Meta“练AI肌肉”：前沿大模型不是终点，产品化与投资见顶才是2026主线

U.S. Internet Meta Platforms Inc. RatingOutperform Price Target 870.00 USD META Meta Investment Memo: Pumping AI Iron It’s still January which means gyms are full of new year resolutioners, weather permitting. Inthe AI gym, where gains are measured in tokens and MMLU scores, Meta looks the scrawny kidin a gym full of bodybuilders named Claude, Gemini, GPT, and Grok. But don’t mistake size forstrength. What if Meta is the elite bodybuilder Anatoly in disguise? And isn’t the point to get inshape, not just look bigger? It’s an expensive gym membership, but “the greatest feeling youcan get is the pump”. Expect Meta to still be lifting in December. Close Date26 Jan 2026META Close Price (USD)672.36Price Target (USD)870.00Upside/(Downside)29%52-Week Range796.25/479.80SPX6,950.23FYEDecDiv Yield0.3%Market Cap (USD) (B)1,695.07EV (USD) (B)1,701.68 The frontier model barbell. If Meta’s new model scores at or near peers, amazing. Andif model performance falls well short with no path to close the gap, Meta can wind downinvestment and partner (see: Apple). The most likely and feared outcome, however, is themiddle where Meta’s new model lands “close but not quite,” driving another step up ininvestment with no guarantees they can catch the pack - this outcome is still favorable if... … Meta can get in shape.Model performance KPIs have anointed AI winners and losersfor now, but AI product adoption and monetization matter more than another basis pointimprovement on some LLM test. On this front,Meta’s productization capability may besecond to none.The company can ship products at scale across their owned channels, andwe expect a barrage of shipments this year — remember, Meta AI counts 1B+ monthly userspowered by an average model - what if the next model is good? An expensive gym but contract locked in?The ‘significant’ step up in expenses to cover AItalent, compute, and depreciation flow through is worrisome if it’s too high or returns remainuncertain. But there’s no terminal risk. Any indication around peak investment intensityshould drive upward revisions and stock re-rating - exactly what’s needed to retain AI talent. Investment Implications Backing the scrawny kid to flex their muscle in 2026.Remain Outperform PT $870. DETAILS THE 2026 META SET UP - LLAMA, LLAMA, META DRAMA 2025 was perhaps the most violent, temperamental, and emotional market we’ve seen in quite a while. From DeepSeek toTariffs to OpenAI (not covered) news flow pixie dust turned poison saw us crowning AI winners and losers simultaneouslyresulting in high torque stock swings. The commonality? Underwriting the unknowable. We’re all well aware that the market inthe near-term is a voting machine, but even still we’ve had more than a few senior PM and CIOs acknowledge that 2025 mayhave been their hardest year in the seat. Perhaps no better stock exemplifies this Market’s finickiness than Meta.Over the past 6 months, we’ve seen a ~15-turnswing between Meta and Google’s Forward PE multiples. 15x Turns! How do we explain that? Give Google credit.Of course some of that is warranted as Google managed to capture a regulatory clearing event, Geminimodel success and usage acceleration, TPU enterprise traction, and solid search performance all compressed in this 6-monthwindow. Google also found themselves as the antidote to numerous thematics: Convinced AI is a bubble? Head to Google.Believe in AI but not in the OpenAI complex? Google again. And if you simply ignore the AI debate but acknowledge a strong admarket with more ad dollars coming from publishers needing traffic, Google continues to deliver on search revenues. But thismemo is not about Google. From ‘AI winner’ to ‘AI winner for core’.From mid-2024, Meta was a consensus AI winner. ChatGPT wasn’t coming aftersocial, and if anyone could take advantage of AI for ad targeting, measurement, and creative… well that was Meta in a nutshell.And Meta absolutely delivered with AI-powered efforts including Lattice, GEM, and Andromeda on the ad delivery side andunifying the content recommendation algorithm across surfaces drove accelerating levels of ad revenues and engagement weno longer thought possible for a company of Meta’s size. Some of our previous estimates suggest thatMeta could be seeing a20-25% ROIC on AI investments going towards coreproducts. Revenue growth not a top-3 debate. For as long as we’ve been covering Meta, there’s always been some flavor of a revenueor engagement metric that was the singular KPI on the stock. We don’t expect this lull to last too long as there’s still room todebate what inning we’re in on the laundry list of AI-driven improvements and whether the return of TikTok, entry of ChatGPTinto the ad space, and changes to European operations could impact ad market share. Mostinvestors we speak with feelcomfortable in the ad growth story which remains the primary reason to own the stock here. No obvious terminal risk from Gen AI disruption.Amazingly, Meta’s de-rating oc