $284 billion in 2028, a year-on-year increase of 17%, both significantly above market consensus. Free cash flow pressure in 2027 and 2028 is expected to be between negative $65 billion and $70 billion annually, but this does not account for new AI product monetization.
There is still room for AI-driven improvements in the core advertising business. Content recommendations and engagement, advertising targeting and retrieval, and AI content creation are all improving. JPMorgan anticipates total revenue of $254.3 billion in 2026, $305.3 billion in 2027, and $354.4 billion in 2028. Advertising revenue is expected to be $297 billion in 2027 and $344.3 billion in 2028.
Target price $820, 23 times 2028 price-to-earnings ratio
JPMorgan has raised its target price from $640 to $820, based on approximately 23 times the 2028 GAAP EPS of $35.44. The previous target of $640 was based on 18 times the 2028 GAAP EPS. JPMorgan believes that a 23 times price-to-earnings ratio carries a premium over the S&P 500's 16 times due to higher confidence in Meta's top-line growth, cost efficiency, and AI monetization pathways. The current price is $653.69 (September 9, 2026), with the target price corresponding to approximately 25% upside potential.
Risks include heavy AI investments with delayed monetization, slower-than-expected revenue growth, AI products not resonating, competition, and core advertising impacted by lawsuits.
When Muse agents begin completing transactions for users, can Meta's commission model find a balance between privacy and trust?

Disclaimer
This article is a compilation and interpretation of a third-party brokerage research report (JPMorgan, September 10, 2026) by Chaoxiang Research, combined with publicly available market information. The ratings, target prices, earnings forecasts, and related judgments mentioned in the text are solely the opinions of the analysts at that brokerage and represent their respective institutions' positions, not the views of Chaoxiang Research, nor do they constitute any investment advice.
The market carries risks, and decisions must be made independently. This article should not be used as the basis for buying or selling any securities.
免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。