Meta Stock Is Still Too Cheap

Meta shares surged 11.4% Monday after Wells Fargo lifted its price target to $796 from $640, citing strong early traction for the Muse AI agent and its underlying Spark LLM. Muse reached No. 1 in Apple’s App Store and logged 600,000 downloads in its first five days; Meta stock is up 21% since the product launched two weeks ago. The article argues Meta retains valuation upside at a 25x P/E, with revenue projected to rise 26.5% to $254.1 billion this year before any contribution from Muse’s $20-$100 monthly premium tiers.

Analysis

The market is pricing an option on a new consumer revenue stream before the key unit economics are observable. App-store rank and early downloads are acquisition metrics, not evidence of durable engagement, paid conversion, inference cost per active user, or incremental ad yield; those four variables determine whether Muse expands META's earnings multiple or merely raises AI operating expense. After a sharp one-day repricing, the next 1-3 month catalyst is management disclosure on weekly retention, premium conversion, and whether agent usage improves click-through/conversion within Meta's existing ad stack.

The reported read-through to CPU vendors is likely too indiscriminate. At-scale agent inference is more likely to accrue to META's internal infrastructure and accelerator ecosystem—principally NVDA and AMD, with ARM benefiting only if on-device inference becomes a material product feature—than to INTC simply from higher agent usage. The more important competitive effect is on GOOG: a successful assistant embedded across Meta's social graph, messaging, commerce discovery, and advertiser data could raise switching costs and force higher AI-related traffic-acquisition and infrastructure spend across the consumer internet.

Consensus may be underestimating the strategic value of an agent that improves ad targeting and merchant conversion even if subscriptions remain niche; modest improvements to auction efficiency can monetize Meta's installed audience faster than a stand-alone consumer subscription model. Conversely, the valuation move is vulnerable if incremental inference costs exceed monetization, if retention normalizes after launch novelty, or if regulators characterize agent behavior/data use as an extension of Meta's existing privacy and youth-safety exposure. The thesis is falsified by weak retention or paid conversion at the next earnings call, a material upward revision to AI opex/capex without corresponding ad-performance commentary, or evidence that usage is cannibalizing rather than lifting core engagement.

Key Decisions for Investors

• Do not chase META after the gap; establish a 6-12 month long only on a pullback or after independently verified retention and monetization disclosure. Underwrite upside from ad-efficiency gains rather than subscription revenue; reduce if management raises 2027 AI spend materially without quantifying revenue or auction-performance benefit.
• Express the infrastructure read-through selectively: long NVDA or AMD versus short INTC over the next 1-3 months, sized modestly. The premise is that incremental agent inference favors established accelerators rather than generalized CPU demand; exit if Meta identifies CPUs as a meaningful production-inference architecture or if accelerator capex guidance weakens.
• Monitor a long META / short GOOG relative-value position after the next reported product metrics. It becomes actionable only if Muse shows sustained retention and Meta ties it to commerce or advertising conversion; the principal risk is Google's broader distribution advantage and faster Gemini monetization, which would compress the relative spread.
• Set an earnings-event alert rather than buying near-dated calls: require disclosure of weekly/monthly active users, paid conversion, revenue per user, and inference cost. Without those data, current enthusiasm is sentiment-driven and option premiums are likely to overprice a binary monetization reveal.