Meta’s AI Bet
From Open Source to the OpenAI Playbook - Premium Report
When I analyzed Meta’s Q3 2025 results in November, the strategic logic seemed sound: front-load infrastructure capacity, bet on superintelligence arriving within 2-7 years, use distribution moats to achieve escape velocity.
Zuckerberg’s framework was clear—best case, AGI arrives in 2-3 years, and Meta is positioned; medium case, it takes 5-7 years and compute powers the core business; worst case, they slow building while growing into capacity.
What’s changed in six weeks? Everything and nothing.
A recent FT’s deep dive confirms what I suspected: the internal disorder beneath the confident earnings calls, the tensions between “Friends of Zuckerberg” and the new AI-native leadership, and the scramble to recover from Llama 4’s April failure.
The numbers we tracked—$70-72B 2025 capex, free cash flow compression from $54B to $20B, the acceleration curve reaching $32B quarterly by late 2026—these weren’t just financial projections. They were pressure gauges on an organization undergoing its most turbulent transformation in two decades.
