Meta (AI Infrastructure)
Not a public cloud; largest captive GPU fleet + MTIA silicon.
Meta runs one of the world's largest captive AI infrastructures to train Llama models and power recommendation/ranking, rather than selling cloud. It is among Nvidia's biggest customers (targeting well over a million GPU-equivalents) and designs its own MTIA inference accelerator. In 2025 Meta announced multi-hundred-billion-dollar capex plans and gigawatt-scale campuses (Prometheus, Hyperion) tied to nuclear and gas power deals.
Custom AI chip
MTIA (inference)
GPU fleet
>1M GPU-equivalents targeted
Mega campuses
Prometheus, Hyperion (GW-scale)
Models
Llama family
How it fits the stack
Meta (AI Infrastructure) with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
Meta (AI Infrastructure) in the AI stack. Meta (AI Infrastructure) with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 8 entities, 7 relationships
- Meta (AI Infrastructure) —depends on→ AMD
- Meta (AI Infrastructure) —uses→ Arm Holdings
- Meta (AI Infrastructure) —depends on→ Electricity Grid & Utilities
- Meta (AI Infrastructure) —uses→ Liquid Cooling
- Meta (AI Infrastructure) —depends on→ Nuclear Power (incl. SMRs)
- Meta (AI Infrastructure) —depends on→ Nvidia
- Meta (AI Infrastructure) —supplies→ Wiwynn
Depends on ↑ · 7