Everything About AI
NextFin News - Meta Platforms shares climbed in premarket trading on Thursday after JPMorgan analyst Doug Anmuth upgraded the stock to Overweight from Neutral and raised his price target to $820 from $640, arguing that the company's newly launched Muse AI assistant has turned Meta's artificial-intelligence strategy from a cost center into a visible revenue channel. The upgrade, which implies roughly 25% upside from Meta's $653.69 close on Wednesday, marks a decisive reversal for Anmuth, who in February 2022 became the first major Wall Street analyst to downgrade Meta after years of bullishness.
The move is less about a single product launch than about a shift in how the market is being asked to value Meta's AI spending. For the better part of two years, investors tolerated Meta's heavy artificial-intelligence and metaverse outlays largely on faith. Muse — which began rolling out on Tuesday, September 8 — is the first concrete answer to the question investors kept asking: what exactly does this spend produce?
The Upgrade and the Product Behind It
JPMorgan's note, issued September 10, 2026, frames the upgrade around two distinct revenue streams that Muse unlocks. The first is consumer-facing: the assistant itself, which Meta is already monetizing through paid tiers. The second is the infrastructure layer beneath it — a model application-programming interface and business agents that sell AI capability to other companies. Both sit on top of what JPMorgan expects to be continued advertising gains from AI-driven recommendations and targeting inside Facebook, Instagram, and WhatsApp.
Muse, known internally as "Hatch" and modeled on the open-source agent OpenClaw, is designed to act on a user's behalf rather than merely answer questions. It can send emails, book travel, and even sell a car, connecting across email, calendar, payments, health, shopping, and the smart home. Each agent runs on its own virtual machine — a cloud-based emulation of a personal computer — so tasks continue in the background even when the user is not actively engaged. The service is available in the U.S. initially through a dedicated Muse app, the web at muse.ai, and chats within WhatsApp, with smart-glass integration planned.
Meta is pricing the product with a wide net: a free tier for casual use, a "Power" plan at $20 a month, and a "Maximum" plan at $100 a month for heavy usage. The company requires a payment card to begin, and the app includes a usage meter that warns users when free capacity is exhausted. That structure matters because it converts Meta's roughly 4 billion monthly active users across its family of apps into a funnel with a clear monetization ladder — free, $20, $100 — rather than an abstract engagement story.
Trust, however, is the product's central friction. Muse needs access to the most sensitive data a person has. Meta's answer is a "Secure Credential Store" that keeps passwords and credit-card details unreadable to the agent itself, with the assistant required to confirm with the user before taking sensitive actions such as payments or sending messages. CEO Mark Zuckerberg framed the launch as the centerpiece of his plan to deliver "personal superintelligence" to the billions who use Meta's services daily.
Why the Rating Reversal Matters
The upgrade carries symbolic weight beyond the numbers. Anmuth has been Meta's most prominent Wall Street skeptic since February 3, 2022, when he cut the stock to Neutral from Overweight — his first downgrade since Meta's 2012 initial public offering — after the company warned of advertising slowdown and an expensive, uncertain transition to the metaverse. At the time, he slashed his price target to $284 from $385 and wrote that he was moving "to the sidelines." His return to Overweight is a signal that the firm now sees the AI transition as de-risked enough to justify a buy-equivalent stance.
The price target itself tells part of the story. At $820, JPMorgan's new target sits above the $780 target held by KeyBanc Capital Markets and the $750 target at Mizuho Securities. It also sits below the $825 target Anmuth reaffirmed in April after the release of Muse Spark, Meta's earlier AI model. The modest trim from $825 to $820 alongside a rating upgrade from Neutral is an unusual combination — it suggests JPMorgan is upgrading the certainty of the thesis even as it slightly lowers the terminal valuation, likely reflecting the reality that Meta has tempered expectations for a frontier large-language model in the first half of 2026 in favor of "good" initial models that show a trajectory toward the frontier over the course of the year.
Other analysts are moving in the same direction, though with different emphasis. Mizuho Securities wrote in a client note:
We believe Meta's Muse consumer AI agent marks the beginning of a substantial product cycle for Meta that is not priced into shares.
KeyBanc Capital Markets said it continues to "believe the market underestimates Meta's AI positioning and product cycle." The common thread is not that AI will work — that was never really in doubt for a company with Meta's data and distribution — but that the monetization mechanism is finally visible.
The valuation backdrop adds urgency to the debate. According to one independent valuation service, Meta traded at $653.69 on Wednesday, approximately 23% below its proprietary intrinsic-value estimate of $850.42, with a trailing price-to-earnings ratio of about 24.6 times — slightly below its own five-year median of roughly 26 times. That places the stock near its historical earnings multiple even as the company is being asked to price a new revenue stream that has no historical precedent. The tension between those two facts — cheap by the old yardstick, unproven by the new one — is what the upgrade is really trying to resolve.
The Mechanism: From Engagement to Recurring Revenue
The deeper story behind the upgrade is a change in Meta's revenue architecture. For most of its history, Meta has been an advertising company with an engagement problem: it must keep users scrolling to sell ads against their attention. That model is cyclical — it rises and falls with advertiser budgets, privacy regulation, and competition from TikTok and others. AI changes the equation by adding a second engine: recurring subscription and API revenue that is far less sensitive to the advertising cycle.
The transmission channel runs like this. Muse increases the value of time spent inside Meta's ecosystem because it completes tasks rather than just displaying content. A user who books travel, manages a calendar, or runs errands through Muse is more deeply embedded in Meta's infrastructure than a user who passively scrolls. That depth produces two revenue effects. First, it supports the subscription tiers directly. Second, it generates richer first-party signals about user intent — travel plans, purchase timing, health interests — which improve ad targeting in a privacy-constrained world where third-party cookies are dying.
