The Forcing Functions
1. Compute Scarcity
Everyone wants GPUs. Supply is constrained.
Wait = fall behind in AI race
2. Power Scarcity
Grid can’t keep up. Data centers need GWs.
Secure power now or lose the future
3. Dependency Trap
Meta depends on partners who are also competitors.
Build or be hostage
The Window of Opportunity
| Timeline | Milestone |
|---|---|
| 2025 | $72B CapEx (NOW) |
| 2027 | Infrastructure operational |
| 2030 | Full vertical integration |
| Future | Infrastructure moat locked |
The Zuckerberg Logic: Three Scenarios
Best Case (2-3 years)
Superintelligence arrives early → Meta ideally positioned
Medium Case (5-7 years)
AI progress steady, not explosive → Extra compute powers core biz
Worst Case (overbuilt)
AI winter / slower adoption → Slow building, grow into capacity
The Asymmetric Bet
If overbuild: Slow down, grow into capacity. Painful but survivable.
If underbuild: Miss the AI transition entirely. Existential.
The asymmetry favors building. That’s why $72B now, not later.
This is part of a comprehensive analysis. Read the full analysis on The Business Engineer.









