

Meta · v2 · seit 2023-07-18 · 13× · zuletzt 25. Sept. 2026
Llama 2 ist eine von Meta AI entwickelte Familie offener, vortrainierter und instruktionsfeinabgestimmter Sprachmodelle mit 7 bis 70 Milliarden Parametern. Die Modelle werden als Gewichte zum Download angeboten (kein proprietärer API-Betrieb durch Meta selbst) und unter einer eigenen Community-Lizenz veröffentlicht, die kommerzielle Nutzung mit Einschränkungen erlaubt. Llama 2 bietet gegenüber dem Vorgänger ein verdoppeltes Kontextfenster und wurde auf 40% mehr Daten trainiert; die Chat-Variante ist auf Dialoganwendungen optimiert.
Features
| Key-Benchmark (%) | MMLU 70B: 68,9% (offizielles Llama-2-Paper) |
| Kontextfenster (Token) | 4.096 Token |
| Lizenz | Llama 2 Community License (eigene Meta-Lizenz, kommerzielle Nutzung erlaubt, Einschränkung ab 700 Mio. aktiven Nutzern) |
| Multimodalität | Nein – reines Text-Modell (7B, 13B, 70B Parameter) |
| Plattform | Modellgewichte zum Download; verfügbar über Azure AI, AWS, Hugging Face; lokal ausführbar (u.a. Windows) |
| Release-Datum | 18. Juli 2023 |
Weitere Produkte in dieser Kategorie: Open-Source-Sprachmodelle
Belege (13)
Unternehmens-Analyse: Meta
Data from the most recent quarter supports a strong operating backdrop (Q3 2026 revenue +26% YoY; ad impressions +10% YoY), but near-term earnings quality is clouded by large legal charges and a structurally higher AI/infrastructure cost trajectory. With the stock trading around ~21x trailing earnings (Nov 17, 2026) and with ongoing regulatory/legal headline risk (including a reported Italian privacy fine in mid-November 2026), the risk-adjusted setup is balanced: attractive fundamentals, but limited tolerance for negative surprises. A HOLD is warranted pending clearer evidence that margin resilience can be maintained alongside AI scaling and that legal/regulatory costs do not recur at similar magnitude.
Zusammenfassung
Meta Platforms (META) operates a two-segment model: (1) Family of Apps (FoA)—Facebook, Instagram, WhatsApp, Messenger—monetized primarily via performance advertising; and (2) Reality Labs (RL)—AR/VR hardware, software, and AI-enabled wearables, currently loss-making but positioned as a longer-duration option on new computing platforms. Core competencies are global-scale consumer product distribution, ad targeting/measurement, and increasingly AI infrastructure and model deployment that improves ranking, recommendations, and ad performance. Market position remains top-tier in global digital advertising, with competitive advantages in user scale, engagement, and a broad advertiser base. In the last ~90 days, the most material company-specific datapoint is Meta’s Q3 2026 results and commentary: revenue of $64.60B (+26% YoY) with ad impressions up 10% YoY across FoA, indicating continued demand and effective monetization improvements. However, reported profitability was distorted by a large legal-related charge: Q3 2026 expenses included a $9.89B charge tied to an August 2026 multistate attorneys general agreement; Meta also highlighted rising infrastructure expenses and third-party AI token costs. Excluding the legal charge, Meta indicated operating income would have increased YoY, supporting the view that underlying operating leverage remains intact despite elevated AI-related cost growth. Reality Labs revenue was $443M (down 6% YoY) due to lower Quest headset sales, partially offset by growth in AI glasses revenue, reinforcing that RL is still early and volatile. Valuation/metrics (as of Nov 17, 2026): META last price is $626.41; using EUR/USD ~1.0997 (USD per EUR), this implies ~€569.62 per share. Trailing P/E is ~21.36x and TTM EPS ~29.32. Market cap is ~US$1.61T. Near-term investor focus is likely to remain on (a) durability of ad growth, (b) the pace of AI infrastructure spend and its margin impact, and (c) regulatory/legal overhangs (e.g., a recent Italian privacy fine reported in mid-November 2026). Outlook (short- to medium-term): data from the latest quarter supports strong top-line momentum and improving AI-driven monetization, but expense volatility (legal items) and structurally higher AI/infrastructure costs can create earnings variability. The base case is continued revenue growth into 2027 supported by AI improvements, with margins dependent on the slope of capex/opex and the frequency of regulatory/legal costs.
