Meta AI — Company Report 2026

Abstract

Meta is executing the largest AI infrastructure capex cycle in corporate history. The company committed $135–145B in 2026 (up 88% from 2025) to build proprietary compute, train Llama models at scale, and embed AI across advertising and consumer products. Unlike competitors who monetize AI as a service, Meta is building compute for internal products first, then monetizing surplus capacity through Meta Compute. The thesis rests on three pillars: (1) Llama's 650M+ downloads create an open-source ecosystem that drives adoption and reduces Meta's own inference costs; (2) Meta AI's 1B+ monthly users across WhatsApp, Instagram, and Facebook provide distribution no standalone AI product can match; (3) AI-driven ad ranking (GEM, Andromeda, Lattice) is already delivering 8.3% click lift and 15.7% conversion uplift in Q2 2026. However, negative free cash flow ($2.1B in Q3) and unproven monetization paths for Meta Compute and Meta AI create material downside risk. This report evaluates Meta's competitive position, capex ROI timeline, regulatory exposure, and three scenarios for institutional investors.

Key findings

1. Meta's 2026 capex of $135–145B (52% of revenue) is 2.7x the historical norm; if sustained, this implies $1.3T spend over 10 years and negative FCF until at least 2028.
2. Meta AI reached 1B+ monthly active users by Q3 2025, with 60% growth in daily active users post-Muse Spark rebuild (July 2026); WhatsApp drives 35% of engagement.
3. Ad ranking AI delivered 8.3% more clicks and 15.7% conversion uplift in Q2 2026; advertisers paid 14% more per ad in Q3 2026 despite rising competition.
4. Llama has been downloaded 650M+ times as of June 2026; Meta's open-weights strategy commoditizes the model layer, reducing competitors' pricing power and Meta's own inference costs.
5. Meta Superintelligence Lab released Muse Spark (multimodal reasoning model) in April 2026; Muse agents launched September 2026; roadmap targets agentic AI on consumer platforms by 2027–2028.
6. Meta's capex-to-revenue ratio is 52% (vs. historical 20–30% norm); FCF was negative $2.1B in Q3 2026, marking the first sustained cash burn in Meta's public history.
7. Regulatory exposure includes FTC antitrust scrutiny on market power, EU data privacy questions on AI training sourcing, and calls for transparency in ad ranking algorithms.
8. Bull case assumes capex is one-time; ad ARPU accelerates; Meta Compute generates $10B+ revenue by 2028 (META $700+). Base case: capex moderates by 2027; $500–600 target. Bear case: capex persists; ad ROI plateaus; Llama commoditized (META $300–400).

Sources & methodology

Prepared by BR Research under BR Rating v1. 14 primary sources, accessed between 2026-11-14. Figures carry the date they were observed; nothing in this report is investment advice.

The full source list is part of the report.