Meta Q3: Revenue $64.6B +26%, beats consensus. Adj EPS $7.40 vs $6.54E. GAAP operating income -37% is purely the $9.9B one-time legal charge — strip it and core operating income grew. AI capex $34.9B (YoY +80%), full-year guide raised to $135-145B. Q4 guide $73-76B.
Meta Muse Spark Open's open-weight release reshapes competitive dynamics in the AI model market. The performance gap between closed models (GPT-5, Claude 4, Gemini Ultra) and open-weight models effectively disappears with this 816B release. For enterprises, if performance is equivalent, licensing cost and data control become the deciding variables — and open weights solve both. Paradoxically, the structural beneficiaries are cloud infrastructure companies. Running open models on-premises requires GPU cloud or on-prem server investment. AWS, GCP, and Azure GPU instance demand grows with open-weight model proliferation. The interpretation that NVDA's short-term pullback reflects open-weight trends is incorrect; open weights expand compute demand.
Meta's Muse Spark Open 816B parameter release is an open-weight strategy at a scale inflection point. 42B active parameters in MoE architecture, 1M context, and 2-Mac-Studio operation means enterprises can self-host GPT-5-class models on their own infrastructure. Meta's strategy appears to sacrifice direct revenue, but the actual monetization has two tracks. First, establishing Llama-lineage models as the de facto industry standard, then monetizing via private fine-tuning services and cloud APIs. Second, ad AI enhancement: integrating 816B-scale models into Meta's internal ad targeting raises ad revenue per impression. There's also the free R&D effect of external researchers improving the open model. Meta's ad ROAS improvement post-Llama 2 open release in 2024 is the empirical precedent for this strategy.
Meta Q4 guide $73-76B (midpoint in-line with $73.7B consensus). Muse AI: 5M users. WhatsApp, Instagram, Threads — 3B+ user base as distribution for AI products is unmatched. Class C shares = Zuckerberg keeps voting control while raising capital.
Meta Q3 AI capex $34.9B — in a single quarter, up 80% from $19.4B a year ago. Annualized ~$140B. Full-year capex guidance raised from $115-135B to $135-145B. Big-4 combined ~$725B AI capex cycle keeps accelerating.
Meta Q3: Revenue $64.6B +26% YoY, beats $63.4B. Adj EPS $7.40 vs $6.54 est — 13% beat. GAAP operating income fell 37% but that's entirely the $9.9B one-time legal settlement. Core operating income $22.9B grew YoY.
Open weights means companies run it themselves — no direct revenue for Meta. But the developer ecosystem lock-in and ad AI advancement may be the bigger long-term play.
Meta Muse Spark Open: 816B params, 42B active — comparable to GPT-5? 1M context running on 2 Mac Studios is a massive deal for Mac users. Open weights changes everything.