Key Takeaways
- Meta’s core earnings engine remains FoA advertising; Q3 2026 revenue grew 26% YoY to $64.60B with ad impressions up 10% YoY, indicating strong demand and monetization execution. (Q3 2026 release)
- Reported Q3 2026 profitability was heavily impacted by a $9.89B legal-related charge tied to an August 2026 multistate AG agreement; excluding this, Meta stated operating income would have increased YoY. (Q3 2026 release/call transcript)
- AI is a dual driver: it is improving ranking/recommendations and ad performance, but also increasing cost of revenue and infrastructure expenses (data centers, cloud, and third-party AI token costs). (SEC filing/Q3 materials)
- Reality Labs remains a drag and is still proving product-market fit: Q3 RL revenue was $443M (down 6% YoY) with weaker Quest sales, partly offset by AI glasses growth. (earnings call transcript)
- At ~21.36x trailing P/E (Nov 17, 2026), valuation is not “low,” so the investment case depends on sustained revenue growth and disciplined cost/capex execution rather than multiple expansion alone. (market data)
Action-Ideen
Accumulate on the view that FoA ad growth is being structurally supported by AI-driven ranking and measurement improvements. Q3 2026 showed 26% YoY revenue growth and 10% YoY ad impression growth, while management commentary indicates AI investments are benefiting the core business. The large Q3 legal charge is material but non-recurring in nature; normalized operating performance (as described by the company) was stronger than GAAP results suggest. Entry should be sized with awareness that AI infrastructure costs can pressure margins even in a strong revenue environment.
Horizont: 12 Mon.
Maintain exposure if already owned, emphasizing that the stock’s trailing valuation (~21x P/E) implies the market already prices in continued solid growth. The Q3 2026 print supports strong fundamentals, but the path of expenses (legal items plus AI-related opex/capex) can create quarter-to-quarter volatility. A hold stance is appropriate while monitoring whether revenue growth remains strong through 2027 and whether cost growth moderates as AI infrastructure scales.
Horizont: 6 Mon.
Reduce exposure if portfolio constraints require lower volatility or if the investment mandate cannot tolerate regulatory/legal headline risk. Recent enforcement actions (e.g., Italy privacy fine reported Nov 2026) highlight that compliance costs and product changes can be recurring. If AI-related costs continue to rise (infrastructure, cloud, token costs) without commensurate margin resilience, the risk-adjusted return profile may deteriorate even if revenue remains strong.
Horizont: 3 Mon.
Contrarian Insights
- • Consensus often frames Meta’s AI spend primarily as a growth catalyst; a more conservative read is that third-party AI token costs and infrastructure expenses can become a structurally higher cost base, limiting margin expansion even with strong revenue growth. (SEC filing commentary on cost drivers)
- • Reality Labs is widely viewed as a near-term drag; contrarian upside is that AI glasses revenue growth (noted in Q3 commentary) could indicate earlier-than-expected traction in wearables, potentially improving the narrative even before RL becomes profitable. (earnings call transcript)
Quellen (7)
- https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Third-Quarter-2026-Results/default.aspx
- https://www.sec.gov/Archives/edgar/data/1326801/000162828026066773/meta-20260930.htm
- https://s21.q4cdn.com/399680738/files/content_files/META-Q3-2026-Earnings-Call-Transcript.pdf
- https://apnews.com/article/9b3073c52464d5cd0e8facbc57285194
- https://finance.yahoo.com/quote/META/analysis/
- https://stockanalysis.com/stocks/meta/forecast/
- https://wise.com/us/currency-converter/eur-to-usd-rate/history/17-11-2026