This is where the second-order effect becomes material. The market has largely priced Meta as an ad company that spends heavily on AI. What the upgrade is really saying is that Meta is becoming a platform company that earns from both ends: the consumer pays for convenience, and the advertiser pays for access to a user whose intent is now observable at the task level rather than the click level. That dual revenue stream should, if the thesis holds, support a higher valuation multiple than a pure advertising business commands, because recurring revenue is discounted at a lower rate than cyclical ad revenue.
Consider the arithmetic of the funnel. If even 3% of Meta's 4 billion monthly active users converted to a paid plan at a blended $40 a month, that would imply roughly $1.4 billion in monthly recurring revenue, or about $17 billion annualized — a figure that would rank as a meaningful standalone business before any API or business-agent revenue is counted. The point is not that this conversion is likely; it is that the upside is no longer theoretical. The pricing tiers exist, the distribution exists, and the payment relationship exists.
The financial guardrails Meta has set make the bet more palatable. The company has guided to revenue growth pushing into the 30% range for 2026, with total expenses of $162 billion to $169 billion and capital expenditures of $115 billion to $135 billion. Anmuth noted that Meta has established bottom-line guardrails with operating income growth expected in 2026, and Reality Labs operating losses projected to be similar to last year, representing a peak in 2026. In other words, the AI spend is no longer an open-ended burn — it is a bounded investment with a defined path to profitability.
The capital-expenditure figure deserves its own scrutiny. At $115 billion to $135 billion, Meta's 2026 capex guidance exceeds the entire market capitalization of most S&P 500 companies. That is the number the upgrade must justify. JPMorgan's argument is effectively that this spend is building a durable asset — models, inference infrastructure, and agent distribution — that will generate returns for years, rather than a one-time expense that vanishes from the income statement with nothing to show. Whether investors accept that framing will determine whether the stock re-rates or merely rallies.
The Counter-Thesis: Why This Might Not Work
The strongest argument against the upgrade is that consumers will not trust a Meta product with their most sensitive data, no matter how many credential stores the company builds. Meta's brand has been shaped by a decade of privacy scandals, and asking users to hand an AI agent access to their email, calendar, payments, and health data asks for a level of trust the company has not earned. If adoption skews toward the free tier and the $20 and $100 plans attract only a small fraction of users, the subscription story collapses into a marginal revenue line that does not justify the capital expenditure.
A second risk is competitive. The AI-agent race includes Microsoft, Google, and Apple, all of which have deeper enterprise relationships and, in Apple's case, a hardware ecosystem that controls the point of interaction. If users manage their digital lives through an agent embedded in their phone's operating system or their office suite, Meta's agent becomes a secondary tool rather than a primary gateway.
There is also a timing risk embedded in the valuation. Meta has tempered expectations for a frontier large-language model in the first half of 2026, emphasizing "good" initial models that show a trajectory toward the frontier over the course of the year. If competitors release demonstrably superior frontier models while Meta is still iterating, the company could find itself selling a mid-tier product at premium prices — a position that rarely holds in technology markets.
The falsifying signal is specific and observable: if Muse's paid-subscription conversion rate is below 2% of monthly active users within two quarters of launch, or if Meta does not disclose any meaningful paid-user figure by its next two earnings reports, the recurring-revenue thesis is wrong and the stock should revert toward its historical advertising multiple. A second falsifier: if Reality Labs losses exceed the 2026 peak guidance while AI revenue remains immaterial, the "bounded investment" narrative breaks.
What to Watch and the Outlook
In the short term, the stock reaction will be driven by sentiment and positioning. Meta's shares rose about 1% in premarket trading on Thursday, following a 0.78% gain on Wednesday that closed the stock at $653.69. The shares have been volatile in recent weeks, falling as much as 3.7% in a single session in mid-August before recovering. Near-term direction will depend less on Muse's technology than on whether Meta's next earnings report includes a paid-user metric that the market can underwrite.
Over the medium term, the fundamental question is whether Muse produces measurable revenue and improves ad performance simultaneously. The base case is that paid tiers attract a low-single-digit percentage of users, API and business-agent revenue grows steadily, and AI-driven ad targeting lifts revenue per user enough to offset the capital intensity. The upside case is that Muse becomes a primary interface for a meaningful share of Meta's 4 billion users, pushing subscription revenue into the tens of billions and supporting a sustained re-rating. The downside case is that trust concerns limit adoption to the free tier, capital spending stays elevated, and the stock falls back toward its historical price-to-earnings range.
Long term, the structural question is whether AI converts Meta from a cyclical advertising company into a hybrid platform with recurring revenue. If it does, the current valuation looks cheap even at $820. If it does not, the upgrade will be remembered as another premature bet on a technology transition that took longer to pay than investors hoped.
The cyclical-versus-structural call cuts both ways here, and it is worth separating the two legs rather than blending them. The cyclical leg is advertising: ad budgets will recover as the macro improves, and AI targeting should lift pricing power regardless of what happens with Muse. That part of the thesis is mean-reverting and relatively safe. The structural leg is the agent itself: if Muse becomes a primary interface for digital life, it is a regime change that will not revert on its own. The upgrade is priced as if the structural leg is already working. The evidence so far supports the cyclical leg; the structural leg is what the next two quarters must prove.
The market is being asked to believe that Meta's AI spend has finally found its payoff mechanism. The upgrade says the mechanism is visible; the next two earnings reports will say whether it is real.
Explore more exclusive insights at nextfin.ai.
